2.0.0 Alpha: Data Refinery

2026-08-08 21:31:56 +08:00
parent fa75081d4d
commit 562775e5db
48 changed files with 4172 additions and 661 deletions
@@ -47,6 +47,7 @@ coverage.xml

 local_settings.py
 db.sqlite3
+*.db.bak

 instance/

@@ -66,7 +67,9 @@ venv.bak/

 output/
 output_arena/
+database/backups/
+database/.pipeline.lock
 arena/
 scripts/
 experiment
-yrtv.zip
+yrtv.zip
@@ -0,0 +1,32 @@
+PYTHON := .venv/bin/python
+
+.PHONY: install run test check l1 l2 l3 l3-all pipeline
+
+install:
+	python3 -m venv .venv
+	$(PYTHON) -m pip install -r requirements.txt
+
+run:
+	$(PYTHON) -m web.app
+
+test:
+	$(PYTHON) -m unittest discover -v
+
+check:
+	$(PYTHON) -m compileall -q web database tests wsgi.py
+	$(PYTHON) -m unittest discover -v
+
+l1:
+	$(PYTHON) database/L1/L1_Builder.py
+
+l2:
+	$(PYTHON) database/L2/L2_Builder.py
+
+l3:
+	$(PYTHON) database/L3/L3_Builder.py
+
+l3-all:
+	$(PYTHON) database/L3/L3_Builder.py --force
+
+pipeline:
+	$(PYTHON) -c "from web.app import create_app; create_app(); from database.job_store import JobStore; from database.pipeline import run_pipeline; job_id = JobStore().create_job('manual_pipeline', created_by='cli'); print('job_id=', job_id); raise SystemExit(0 if run_pipeline(job_id) else 1)"
@@ -1,154 +1,269 @@
-# YRTV 项目说明 till 1.0.2hotfix
+# YRTV 2.0.0 Alpha

-## 项目概览
-YRTV 是一个基于 CS2 比赛数据的综合分析与战队管理平台。它集成了数据采集、ETL 清洗建模、特征挖掘以及现代化的 Web 交互界面。
-核心目标是为战队提供数据驱动的决策支持，包括战术分析、队员表现评估、阵容管理（Clubhouse）以及实时战术板功能。
+YRTV 是面向固定 CS2 战队的私人数据站。它不是公共玩家排行榜，而是让队员拥有类似职业选手的个人主页，并为战队提供比赛档案、队内比较、阵容分析、对手情报和战术工具。

---
+当前 Alpha 版本已经建立可重复运行的数据流水线、数据库治理和职业主页数据集市，重点服务 active roster。

-您可以使用以下命令快速配置环境：
-pip install -r requirements.txt
+## 当前基线

-数据来源与处理核心包括：
- 比赛页面的 iframe JSON 数据（`iframe_network.json`）
- 可选的 demo 文件（`.zip/.dem`）
- L1A/L2/L3 分层数据库建模与校验
+- 208 场比赛
+- 1,181 名采集到的玩家
+- 2,080 条玩家比赛记录
+- 4,315 个回合
+- 33,560 条回合事件
+- 38,423 条经济记录
+- 9 名 active roster 队员
+- 885 条 roster 逐场历史
+- 76 条地图统计
+- 236 条武器统计
+- 54 条时间窗口统计
+- 54 条个人职业纪录
+- 24 项自动化测试通过
+- 28 项数据完整性检查通过

-## v3.0.0 Release 更新要点
- **核心算法升级**: 严格确立 Active Roster (Lineup 1) 为战队平均数据计算基准，修复了雷达图与平均数据的计算偏差。
- **Clubhouse 增强**: 
-    - 布局优化为 3 列网格。
-    - 新增 **OVR (Overall Score)** 显示，优先展示真实评分 (Real Rating)，直观反映选手综合实力。
- **Tactics 系统**: 
-    - 统一评分逻辑：全站优先采用 L3 `core_avg_rating2` (真实评分)，智能回退至 `basic_avg_rating`。
-    - Data Center 数据中心现在完整映射了 Utility、Trading 等高阶战术数据。
- **稳定性修复**: 修正了特征服务中的语法错误，增强了对缺失数据的鲁棒性处理。
+数据规模会随导入变化，Admin 数据完整性中心显示的结果是运行时事实。

-## Web 交互系统 (Core)
-基于 Flask + TailwindCSS + Alpine.js 构建的现代化 Web 应用。
+## 核心功能

-### 核心功能模块
-1.  **Clubhouse (战队管理)**
-    -   **Roster Management**: 拖拽式管理当前激活阵容 (Active Roster)。
-    -   **Scout System**: 全库模糊搜索玩家，支持按 Rating/Matches/KD 排序筛选。
-    -   **Contract System**: 模拟签约/解约流程 (Sign/Release)，管理战队资产。
-    -   **Identity**: 统一的头像与 ID 显示逻辑 (SteamID/Name)，支持自动生成首字母头像。
+### 玩家职业主页

-2.  **Tactics Board (战术终端)**
-    -   **SPA 架构**: 基于 Alpine.js 的单页应用，无刷新切换四大功能区。
-    -   **Board (战术板)**: 集成 Leaflet.js 的交互式地图，支持战术点位标记。
-    -   **Data (数据中心)**: 实时查看全队近期数据表现，集成 Utility/Trading 等高阶战术指标。
-    -   **Analysis (深度分析)**: 
-        -   **Chemistry**: 任意组合 (2-5人) 的共同比赛胜率与数据分析。
-        -   **Depth**: 阵容深度与位置分析。
-    -   **Economy (经济计算)**: 简单的经济局/长枪局计算器。
+- Rating、K/D、ADR、KAST 等生涯数据
+- Aim、Clutch、Pistol、Defense、Utility、Stability、Economy、Pace 八维能力
+- 生涯、最近 10/20/30 场、最近 30/90 天阶段统计
+- 可切换时间窗口的 Rating 趋势
+- 最高 Rating、最多击杀、最高 ADR、最高 K/D、最多爆头和最长连胜
+- 每项个人纪录可追溯到具体比赛
+- 地图表现、比赛历史、Party 信息、队内排名和留言板
+- 缺失或尚未实现的指标显示为 `N/A`，不使用伪造分数

-3.  **Match Center (比赛中心)**
-    -   **List View**: 
-        -   显示比赛平均 ELO。
-        -   **Party Identification**: 自动识别组排车队 (👥 2-5)，并用颜色区分规模 (Indigo/Blue/Purple/Orange)。
-        -   **Result Tracking**: 基于 "Our Team" (Active Roster) 的胜负判定 (VICTORY/DEFEAT/CIVIL WAR)。
-    -   **Detail View**:
-        -   按 Rating 降序排列双方队员。
-        -   高亮显示组排关系。
-        -   集成 Round-by-Round 经济与事件详情。
+### 比赛中心

-4.  **Player Profile (玩家档案)**
-    -   综合能力雷达图 (八维数据: Aim, Clutch, Pistol, Defense, Util, Stability, Economy, Pace)。
-    -   近期 Rating/KD/ADR 趋势折线图。
-    -   详细的历史比赛记录（含 Party info 与 Result）。
-    -   头像上传与管理。
+- 比赛列表、地图、比分、平均 ELO 和己方结果
+- Active roster 与 Party 识别
+- 双方玩家表现和 Rating 排序
+- Head-to-head 击杀矩阵
+- 回合事件、经济和装备信息
+- 原始比赛数据查看

-## 自动化与运维
-新增 `ETL/refresh.py` 自动化脚本，用于一键执行全量数据刷新：
- 自动清理旧数据库。
- 顺序执行 L1A -> L2 -> L3 构建。
- 自动处理 schema 迁移。
+### 战队与战术

-## 数据流程
-1. **下载与落盘**  
-   通过 `downloader/downloader.py` 抓取比赛页面数据，生成 `output_arena/<match_id>/iframe_network.json`，并可同时下载 demo 文件。
-2. **L1A 入库（原始 JSON）**  
-   `ETL/L1A.py` 将 `output_arena/*/iframe_network.json` 批量写入 `database/L1A/L1A.sqlite`。
-3. **L2 入库（结构化事实表/维度表）**  
-   `ETL/L2_Builder.py` 读取 L1A 数据，按 `database/L2/schema.sql` 构建维度表与事实表，生成 `database/L2/L2_Main.sqlite`。
-4. **L3 入库（特征集市）**
-   `ETL/L3_Builder.py` 读取 L2 数据，计算 Basic 及 6 大挖掘能力维度特征，生成 `database/L3/L3_Features.sqlite`。
-5. **质量校验与覆盖分析**  
-   `ETL/verify/verify_L2.py` 与 `ETL/verify/verify_deep.py` 用于 L2 字段覆盖与逻辑检查。
+- Active roster 管理
+- 玩家搜索、签入和移出
+- 2-5 人同队比赛与 Chemistry 分析
+- 对手档案和真实交手记录
+- 地图战术板、阵容数据中心和经济工具
+- Wiki、玩家标签、备注和评论
+
+### 数据运营
+
+- Admin 上传 `iframe_network.json`
+- 自动识别唯一 `g161-*` 比赛 ID
+- JSON 结构、必要接口、哈希和重复比赛校验
+- 后台执行 L1 → L2 → L3
+- 实时查看作业阶段、进度、日志和耗时
+- 数据完整性中心与 JSON 报告
+
+## 快速开始
+
+环境要求：
+
+- macOS/Linux
+- Python 3.9+
+- SQLite 3
+
+安装并启动：
+
+```bash
+make install
+
+export SECRET_KEY='replace-with-a-random-secret'
+export ADMIN_TOKEN='replace-with-an-admin-token'
+
+make run
+```
+
+默认地址：
+
+- 应用：`http://127.0.0.1:5000`
+- Admin：`/admin/`
+- 比赛导入：`/admin/import-match`
+- 数据完整性：`/admin/data-integrity`
+
+生产进程入口：
+
+```bash
+.venv/bin/gunicorn wsgi:app
+```
+
+## 常用命令
+
+```bash
+make run       # 启动 Flask
+make check     # 编译检查 + 自动化测试
+make pipeline  # 备份后执行完整 L1 -> L2 -> L3
+
+make l1        # 仅构建 L1
+make l2        # 仅构建 L2
+make l3        # 仅构建 active roster L3
+make l3-all    # 为全部采集玩家构建 L3
+```
+
+正常维护优先使用 `make pipeline`。单层命令主要用于开发和排错。
+
+## 比赛导入
+
+推荐从 Admin 页面上传完整的 `iframe_network.json`。
+
+导入流程：
+
+1. 验证 UTF-8 和 JSON 结构。
+2. 从网络 URL 中提取唯一比赛 ID。
+3. 检查 match 和 round 必要接口。
+4. 计算 SHA256，拒绝相同数据重复导入。
+5. 保存到 `output_arena/<match_id>/iframe_network.json`。
+6. 创建 `etl_jobs` 作业。
+7. 备份 L1/L2/L3。
+8. 串行执行三个 Builder。
+9. 验证目标比赛具有 10 名玩家和回合事实。
+10. 成功提交；失败自动恢复备份。
+
+仓库当前不包含自动访问 5E 网页的下载器，因此首页 URL 输入不会抓取数据。
+
+## 数据架构
+
+```text
+iframe_network.json
+        |
+        v
+L1 raw capture
+        |
+        v
+L2 normalized facts
+        |
+        v
+L3 roster features and profile marts
+        |
+        v
+Flask services and player profiles
+```
+
+### L1：原始层
+
+- 数据库：`database/L1/L1.db`
+- Builder：`database/L1/L1_Builder.py`
+- Grain：每场比赛一份完整网络抓包
+- 核心表：`raw_iframe_network`
+
+### L2：事实层
+
+- 数据库：`database/L2/L2.db`
+- Schema：`database/L2/schema.sql`
+- Builder：`database/L2/L2_Builder.py`
+- 核心表：
+  - `dim_players`
+  - `dim_maps`
+  - `fact_matches`
+  - `fact_match_teams`
+  - `fact_match_players`
+  - `fact_match_players_t`
+  - `fact_match_players_ct`
+  - `fact_rounds`
+  - `fact_round_events`
+  - `fact_round_player_economy`
+
+### L3：特征与主页集市
+
+- 数据库：`database/L3/L3.db`
+- Schema：`database/L3/schema.sql`
+- Builder：`database/L3/L3_Builder.py`
+- 核心表：
+  - `dm_player_features`
+  - `dm_player_match_history`
+  - `dm_player_map_stats`
+  - `dm_player_weapon_stats`
+  - `dm_player_period_stats`
+  - `dm_player_records`
+
+### Web：应用状态
+
+- 数据库：`database/Web/Web_App.sqlite`
+- Schema：`database/Web/schema.sql`
+- 当前 schema version：2
+- 保存 lineup、玩家备注、评论、Wiki、战术板、导入登记和 ETL 作业
+
+## 数据库治理
+
+- 所有运行路径集中定义在 `database/paths.py`
+- 完整编排入口为 `database/pipeline.py`
+- 同一时间只允许一个 pipeline
+- Pipeline 运行前备份 L1/L2/L3
+- 失败时恢复三层数据库，Web 作业日志继续保留
+- 备份位于 `database/backups/`
+- 自动保留最近 3 组备份
+- Web schema 使用 `schema_migrations` 记录版本
+- 高频玩家历史、Party、事件和经济查询具有专用索引
+- 数据库和目录规则详见 `database/README.md`
+
+## 数据质量
+
+Admin 数据完整性中心检查：
+
+- 四个 SQLite 数据库的 `quick_check`
+- 必要表和 Web schema version
+- 玩家比赛、回合事件的引用完整性
+- 每场比赛玩家数量
+- 玩家身份覆盖
+- 高频查询索引
+- Active roster 的 L3 特征覆盖
+- 逐场历史与总场次一致性
+- 真实队内 percentile
+- 地图、武器、时间窗口和职业纪录集市
+- 占位空间指标
+- Web 外键、active lineup 和 pipeline 并发
+- 备份数量与存储规模
+
+运行测试：
+
+```bash
+make check
+```

 ## 目录结构
-```
+
+```text
 yrtv/
-├── downloader/                 # 下载器（抓取 iframe JSON 与 demo）
-├── ETL/                        # ETL 脚本
-│   ├── L1A.py
-│   ├── L2_Builder.py
-│   ├── L3_Builder.py
-│   ├── refresh.py              # [NEW] 一键刷新脚本
-│   └── verify/
-├── database/                   # SQLite 数据库存储
-│   ├── L1A/                    
-│   ├── L2/                     
-│   ├── L3/                     
-│   └── original_json_schema/   
-├── web/                        # [NEW] Web 应用程序
-│   ├── app.py                  # 应用入口
-│   ├── routes/                 # 路由 (matches, players, teams, tactics)
-│   ├── services/               # 业务逻辑 (stats, web)
-│   ├── templates/              # Jinja2 模板 (TailwindCSS + Alpine.js)
-│   └── static/                 # 静态资源 (CSS, JS, Uploads)
-└── utils/
-    └── json_extractor/         # JSON Schema 抽取工具
+├── database/
+│   ├── L1/                 # 原始抓包与 Builder
+│   ├── L2/                 # 事实层、Schema、Processor
+│   ├── L3/                 # 特征层、Schema、Processor
+│   ├── Web/                # 应用数据库 Schema
+│   ├── paths.py            # 统一路径
+│   ├── maintenance.py      # 备份、恢复、健康检查
+│   ├── job_store.py        # ETL 作业状态
+│   └── pipeline.py         # 完整流水线
+├── tests/                  # 自动化测试
+├── utils/                  # JSON 结构分析工具
+├── web/
+│   ├── routes/
+│   ├── services/
+│   ├── templates/
+│   └── static/
+├── Makefile
+├── requirements.txt
+└── wsgi.py
 ```

-## 环境要求
- Python 3.11.4+
- Flask, Jinja2
- Playwright（下载器依赖）
- pandas, numpy（数据处理依赖）
+## Alpha 限制

-## 数据库层级说明
-### L1A
- **用途**：保存原始 iframe JSON
- **输入**：`output_arena/*/iframe_network.json`
- **输出**：`database/L1A/L1A.sqlite`
- **脚本**：`ETL/L1A.py`
+- 当前主要数据源为 5E iframe 网络响应。
+- 不包含自动网页下载器和 Demo parser。
+- 认证仍是单一 Admin Token，适合私人部署，不适合开放注册。
+- SQLite 适合当前单战队规模，不面向高并发多租户。
+- 部分高级空间能力需要地图边界、路径和 Demo 数据，当前显示 `N/A`。
+- `StatsService` 仍保留部分兼容逻辑，后续会继续按领域拆分。

-### L1B
- **用途**：保存 demo 解析后的原始数据（由 demoparser2 产出）
- **输出**：`database/L1B/L1B.sqlite`
- 当前仓库提供目录与说明，解析流程需结合外部工具执行
+## 版本

-### L2
-结构化事实表/维度表数据库，覆盖比赛、玩家、回合与经济等数据：
- **Schema**：`database/L2/schema.sql`
- **输出**：`database/L2/L2_Main.sqlite`
- **核心表**：
-  - `dim_players`、`dim_maps`
-  - `fact_matches`、`fact_match_teams`
-  - `fact_match_players`、`fact_match_players_t`、`fact_match_players_ct`
-  - `fact_rounds`、`fact_round_events`、`fact_round_player_economy`
+当前版本：`2.0.0 Alpha`

-### L3
-玩家特征集市 (Player Features Data Mart)，聚合 Basic 及 6 大挖掘能力维度 (STA, BAT, HPS, PTL, T/CT, UTIL)。
- **Schema**：`database/L3/schema.sql`
- **输出**：`database/L3/L3_Features.sqlite`
- **脚本**：`ETL/L3_Builder.py`
- **核心表**：`dm_player_features` (玩家聚合画像)
-
-## JSON Schema 抽取工具
-用于分析大量 `iframe_network.json` 的字段结构与覆盖情况，支持动态 Key 归并与多格式输出。
-
-输出内容通常位于 `output_reports/` 或 `database/original_json_schema/`，包括：
- `schema_summary.md`：结构概览
- `schema_flat.csv`：扁平字段列表
- `uncovered_features.csv`：未覆盖字段清单
-
-## 数据源互斥说明
-L2 中 `fact_matches.data_source_type` 用于区分数据来源与字段覆盖范围：
- `classic`：含 round_list 详细回合与坐标信息
- `leetify`：含 leetify 评分与经济信息
- `unknown`：无法识别来源
-
-入库逻辑保持互斥：同一场比赛只会按其来源覆盖相应字段，避免重复或冲突。
+这一版本的目标是建立可信、可恢复、可持续导入的私人战队 HLTV 基线，而不是冻结产品功能。
@@ -14,13 +14,18 @@ import os
 import json
 import sqlite3
 import glob
-import argparse  # Added
+import argparse
+import sys

-# Paths
 BASE_DIR = os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
-OUTPUT_ARENA_DIR = os.path.join(BASE_DIR, 'output_arena')
-DB_DIR = os.path.join(BASE_DIR, 'database', 'L1')
-DB_PATH = os.path.join(DB_DIR, 'L1.db')
+if BASE_DIR not in sys.path:
+    sys.path.insert(0, BASE_DIR)
+
+from database.paths import L1_DB, L1_DIR, OUTPUT_ARENA
+
+OUTPUT_ARENA_DIR = str(OUTPUT_ARENA)
+DB_DIR = str(L1_DIR)
+DB_PATH = str(L1_DB)

 def init_db():
    if not os.path.exists(DB_DIR):
@@ -65,6 +70,7 @@ def process_files():
    
    count = 0
    skipped = 0
+    errors = 0
    
    for file_path in files:
        try:
@@ -92,11 +98,14 @@ def process_files():
                conn.commit()
                
        except Exception as e:
+            errors += 1
            print(f"Error processing {file_path}: {e}")
            
    conn.commit()
    conn.close()
-    print(f"Finished. Processed: {count}, Skipped: {skipped}.")
+    print(f"Finished. Processed: {count}, Skipped: {skipped}, Errors: {errors}.")
+    if errors:
+        raise RuntimeError(f"L1 ingestion failed for {errors} file(s)")

 if __name__ == '__main__':
-    process_files()
+    process_files()
@@ -1,16 +1,25 @@
-L1A 5eplay平台网页爬虫原始数据。
+# L1 Raw Match Store

-## ETL Step 1:
-从原始json数据库提取到L1A级数据库中。
-`output_arena/*/iframe_network.json` -> `database/L1A/L1A.sqlite`
+L1 stores one complete 5E network capture per match without transforming its
+payload.

-### 脚本说明
- **脚本位置**: `ETL/L1A.py`
- **功能**: 自动遍历 `output_arena` 目录下所有的 `iframe_network.json` 文件，提取原始内容并以 `match_id` (文件夹名) 为主键存入 `L1A.sqlite` 数据库的 `raw_iframe_network` 表中。
+## Runtime Files

-### 运行方式
-使用项目指定的 Python 环境运行脚本：
+- Database: `database/L1/L1.db`
+- Builder: `database/L1/L1_Builder.py`
+- Input: `output_arena/<match_id>/iframe_network.json`
+- Primary key: `raw_iframe_network.match_id`
+
+## Commands

 ```bash
-C:/ProgramData/anaconda3/python.exe ETL/L1A.py
+make l1
+make pipeline
 ```
+
+Normal ingestion is incremental. `--force` re-reads every capture currently
+present in `output_arena`.
+
+`L1A.db` and the historical `database/L1A/L1A.sqlite` path are retired. L1B is
+reserved for a future demo-parser source and is not part of the runtime
+pipeline.
@@ -7,14 +7,20 @@ from dataclasses import dataclass, field
 from typing import List, Dict, Optional, Any, Tuple
 from datetime import datetime

+PROJECT_ROOT = os.path.dirname(os.path.dirname(os.path.dirname(os.path.abspath(__file__))))
+if PROJECT_ROOT not in sys.path:
+    sys.path.insert(0, PROJECT_ROOT)
+
+from database.paths import L1_DB, L2_DB, L2_SCHEMA
+
 # Setup logging
 logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s')
 logger = logging.getLogger(__name__)

 # Constants
-L1A_DB_PATH = 'database/L1/L1.db'
-L2_DB_PATH = 'database/L2/L2.db'
-SCHEMA_PATH = 'database/L2/schema.sql'
+L1A_DB_PATH = str(L1_DB)
+L2_DB_PATH = str(L2_DB)
+SCHEMA_PATH = str(L2_SCHEMA)

 # --- Data Structures for Unification ---

@@ -1238,6 +1244,10 @@ def process_matches():
    l1_conn.close()
    l2_conn.close()
    logger.info(f"\nDone. Processed {count} matches ({success_count} success, {error_count} errors).")
+    if error_count:
+        raise RuntimeError(
+            f"L2 build failed for {error_count}/{count} matches"
+        )

 if __name__ == "__main__":
    process_matches()
@@ -636,3 +636,24 @@ SELECT
 FROM fact_match_players fmp
 JOIN fact_matches fm ON fmp.match_id = fm.match_id
 GROUP BY fmp.steam_id_64, fm.map_name;
+
+-- ==========================================
+-- Operational query indexes
+-- ==========================================
+CREATE INDEX IF NOT EXISTS idx_match_players_player_match
+ON fact_match_players(steam_id_64, match_id);
+
+CREATE INDEX IF NOT EXISTS idx_match_players_match_team
+ON fact_match_players(match_id, team_id, steam_id_64);
+
+CREATE INDEX IF NOT EXISTS idx_match_players_party
+ON fact_match_players(match_id, match_team_id, steam_id_64);
+
+CREATE INDEX IF NOT EXISTS idx_round_events_victim
+ON fact_round_events(victim_steam_id, match_id);
+
+CREATE INDEX IF NOT EXISTS idx_economy_player_match
+ON fact_round_player_economy(steam_id_64, match_id, round_num);
+
+CREATE INDEX IF NOT EXISTS idx_matches_map_time
+ON fact_matches(map_name, start_time DESC);
@@ -6,6 +6,8 @@ import sqlite3
 import json
 import argparse
 import concurrent.futures
+from collections import defaultdict, deque
+from typing import Optional

 # Setup logging
 logging.basicConfig(level=logging.INFO, format='%(asctime)s - %(levelname)s - %(message)s')
@@ -15,10 +17,14 @@ logger = logging.getLogger(__name__)
 BASE_DIR = os.path.dirname(os.path.dirname(os.path.abspath(__file__)))  # Points to database/ directory
 PROJECT_ROOT = os.path.dirname(BASE_DIR)  # Points to project root
 sys.path.insert(0, PROJECT_ROOT)  # Add project root to Python path
-L2_DB_PATH = os.path.join(BASE_DIR, 'L2', 'L2.db')
-L3_DB_PATH = os.path.join(BASE_DIR, 'L3', 'L3.db')
-WEB_DB_PATH = os.path.join(BASE_DIR, 'Web', 'Web_App.sqlite')
-SCHEMA_PATH = os.path.join(BASE_DIR, 'L3', 'schema.sql')
+
+from database.paths import L2_DB, L3_DB, L3_SCHEMA, WEB_DB
+
+L2_DB_PATH = str(L2_DB)
+L3_DB_PATH = str(L3_DB)
+L3_BACKUP_PATH = f"{L3_DB_PATH}.bak"
+WEB_DB_PATH = str(WEB_DB)
+SCHEMA_PATH = str(L3_SCHEMA)

 def _get_existing_columns(conn, table_name):
    cur = conn.execute(f"PRAGMA table_info({table_name})")
@@ -76,7 +82,28 @@ def _get_team_players():
    try:
        conn = sqlite3.connect(WEB_DB_PATH)
        cursor = conn.cursor()
-        cursor.execute("SELECT player_ids_json FROM team_lineups")
+        columns = {
+            row[1] for row in cursor.execute("PRAGMA table_info(team_lineups)")
+        }
+        if 'is_active' in columns:
+            cursor.execute(
+                """
+                SELECT player_ids_json
+                FROM team_lineups
+                WHERE is_active = 1
+                ORDER BY created_at DESC, id DESC
+                LIMIT 1
+                """
+            )
+        else:
+            cursor.execute(
+                """
+                SELECT player_ids_json
+                FROM team_lineups
+                ORDER BY created_at DESC, id DESC
+                LIMIT 1
+                """
+            )
        rows = cursor.fetchall()
        
        steam_ids = set()
@@ -150,7 +177,25 @@ def _build_player_record(steam_id: str):
            "error": str(e),
        }

-def main(force_all: bool = False, workers: int = 1):
+def _backup_l3_database(source_path=L3_DB_PATH, backup_path=L3_BACKUP_PATH):
+    if not os.path.exists(source_path):
+        return None
+
+    source = sqlite3.connect(source_path)
+    backup = sqlite3.connect(backup_path)
+    try:
+        source.backup(backup)
+        result = backup.execute("PRAGMA quick_check").fetchone()[0]
+        if result != 'ok':
+            raise RuntimeError(f"L3 backup quick_check failed: {result}")
+    finally:
+        source.close()
+        backup.close()
+    logger.info("L3 backup created at %s", backup_path)
+    return backup_path
+
+
+def main(force_all: bool = False, workers: int = 1, create_backup: bool = True):
    """
    Main L3 feature building pipeline using modular processors
    """
@@ -158,6 +203,9 @@ def main(force_all: bool = False, workers: int = 1):
    logger.info("Starting L3 Builder with 5-Tier Architecture")
    logger.info("========================================")
    
+    if create_backup:
+        _backup_l3_database()
+
    # 1. Ensure Schema is up to date
    init_db()
    
@@ -181,6 +229,7 @@ def main(force_all: bool = False, workers: int = 1):
    conn_l3 = sqlite3.connect(L3_DB_PATH)
    
    try:
+        conn_l3.execute("BEGIN IMMEDIATE")
        cursor_l2 = conn_l2.cursor()
        if force_all:
            logger.info("Force mode enabled: building L3 for all players in L2.")
@@ -240,7 +289,6 @@ def main(force_all: bool = False, workers: int = 1):
                        )
                        success_count += 1
                    if processed_count % 2 == 0:
-                        conn_l3.commit()
                        logger.info(f"Progress: {processed_count}/{total_players} ({success_count} success, {error_count} errors)")
        else:
            for idx, row in enumerate(players, 1):
@@ -268,10 +316,20 @@ def main(force_all: bool = False, workers: int = 1):

                processed_count = idx
                if processed_count % 2 == 0:
-                    conn_l3.commit()
                    logger.info(f"Progress: {processed_count}/{total_players} ({success_count} success, {error_count} errors)")
-        
-        # Final commit
+
+        if error_count:
+            raise RuntimeError(
+                f"L3 feature build failed for {error_count}/{total_players} players"
+            )
+
+        processed_ids = [str(row[0]) for row in players]
+        _update_percentiles(conn_l3, processed_ids)
+        _rebuild_auxiliary_marts(conn_l2, conn_l3, processed_ids)
+
+        quick_check = conn_l3.execute("PRAGMA quick_check").fetchone()[0]
+        if quick_check != 'ok':
+            raise RuntimeError(f"L3 quick_check failed before commit: {quick_check}")
        conn_l3.commit()
        
        logger.info("========================================")
@@ -283,9 +341,11 @@ def main(force_all: bool = False, workers: int = 1):
        logger.info("========================================")
        
    except Exception as e:
+        conn_l3.rollback()
        logger.error(f"Fatal error during L3 build: {e}")
        import traceback
        traceback.print_exc()
+        raise
        
    finally:
        conn_l2.close()
@@ -313,7 +373,7 @@ def _get_round_count(steam_id: str, conn_l2: sqlite3.Connection) -> int:


 def _upsert_features(conn_l3: sqlite3.Connection, steam_id: str, features: dict,
-                     match_count: int, round_count: int, conn_l2: sqlite3.Connection | None,
+                     match_count: int, round_count: int, conn_l2: Optional[sqlite3.Connection],
                     first_match_date=None, last_match_date=None):
    """
    Insert or update player features in dm_player_features
@@ -353,12 +413,430 @@ def _upsert_features(conn_l3: sqlite3.Connection, steam_id: str, features: dict,
    
    cursor_l3.execute(sql, values)

+
+def _rebuild_auxiliary_marts(conn_l2, conn_l3, steam_ids):
+    """Rebuild player-grain marts used by profiles and trend APIs."""
+    if not steam_ids:
+        return
+
+    logger.info("Rebuilding L3 match, map and weapon marts")
+    total_history = 0
+    total_maps = 0
+    total_weapons = 0
+    total_periods = 0
+    total_records = 0
+
+    for start in range(0, len(steam_ids), 400):
+        chunk = steam_ids[start:start + 400]
+        placeholders = ','.join('?' for _ in chunk)
+
+        for table in (
+            'dm_player_match_history',
+            'dm_player_map_stats',
+            'dm_player_weapon_stats',
+            'dm_player_period_stats',
+            'dm_player_records',
+        ):
+            conn_l3.execute(
+                f"DELETE FROM {table} WHERE steam_id_64 IN ({placeholders})",
+                chunk,
+            )
+
+        history_rows = conn_l2.execute(
+            f"""
+            SELECT
+                mp.steam_id_64,
+                mp.match_id,
+                m.start_time,
+                mp.rating,
+                mp.kd_ratio,
+                mp.adr,
+                mp.kast,
+                mp.is_win,
+                m.map_name,
+                mp.kills,
+                mp.deaths,
+                mp.headshot_count,
+                (
+                    SELECT AVG(teammate.rating)
+                    FROM fact_match_players teammate
+                    WHERE teammate.match_id = mp.match_id
+                      AND teammate.team_id = mp.team_id
+                      AND teammate.steam_id_64 != mp.steam_id_64
+                ) AS teammate_avg_rating
+            FROM fact_match_players mp
+            JOIN fact_matches m ON m.match_id = mp.match_id
+            WHERE mp.steam_id_64 IN ({placeholders})
+            ORDER BY mp.steam_id_64, m.start_time, mp.match_id
+            """,
+            chunk,
+        ).fetchall()
+
+        history_values = []
+        player_state = defaultdict(lambda: {
+            'sequence': 0,
+            'rating_sum': 0.0,
+            'recent': deque(maxlen=10),
+        })
+        for row in history_rows:
+            steam_id = str(row[0])
+            state = player_state[steam_id]
+            rating = float(row[3] or 0.0)
+            state['sequence'] += 1
+            state['rating_sum'] += rating
+            state['recent'].append(rating)
+            history_values.append((
+                steam_id,
+                row[1],
+                row[2],
+                state['sequence'],
+                row[3],
+                row[4],
+                row[5],
+                row[6],
+                row[7],
+                row[8],
+                None,
+                row[12],
+                state['rating_sum'] / state['sequence'],
+                sum(state['recent']) / len(state['recent']),
+            ))
+
+        conn_l3.executemany(
+            """
+            INSERT INTO dm_player_match_history (
+                steam_id_64, match_id, match_date, match_sequence,
+                rating, kd_ratio, adr, kast, is_win, map_name,
+                opponent_avg_elo, teammate_avg_rating,
+                cumulative_rating, rolling_10_rating
+            ) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)
+            """,
+            history_values,
+        )
+        total_history += len(history_values)
+
+        map_rows = conn_l2.execute(
+            f"""
+            SELECT
+                mp.steam_id_64,
+                m.map_name,
+                COUNT(*) AS matches,
+                SUM(CASE WHEN mp.is_win = 1 THEN 1 ELSE 0 END) AS wins,
+                AVG(mp.rating) AS avg_rating,
+                AVG(mp.kd_ratio) AS avg_kd,
+                AVG(mp.adr) AS avg_adr,
+                AVG(mp.kast) AS avg_kast,
+                MAX(mp.rating) AS best_rating,
+                MIN(mp.rating) AS worst_rating
+            FROM fact_match_players mp
+            JOIN fact_matches m ON m.match_id = mp.match_id
+            WHERE mp.steam_id_64 IN ({placeholders})
+              AND m.map_name IS NOT NULL
+              AND m.map_name != ''
+            GROUP BY mp.steam_id_64, m.map_name
+            """,
+            chunk,
+        ).fetchall()
+        map_values = [
+            tuple(row[:4]) + (
+                (row[3] or 0) / row[2] if row[2] else 0.0,
+            ) + tuple(row[4:])
+            for row in map_rows
+        ]
+        conn_l3.executemany(
+            """
+            INSERT INTO dm_player_map_stats (
+                steam_id_64, map_name, matches, wins, win_rate,
+                avg_rating, avg_kd, avg_adr, avg_kast,
+                best_rating, worst_rating
+            ) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)
+            """,
+            map_values,
+        )
+        total_maps += len(map_values)
+
+        round_counts = {
+            str(row[0]): int(row[1] or 0)
+            for row in conn_l2.execute(
+                f"""
+                SELECT steam_id_64, SUM(round_total)
+                FROM fact_match_players
+                WHERE steam_id_64 IN ({placeholders})
+                GROUP BY steam_id_64
+                """,
+                chunk,
+            )
+        }
+        weapon_rows = conn_l2.execute(
+            f"""
+            SELECT
+                attacker_steam_id,
+                weapon,
+                COUNT(*) AS total_kills,
+                SUM(CASE WHEN is_headshot = 1 THEN 1 ELSE 0 END) AS total_headshots,
+                COUNT(DISTINCT match_id || ':' || round_num) AS usage_rounds
+            FROM fact_round_events
+            WHERE event_type = 'kill'
+              AND attacker_steam_id IN ({placeholders})
+              AND weapon IS NOT NULL
+              AND weapon != ''
+            GROUP BY attacker_steam_id, weapon
+            """,
+            chunk,
+        ).fetchall()
+        weapon_values = []
+        for row in weapon_rows:
+            rounds = round_counts.get(str(row[0]), 0)
+            kills = int(row[2] or 0)
+            headshots = int(row[3] or 0)
+            usage_rounds = int(row[4] or 0)
+            hs_rate = headshots / kills if kills else 0.0
+            usage_rate = usage_rounds / rounds if rounds else 0.0
+            kills_per_round = kills / rounds if rounds else 0.0
+            effectiveness = kills / usage_rounds if usage_rounds else 0.0
+            weapon_values.append((
+                str(row[0]),
+                row[1],
+                kills,
+                headshots,
+                hs_rate,
+                usage_rounds,
+                usage_rate,
+                kills_per_round,
+                effectiveness,
+            ))
+        conn_l3.executemany(
+            """
+            INSERT INTO dm_player_weapon_stats (
+                steam_id_64, weapon_name, total_kills, total_headshots,
+                hs_rate, usage_rounds, usage_rate,
+                avg_kills_per_round, effectiveness_score
+            ) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?)
+            """,
+            weapon_values,
+        )
+        total_weapons += len(weapon_values)
+
+        period_values = _calculate_period_rows(history_rows)
+        conn_l3.executemany(
+            """
+            INSERT INTO dm_player_period_stats (
+                steam_id_64, period_key, period_label,
+                period_start, period_end, matches, wins, win_rate,
+                avg_rating, avg_kd, avg_adr, avg_kast,
+                total_kills, total_deaths, sample_reliable
+            ) VALUES (?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?, ?)
+            """,
+            period_values,
+        )
+        total_periods += len(period_values)
+
+        record_values = _calculate_record_rows(history_rows)
+        conn_l3.executemany(
+            """
+            INSERT INTO dm_player_records (
+                steam_id_64, record_key, record_label, record_value,
+                match_id, map_name, match_date
+            ) VALUES (?, ?, ?, ?, ?, ?, ?)
+            """,
+            record_values,
+        )
+        total_records += len(record_values)
+
+    logger.info(
+        "Auxiliary marts rebuilt: %s history, %s map, %s weapon, "
+        "%s period, %s record rows",
+        total_history,
+        total_maps,
+        total_weapons,
+        total_periods,
+        total_records,
+    )
+
+
+def _group_player_match_rows(history_rows):
+    grouped = defaultdict(list)
+    for row in history_rows:
+        grouped[str(row['steam_id_64'])].append(row)
+    for rows in grouped.values():
+        rows.sort(key=lambda row: (row['start_time'] or 0, row['match_id']))
+    return grouped
+
+
+def _safe_average(rows, key):
+    values = [float(row[key]) for row in rows if row[key] is not None]
+    return sum(values) / len(values) if values else None
+
+
+def _calculate_period_rows(history_rows):
+    result = []
+    for steam_id, all_rows in _group_player_match_rows(history_rows).items():
+        latest_time = max(int(row['start_time'] or 0) for row in all_rows)
+        period_groups = [
+            ('career', '生涯', all_rows),
+            ('last_10', '最近 10 场', all_rows[-10:]),
+            ('last_20', '最近 20 场', all_rows[-20:]),
+            ('last_30', '最近 30 场', all_rows[-30:]),
+            (
+                'days_30',
+                '最近 30 天',
+                [
+                    row for row in all_rows
+                    if int(row['start_time'] or 0) >= latest_time - 30 * 86400
+                ],
+            ),
+            (
+                'days_90',
+                '最近 90 天',
+                [
+                    row for row in all_rows
+                    if int(row['start_time'] or 0) >= latest_time - 90 * 86400
+                ],
+            ),
+        ]
+
+        for period_key, period_label, rows in period_groups:
+            if not rows:
+                continue
+            matches = len(rows)
+            wins = sum(1 for row in rows if row['is_win'])
+            kills = sum(int(row['kills'] or 0) for row in rows)
+            deaths = sum(int(row['deaths'] or 0) for row in rows)
+            result.append((
+                steam_id,
+                period_key,
+                period_label,
+                min(int(row['start_time'] or 0) for row in rows),
+                max(int(row['start_time'] or 0) for row in rows),
+                matches,
+                wins,
+                wins / matches,
+                _safe_average(rows, 'rating'),
+                kills / deaths if deaths else float(kills),
+                _safe_average(rows, 'adr'),
+                _safe_average(rows, 'kast'),
+                kills,
+                deaths,
+                1 if matches >= 10 else 0,
+            ))
+    return result
+
+
+def _calculate_record_rows(history_rows):
+    result = []
+    metric_definitions = (
+        ('highest_rating', '最高 Rating', 'rating'),
+        ('most_kills', '单场最多击杀', 'kills'),
+        ('highest_adr', '单场最高 ADR', 'adr'),
+        ('highest_kd', '单场最高 K/D', 'kd_ratio'),
+        ('most_headshots', '单场最多爆头', 'headshot_count'),
+    )
+
+    for steam_id, rows in _group_player_match_rows(history_rows).items():
+        for record_key, record_label, field in metric_definitions:
+            candidates = [row for row in rows if row[field] is not None]
+            if not candidates:
+                continue
+            best = max(
+                candidates,
+                key=lambda row: (
+                    float(row[field]),
+                    int(row['start_time'] or 0),
+                ),
+            )
+            result.append((
+                steam_id,
+                record_key,
+                record_label,
+                float(best[field]),
+                best['match_id'],
+                best['map_name'],
+                best['start_time'],
+            ))
+
+        longest_streak = 0
+        current_streak = 0
+        streak_end = None
+        for row in rows:
+            if row['is_win']:
+                current_streak += 1
+                if current_streak >= longest_streak:
+                    longest_streak = current_streak
+                    streak_end = row
+            else:
+                current_streak = 0
+        if streak_end is not None:
+            result.append((
+                steam_id,
+                'longest_win_streak',
+                '最长连胜',
+                float(longest_streak),
+                streak_end['match_id'],
+                streak_end['map_name'],
+                streak_end['start_time'],
+            ))
+    return result
+
+
+def _update_percentiles(conn_l3, steam_ids):
+    """Calculate a real percentile among eligible players in this build."""
+    if not steam_ids:
+        return
+
+    score_rows = []
+    for start in range(0, len(steam_ids), 400):
+        chunk = steam_ids[start:start + 400]
+        placeholders = ','.join('?' for _ in chunk)
+        conn_l3.execute(
+            f"""
+            UPDATE dm_player_features
+            SET tier_percentile = NULL
+            WHERE steam_id_64 IN ({placeholders})
+            """,
+            chunk,
+        )
+        score_rows.extend(conn_l3.execute(
+            f"""
+            SELECT steam_id_64, score_overall
+            FROM dm_player_features
+            WHERE steam_id_64 IN ({placeholders})
+              AND score_overall > 0
+            """,
+            chunk,
+        ).fetchall())
+
+    if not score_rows:
+        return
+
+    scores = [float(row[1]) for row in score_rows]
+    percentile_values = []
+    for row in score_rows:
+        score = float(row[1])
+        percentile = sum(value <= score for value in scores) / len(scores) * 100
+        percentile_values.append((round(percentile, 2), str(row[0])))
+
+    conn_l3.executemany(
+        """
+        UPDATE dm_player_features
+        SET tier_percentile = ?
+        WHERE steam_id_64 = ?
+        """,
+        percentile_values,
+    )
+    logger.info("Updated percentiles for %s eligible players", len(score_rows))
+
+
 def _parse_args():
    parser = argparse.ArgumentParser()
    parser.add_argument("--force", action="store_true")
    parser.add_argument("--workers", type=int, default=1)
+    parser.add_argument("--no-backup", action="store_true")
    return parser.parse_args()

 if __name__ == "__main__":
    args = _parse_args()
-    main(force_all=args.force, workers=args.workers)
+    main(
+        force_all=args.force,
+        workers=args.workers,
+        create_backup=not args.no_backup,
+    )
@@ -65,8 +65,8 @@ class CompositeProcessor(BaseFeatureProcessor):
        # Classify tier based on overall score
        features['tier_classification'] = CompositeProcessor._classify_tier(features['score_overall'])
        
-        # Percentile rank (placeholder - requires all players)
-        features['tier_percentile'] = min(features['score_overall'], 100.0)
+        # Filled by L3_Builder after every eligible player has been calculated.
+        features['tier_percentile'] = None
        
        return features
    
@@ -266,13 +266,13 @@ class CompositeProcessor(BaseFeatureProcessor):
        STABILITY Score (0-100) | 8%
        """
        # Extract features
-        volatility = features.get('meta_rating_volatility', 0.0)
-        loss_rating = features.get('meta_loss_rating', 0.0)
-        consistency = features.get('meta_rating_consistency', 0.0)
-        tilt_resilience = features.get('int_pressure_tilt_resistance', 0.0)
-        map_stable = features.get('meta_map_stability', 0.0)
-        elo_stable = features.get('meta_elo_tier_stability', 0.0)
-        recent_form = features.get('meta_recent_form_rating', 0.0)
+        volatility = features.get('meta_rating_volatility') or 0.0
+        loss_rating = features.get('meta_loss_rating') or 0.0
+        consistency = features.get('meta_rating_consistency') or 0.0
+        tilt_resilience = features.get('int_pressure_tilt_resistance') or 0.0
+        map_stable = features.get('meta_map_stability') or 0.0
+        elo_stable = features.get('meta_elo_tier_stability') or 0.0
+        recent_form = features.get('meta_recent_form_rating') or 0.0
        
        # Normalize
        # Volatility: Reverse score. 100 - (Vol * 220)
@@ -281,8 +281,8 @@ class CompositeProcessor(BaseFeatureProcessor):
        loss_score = min((loss_rating / 1.00) * 100, 100)
        cons_score = min((consistency / 70) * 100, 100)
        tilt_score = min((tilt_resilience / 0.80) * 100, 100)
-        map_score = min((map_stable / 0.25) * 100, 100)
-        elo_score = min((elo_stable / 0.48) * 100, 100)
+        map_score = max(0, min(100, 100 - (map_stable / 0.25) * 100))
+        elo_score = max(0, min(100, 100 - (elo_stable / 0.48) * 100))
        recent_score = min((recent_form / 1.15) * 100, 100)
        
        # Weighted Sum
@@ -337,12 +337,12 @@ class CompositeProcessor(BaseFeatureProcessor):
        PACE Score (0-100) | 5%
        """
        # Extract features
-        early_kill_pct = features.get('int_timing_early_kill_share', 0.0)
-        aggression = features.get('int_timing_aggression_index', 0.0)
-        trade_speed = features.get('int_trade_response_time', 0.0)
-        trade_kill = features.get('int_trade_kill_count', 0)
-        teamwork = features.get('int_teamwork_score', 0.0)
-        first_contact = features.get('int_timing_first_contact_time', 0.0)
+        early_kill_pct = features.get('int_timing_early_kill_share') or 0.0
+        aggression = features.get('int_timing_aggression_index') or 0.0
+        trade_speed = features.get('int_trade_response_time') or 0.0
+        trade_kill = features.get('int_trade_kill_count') or 0
+        teamwork = features.get('int_teamwork_score') or 0.0
+        first_contact = features.get('int_timing_first_contact_time') or 0.0
        
        # Normalize
        early_score = min((early_kill_pct / 0.44) * 100, 100)
@@ -353,7 +353,7 @@ class CompositeProcessor(BaseFeatureProcessor):
        if trade_speed > 0.01:
            trade_speed_score = min((2.0 / trade_speed) * 100, 100)
        else:
-            trade_speed_score = 100 # Instant trade
+            trade_speed_score = 0
            
        trade_kill_score = min((trade_kill / 650) * 100, 100)
        teamwork_score = min((teamwork / 29) * 100, 100)
@@ -362,13 +362,7 @@ class CompositeProcessor(BaseFeatureProcessor):
        if first_contact > 0.01:
            first_contact_score = min((30 / first_contact) * 100, 100)
        else:
-            first_contact_score = 0 # If 0, probably no data, safe to say 0? Or 100?
-            # 0 first contact time means instant damage.
-            # But "30 / Contact" means smaller contact time gives higher score.
-            # If contact time is 0, score explodes.
-            # Realistically first contact time is > 0.
-            # I will clamp it.
-            first_contact_score = 100 # Assume very fast
+            first_contact_score = 0
            
        # Weighted Sum
        pace_score = (
@@ -416,5 +410,5 @@ def _get_default_composite_features() -> Dict[str, Any]:
        'score_pace': 0.0,
        'score_overall': 0.0,
        'tier_classification': 'Beginner',
-        'tier_percentile': 0.0,
+        'tier_percentile': None,
    }
@@ -466,7 +466,8 @@ class IntelligenceProcessor(BaseFeatureProcessor):
        - int_pos_spatial_iq_score
        - int_pos_avg_distance_from_teammates
        
-        Note: Simplified implementation - full version requires DBSCAN clustering
+        Only geometry-independent values are calculated here. Metrics that
+        require map boundaries, paths or teammate positions remain NULL.
        """
        cursor = conn_l2.cursor()
        
@@ -481,26 +482,23 @@ class IntelligenceProcessor(BaseFeatureProcessor):
        has_position_data = cursor.fetchone()[0] > 0
        
        if not has_position_data:
-            # Return placeholder values if no position data
            return {
-                'int_pos_site_a_control_rate': 0.0,
-                'int_pos_site_b_control_rate': 0.0,
-                'int_pos_mid_control_rate': 0.0,
-                'int_pos_favorite_position': 'unknown',
-                'int_pos_position_diversity': 0.0,
-                'int_pos_rotation_speed': 0.0,
-                'int_pos_map_coverage': 0.0,
-                'int_pos_lurk_tendency': 0.0,
-                'int_pos_site_anchor_score': 0.0,
-                'int_pos_entry_route_diversity': 0.0,
-                'int_pos_retake_positioning': 0.0,
-                'int_pos_postplant_positioning': 0.0,
-                'int_pos_spatial_iq_score': 0.0,
-                'int_pos_avg_distance_from_teammates': 0.0,
+                'int_pos_site_a_control_rate': None,
+                'int_pos_site_b_control_rate': None,
+                'int_pos_mid_control_rate': None,
+                'int_pos_favorite_position': None,
+                'int_pos_position_diversity': None,
+                'int_pos_rotation_speed': None,
+                'int_pos_map_coverage': None,
+                'int_pos_lurk_tendency': None,
+                'int_pos_site_anchor_score': None,
+                'int_pos_entry_route_diversity': None,
+                'int_pos_retake_positioning': None,
+                'int_pos_postplant_positioning': None,
+                'int_pos_spatial_iq_score': None,
+                'int_pos_avg_distance_from_teammates': None,
            }
        
-        # Simplified position analysis (proper implementation needs clustering)
-        # Calculate basic position variance as proxy for mobility
        cursor.execute("""
            SELECT 
                AVG(attacker_pos_x) as avg_x,
@@ -515,34 +513,24 @@ class IntelligenceProcessor(BaseFeatureProcessor):
        pos_row = cursor.fetchone()
        position_count = pos_row[3] if pos_row[3] else 1
        
-        # Position diversity based on unique grid cells visited
        position_diversity = min(position_count / 50.0, 1.0)  # Normalize to 0-1
-        
-        # Map coverage (simplified)
        map_coverage = position_diversity
-        
-        # Site control rates CANNOT be calculated without map-specific geometry data
-        # Each map (Dust2, Mirage, Nuke, etc.) has different site boundaries
-        # Would require: CREATE TABLE map_boundaries (map_name, site_name, min_x, max_x, min_y, max_y)
-        # Commenting out these 3 features:
-        # - int_pos_site_a_control_rate
-        # - int_pos_site_b_control_rate  
-        # - int_pos_mid_control_rate
+
        return {
-            'int_pos_site_a_control_rate': 0.33,  # Placeholder
-            'int_pos_site_b_control_rate': 0.33,  # Placeholder
-            'int_pos_mid_control_rate': 0.34,     # Placeholder
-            'int_pos_favorite_position': 'mid',
+            'int_pos_site_a_control_rate': None,
+            'int_pos_site_b_control_rate': None,
+            'int_pos_mid_control_rate': None,
+            'int_pos_favorite_position': None,
            'int_pos_position_diversity': round(position_diversity, 3),
-            'int_pos_rotation_speed': 50.0,
+            'int_pos_rotation_speed': None,
            'int_pos_map_coverage': round(map_coverage, 3),
-            'int_pos_lurk_tendency': 0.25,
-            'int_pos_site_anchor_score': 50.0,
+            'int_pos_lurk_tendency': None,
+            'int_pos_site_anchor_score': None,
            'int_pos_entry_route_diversity': round(position_diversity, 3),
-            'int_pos_retake_positioning': 50.0,
-            'int_pos_postplant_positioning': 50.0,
+            'int_pos_retake_positioning': None,
+            'int_pos_postplant_positioning': None,
            'int_pos_spatial_iq_score': round(position_diversity * 100, 2),
-            'int_pos_avg_distance_from_teammates': 500.0,
+            'int_pos_avg_distance_from_teammates': None,
        }
    
    @staticmethod
@@ -706,20 +694,20 @@ def _get_default_intelligence_features() -> Dict[str, Any]:
        'int_pressure_big_moment_score': 0.0,
        'int_pressure_tilt_resistance': 0.0,
        # Position Mastery (14)
-        'int_pos_site_a_control_rate': 0.0,
-        'int_pos_site_b_control_rate': 0.0,
-        'int_pos_mid_control_rate': 0.0,
-        'int_pos_favorite_position': 'unknown',
-        'int_pos_position_diversity': 0.0,
-        'int_pos_rotation_speed': 0.0,
-        'int_pos_map_coverage': 0.0,
-        'int_pos_lurk_tendency': 0.0,
-        'int_pos_site_anchor_score': 0.0,
-        'int_pos_entry_route_diversity': 0.0,
-        'int_pos_retake_positioning': 0.0,
-        'int_pos_postplant_positioning': 0.0,
-        'int_pos_spatial_iq_score': 0.0,
-        'int_pos_avg_distance_from_teammates': 0.0,
+        'int_pos_site_a_control_rate': None,
+        'int_pos_site_b_control_rate': None,
+        'int_pos_mid_control_rate': None,
+        'int_pos_favorite_position': None,
+        'int_pos_position_diversity': None,
+        'int_pos_rotation_speed': None,
+        'int_pos_map_coverage': None,
+        'int_pos_lurk_tendency': None,
+        'int_pos_site_anchor_score': None,
+        'int_pos_entry_route_diversity': None,
+        'int_pos_retake_positioning': None,
+        'int_pos_postplant_positioning': None,
+        'int_pos_spatial_iq_score': None,
+        'int_pos_avg_distance_from_teammates': None,
        # Trade Network (8)
        'int_trade_kill_count': 0,
        'int_trade_kill_rate': 0.0,
@@ -60,10 +60,11 @@ class MetaProcessor(BaseFeatureProcessor):
        
        # Get recent matches for volatility
        cursor.execute("""
-            SELECT rating
-            FROM fact_match_players
-            WHERE steam_id_64 = ?
-            ORDER BY match_id DESC
+            SELECT p.rating
+            FROM fact_match_players p
+            JOIN fact_matches m ON m.match_id = p.match_id
+            WHERE p.steam_id_64 = ?
+            ORDER BY m.start_time DESC, p.match_id DESC
            LIMIT 20
        """, (steam_id,))
        
@@ -141,8 +142,35 @@ class MetaProcessor(BaseFeatureProcessor):
        map_ratings = [row[1] for row in cursor.fetchall() if row[1] is not None]
        map_stability = SafeAggregator.safe_stddev(map_ratings, 0.0)
        
-        # ELO tier stability (placeholder)
-        elo_tier_stability = rating_volatility  # Simplified
+        cursor.execute("""
+            SELECT
+                CASE
+                    WHEN p.origin_elo - opponent.avg_elo > 200 THEN 'lower'
+                    WHEN p.origin_elo - opponent.avg_elo < -200 THEN 'higher'
+                    ELSE 'similar'
+                END AS opponent_tier,
+                AVG(p.rating) AS avg_rating
+            FROM fact_match_players p
+            JOIN (
+                SELECT match_id, team_id, AVG(origin_elo) AS avg_elo
+                FROM fact_match_players
+                WHERE origin_elo IS NOT NULL
+                GROUP BY match_id, team_id
+            ) opponent
+              ON opponent.match_id = p.match_id
+             AND opponent.team_id != p.team_id
+            WHERE p.steam_id_64 = ?
+              AND p.origin_elo IS NOT NULL
+              AND p.rating IS NOT NULL
+            GROUP BY opponent_tier
+        """, (steam_id,))
+        elo_tier_ratings = [
+            row[1] for row in cursor.fetchall() if row[1] is not None
+        ]
+        elo_tier_stability = SafeAggregator.safe_stddev(
+            elo_tier_ratings,
+            0.0,
+        )
        
        return {
            'meta_rating_volatility': round(rating_volatility, 3),
@@ -378,6 +378,56 @@ CREATE TABLE IF NOT EXISTS dm_player_weapon_stats (
 CREATE INDEX IF NOT EXISTS idx_player_weapon_stats_player ON dm_player_weapon_stats(steam_id_64);
 CREATE INDEX IF NOT EXISTS idx_player_weapon_stats_weapon ON dm_player_weapon_stats(weapon_name);

+-- ============================================================================
+-- Profile Mart: Time-window statistics
+-- ============================================================================
+CREATE TABLE IF NOT EXISTS dm_player_period_stats (
+    steam_id_64 TEXT NOT NULL,
+    period_key TEXT NOT NULL,
+    period_label TEXT NOT NULL,
+    period_start INTEGER,
+    period_end INTEGER,
+    matches INTEGER NOT NULL DEFAULT 0,
+    wins INTEGER NOT NULL DEFAULT 0,
+    win_rate REAL,
+    avg_rating REAL,
+    avg_kd REAL,
+    avg_adr REAL,
+    avg_kast REAL,
+    total_kills INTEGER NOT NULL DEFAULT 0,
+    total_deaths INTEGER NOT NULL DEFAULT 0,
+    sample_reliable BOOLEAN NOT NULL DEFAULT 0,
+    last_updated TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
+
+    PRIMARY KEY (steam_id_64, period_key),
+    FOREIGN KEY (steam_id_64)
+        REFERENCES dm_player_features(steam_id_64) ON DELETE CASCADE
+);
+
+CREATE INDEX IF NOT EXISTS idx_player_period_player
+ON dm_player_period_stats(steam_id_64, period_key);
+
+-- ============================================================================
+-- Profile Mart: Career records linked to the source match
+-- ============================================================================
+CREATE TABLE IF NOT EXISTS dm_player_records (
+    steam_id_64 TEXT NOT NULL,
+    record_key TEXT NOT NULL,
+    record_label TEXT NOT NULL,
+    record_value REAL,
+    match_id TEXT,
+    map_name TEXT,
+    match_date INTEGER,
+    last_updated TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
+
+    PRIMARY KEY (steam_id_64, record_key),
+    FOREIGN KEY (steam_id_64)
+        REFERENCES dm_player_features(steam_id_64) ON DELETE CASCADE
+);
+
+CREATE INDEX IF NOT EXISTS idx_player_records_player
+ON dm_player_records(steam_id_64, record_key);
+
 -- ============================================================================
 -- Schema Summary
 -- ============================================================================
@@ -391,4 +441,6 @@ CREATE INDEX IF NOT EXISTS idx_player_weapon_stats_weapon ON dm_player_weapon_st
 -- dm_player_match_history: Per-match snapshots for trend analysis
 -- dm_player_map_stats: Map-level aggregations
 -- dm_player_weapon_stats: Weapon usage statistics
+-- dm_player_period_stats: Career/recent time-window aggregations
+-- dm_player_records: Career record values and source matches
 -- ============================================================================
@@ -0,0 +1,45 @@
+# Database Governance
+
+The repository intentionally keeps SQLite for the current private-team scale.
+This directory separates four different responsibilities:
+
+| Layer | Database | Grain | Owner |
+|---|---|---|---|
+| L1 | `L1/L1.db` | One raw network capture per match | Import pipeline |
+| L2 | `L2/L2.db` | Normalized match, player, round and event facts | L2 Builder |
+| L3 | `L3/L3.db` | Roster features and profile marts | L3 Builder |
+| Web | `Web/Web_App.sqlite` | Lineups, comments, jobs and editorial data | Flask app |
+
+## Rules
+
+1. Paths are defined only in `database/paths.py`.
+2. Schemas live next to their owning database.
+3. Builders may read the previous layer and write only their own layer.
+4. User-generated Web data is never restored as part of an ETL rollback.
+5. A full import must run through `database/pipeline.py`.
+6. Pipeline runs are serialized by `database/.pipeline.lock`.
+7. L1/L2/L3 are backed up before a full pipeline run.
+8. Missing metrics are stored as `NULL`, not fabricated zero values.
+9. `Admin -> Data Integrity` is the operational source of truth.
+10. Web schema changes increment `Config.WEB_SCHEMA_VERSION`.
+
+## Entry Points
+
+```bash
+make l1          # Import output_arena JSON into L1
+make l2          # Rebuild normalized facts
+make l3          # Rebuild active-roster features
+make pipeline    # Run L1 -> L2 -> L3 with backup and validation
+make check       # Compile and run tests
+```
+
+## Directory Policy
+
+- `L1/`, `L2/`, `L3/`, `Web/`: active code, schema and database.
+- `backups/`: generated rollback snapshots; ignored by Git.
+- `schema_bkp/`: historical schema research only; not used at runtime.
+- `L1B/`: reserved demo-parser integration; not used at runtime.
+- `L3/Roadmap/`: historical design notes; not used at runtime.
+
+Large-scale directory moves are deliberately deferred until the legacy
+builders no longer depend on their current module layout.
@@ -0,0 +1,95 @@
+CREATE TABLE IF NOT EXISTS schema_migrations (
+    version INTEGER PRIMARY KEY,
+    description TEXT NOT NULL,
+    applied_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
+);
+
+CREATE TABLE IF NOT EXISTS comments (
+    id INTEGER PRIMARY KEY AUTOINCREMENT,
+    user_id TEXT,
+    username TEXT,
+    target_type TEXT NOT NULL,
+    target_id TEXT NOT NULL,
+    content TEXT NOT NULL,
+    likes INTEGER NOT NULL DEFAULT 0,
+    is_hidden INTEGER NOT NULL DEFAULT 0,
+    created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
+);
+
+CREATE INDEX IF NOT EXISTS idx_comments_target
+ON comments(target_type, target_id, is_hidden, created_at DESC);
+
+CREATE TABLE IF NOT EXISTS player_metadata (
+    steam_id_64 TEXT PRIMARY KEY,
+    notes TEXT,
+    tags TEXT NOT NULL DEFAULT '[]',
+    updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
+);
+
+CREATE TABLE IF NOT EXISTS strategy_boards (
+    id INTEGER PRIMARY KEY AUTOINCREMENT,
+    title TEXT NOT NULL,
+    map_name TEXT NOT NULL,
+    data_json TEXT NOT NULL,
+    created_by TEXT,
+    created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
+);
+
+CREATE TABLE IF NOT EXISTS team_lineups (
+    id INTEGER PRIMARY KEY AUTOINCREMENT,
+    name TEXT NOT NULL,
+    description TEXT,
+    player_ids_json TEXT NOT NULL DEFAULT '[]',
+    is_active INTEGER NOT NULL DEFAULT 0,
+    created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
+);
+
+CREATE UNIQUE INDEX IF NOT EXISTS idx_team_lineups_single_active
+ON team_lineups(is_active)
+WHERE is_active = 1;
+
+CREATE TABLE IF NOT EXISTS wiki_pages (
+    id INTEGER PRIMARY KEY AUTOINCREMENT,
+    path TEXT NOT NULL UNIQUE,
+    title TEXT NOT NULL,
+    content TEXT NOT NULL DEFAULT '',
+    updated_by TEXT,
+    updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
+    created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP
+);
+
+CREATE TABLE IF NOT EXISTS etl_jobs (
+    id INTEGER PRIMARY KEY AUTOINCREMENT,
+    job_type TEXT NOT NULL,
+    status TEXT NOT NULL DEFAULT 'queued'
+        CHECK (status IN ('queued', 'running', 'succeeded', 'failed')),
+    match_id TEXT,
+    input_path TEXT,
+    current_stage TEXT,
+    progress INTEGER NOT NULL DEFAULT 0
+        CHECK (progress >= 0 AND progress <= 100),
+    message TEXT,
+    log_text TEXT NOT NULL DEFAULT '',
+    created_by TEXT,
+    created_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
+    started_at TIMESTAMP,
+    finished_at TIMESTAMP,
+    duration_seconds REAL
+);
+
+CREATE INDEX IF NOT EXISTS idx_etl_jobs_created
+ON etl_jobs(created_at DESC, id DESC);
+
+CREATE INDEX IF NOT EXISTS idx_etl_jobs_status
+ON etl_jobs(status, created_at);
+
+CREATE TABLE IF NOT EXISTS match_imports (
+    match_id TEXT PRIMARY KEY,
+    content_sha256 TEXT NOT NULL,
+    source_path TEXT NOT NULL,
+    status TEXT NOT NULL DEFAULT 'queued',
+    job_id INTEGER,
+    imported_at TIMESTAMP,
+    updated_at TIMESTAMP NOT NULL DEFAULT CURRENT_TIMESTAMP,
+    FOREIGN KEY (job_id) REFERENCES etl_jobs(id) ON DELETE SET NULL
+);
@@ -0,0 +1,2 @@
+"""Database builders, schemas, maintenance tools and local data stores."""
+
@@ -0,0 +1,198 @@
+import sqlite3
+from typing import Any, Dict, List, Optional
+
+from database.paths import WEB_DB
+
+
+class JobStore:
+    def __init__(self, database_path=WEB_DB):
+        self.database_path = str(database_path)
+
+    def _connect(self):
+        db = sqlite3.connect(self.database_path, timeout=30)
+        db.row_factory = sqlite3.Row
+        db.execute('PRAGMA busy_timeout = 30000')
+        db.execute('PRAGMA foreign_keys = ON')
+        return db
+
+    def create_job(
+        self,
+        job_type: str,
+        match_id: Optional[str] = None,
+        input_path: Optional[str] = None,
+        created_by: Optional[str] = None,
+    ) -> int:
+        db = self._connect()
+        try:
+            cursor = db.execute(
+                """
+                INSERT INTO etl_jobs (
+                    job_type, match_id, input_path, created_by
+                ) VALUES (?, ?, ?, ?)
+                """,
+                [job_type, match_id, input_path, created_by],
+            )
+            db.commit()
+            return int(cursor.lastrowid)
+        finally:
+            db.close()
+
+    def get_job(self, job_id: int) -> Optional[Dict[str, Any]]:
+        db = self._connect()
+        try:
+            row = db.execute(
+                'SELECT * FROM etl_jobs WHERE id = ?',
+                [job_id],
+            ).fetchone()
+            return dict(row) if row else None
+        finally:
+            db.close()
+
+    def list_jobs(self, limit: int = 20) -> List[Dict[str, Any]]:
+        db = self._connect()
+        try:
+            rows = db.execute(
+                """
+                SELECT *
+                FROM etl_jobs
+                ORDER BY created_at DESC, id DESC
+                LIMIT ?
+                """,
+                [max(1, min(int(limit), 100))],
+            ).fetchall()
+            return [dict(row) for row in rows]
+        finally:
+            db.close()
+
+    def start_job(self, job_id: int, stage: str, message: str):
+        db = self._connect()
+        try:
+            db.execute(
+                """
+                UPDATE etl_jobs
+                SET status = 'running',
+                    current_stage = ?,
+                    progress = 1,
+                    message = ?,
+                    started_at = CURRENT_TIMESTAMP
+                WHERE id = ? AND status = 'queued'
+                """,
+                [stage, message, job_id],
+            )
+            db.commit()
+        finally:
+            db.close()
+
+    def update_progress(
+        self,
+        job_id: int,
+        stage: str,
+        progress: int,
+        message: str,
+    ):
+        db = self._connect()
+        try:
+            db.execute(
+                """
+                UPDATE etl_jobs
+                SET current_stage = ?, progress = ?, message = ?
+                WHERE id = ?
+                """,
+                [stage, max(0, min(int(progress), 100)), message, job_id],
+            )
+            db.commit()
+        finally:
+            db.close()
+
+    def append_log(self, job_id: int, text: str):
+        if not text:
+            return
+        db = self._connect()
+        try:
+            db.execute(
+                """
+                UPDATE etl_jobs
+                SET log_text = substr(log_text || ?, -100000)
+                WHERE id = ?
+                """,
+                [text, job_id],
+            )
+            db.commit()
+        finally:
+            db.close()
+
+    def finish_job(
+        self,
+        job_id: int,
+        succeeded: bool,
+        message: str,
+        duration_seconds: float,
+    ):
+        status = 'succeeded' if succeeded else 'failed'
+        db = self._connect()
+        try:
+            db.execute(
+                """
+                UPDATE etl_jobs
+                SET status = ?,
+                    current_stage = ?,
+                    progress = ?,
+                    message = ?,
+                    finished_at = CURRENT_TIMESTAMP,
+                    duration_seconds = ?
+                WHERE id = ?
+                """,
+                [
+                    status,
+                    'complete' if succeeded else 'failed',
+                    100 if succeeded else 0,
+                    message,
+                    round(float(duration_seconds), 3),
+                    job_id,
+                ],
+            )
+            db.execute(
+                """
+                UPDATE match_imports
+                SET status = ?,
+                    imported_at = CASE
+                        WHEN ? = 'succeeded' THEN CURRENT_TIMESTAMP
+                        ELSE imported_at
+                    END,
+                    updated_at = CURRENT_TIMESTAMP
+                WHERE job_id = ?
+                """,
+                [status, status, job_id],
+            )
+            db.commit()
+        finally:
+            db.close()
+
+    def upsert_match_import(
+        self,
+        match_id: str,
+        content_sha256: str,
+        source_path: str,
+        status: str,
+        job_id: int,
+    ):
+        db = self._connect()
+        try:
+            db.execute(
+                """
+                INSERT INTO match_imports (
+                    match_id, content_sha256, source_path, status, job_id
+                ) VALUES (?, ?, ?, ?, ?)
+                ON CONFLICT(match_id) DO UPDATE SET
+                    content_sha256 = excluded.content_sha256,
+                    source_path = excluded.source_path,
+                    status = excluded.status,
+                    job_id = excluded.job_id,
+                    updated_at = CURRENT_TIMESTAMP
+                """,
+                [match_id, content_sha256, source_path, status, job_id],
+            )
+            db.commit()
+        finally:
+            db.close()
+
@@ -0,0 +1,140 @@
+from datetime import datetime, timezone
+import json
+from pathlib import Path
+import shutil
+import sqlite3
+from typing import Dict
+
+from database.paths import BACKUP_ROOT, L1_DB, L2_DB, L3_DB
+
+
+MANAGED_DATABASES = {
+    'l1': L1_DB,
+    'l2': L2_DB,
+    'l3': L3_DB,
+}
+
+
+def quick_check(path: Path) -> str:
+    if not path.exists():
+        return 'missing'
+    db = sqlite3.connect(str(path))
+    try:
+        return str(db.execute('PRAGMA quick_check').fetchone()[0])
+    finally:
+        db.close()
+
+
+def backup_database(source_path: Path, backup_path: Path):
+    if not source_path.exists():
+        raise FileNotFoundError(f'Database does not exist: {source_path}')
+
+    backup_path.parent.mkdir(parents=True, exist_ok=True)
+    source = sqlite3.connect(str(source_path))
+    destination = sqlite3.connect(str(backup_path))
+    try:
+        source.backup(destination)
+        result = destination.execute('PRAGMA quick_check').fetchone()[0]
+        if result != 'ok':
+            raise RuntimeError(
+                f'Backup quick_check failed for {source_path.name}: {result}'
+            )
+    finally:
+        source.close()
+        destination.close()
+
+
+def restore_database(backup_path: Path, target_path: Path):
+    if not backup_path.exists():
+        raise FileNotFoundError(f'Backup does not exist: {backup_path}')
+
+    source = sqlite3.connect(str(backup_path))
+    target = sqlite3.connect(str(target_path), timeout=30)
+    try:
+        source.backup(target)
+        result = target.execute('PRAGMA quick_check').fetchone()[0]
+        if result != 'ok':
+            raise RuntimeError(
+                f'Restored quick_check failed for {target_path.name}: {result}'
+            )
+    finally:
+        source.close()
+        target.close()
+
+
+def create_backup_set(label: str) -> Path:
+    safe_label = ''.join(
+        character for character in str(label)
+        if character.isalnum() or character in {'-', '_'}
+    )
+    if not safe_label:
+        raise ValueError('Backup label is empty after sanitization')
+
+    backup_dir = BACKUP_ROOT / safe_label
+    backup_dir.mkdir(parents=True, exist_ok=True)
+    manifest: Dict[str, object] = {
+        'label': safe_label,
+        'created_at': datetime.now(timezone.utc).isoformat(),
+        'databases': {},
+    }
+
+    for name, source_path in MANAGED_DATABASES.items():
+        backup_path = backup_dir / source_path.name
+        backup_database(source_path, backup_path)
+        manifest['databases'][name] = {
+            'source': str(source_path),
+            'backup': str(backup_path),
+            'size_bytes': backup_path.stat().st_size,
+            'quick_check': quick_check(backup_path),
+        }
+
+    with (backup_dir / 'manifest.json').open('w', encoding='utf-8') as file:
+        json.dump(manifest, file, ensure_ascii=True, indent=2)
+    return backup_dir
+
+
+def restore_backup_set(backup_dir: Path):
+    for name, target_path in MANAGED_DATABASES.items():
+        backup_path = backup_dir / target_path.name
+        restore_database(backup_path, target_path)
+
+
+def prune_backup_sets(keep: int = 3):
+    BACKUP_ROOT.mkdir(parents=True, exist_ok=True)
+    backup_dirs = sorted(
+        [path for path in BACKUP_ROOT.iterdir() if path.is_dir()],
+        key=lambda path: path.stat().st_mtime,
+        reverse=True,
+    )
+    removed = []
+    for path in backup_dirs[max(int(keep), 0):]:
+        shutil.rmtree(path)
+        removed.append(str(path))
+    return removed
+
+
+def backup_storage_status():
+    BACKUP_ROOT.mkdir(parents=True, exist_ok=True)
+    backup_dirs = [path for path in BACKUP_ROOT.iterdir() if path.is_dir()]
+    total_bytes = sum(
+        file.stat().st_size
+        for directory in backup_dirs
+        for file in directory.rglob('*')
+        if file.is_file()
+    )
+    return {
+        'sets': len(backup_dirs),
+        'total_bytes': total_bytes,
+    }
+
+
+def check_managed_databases():
+    return {
+        name: {
+            'path': str(path),
+            'exists': path.exists(),
+            'size_bytes': path.stat().st_size if path.exists() else 0,
+            'quick_check': quick_check(path),
+        }
+        for name, path in MANAGED_DATABASES.items()
+    }
@@ -0,0 +1,36 @@
+from pathlib import Path
+
+
+PROJECT_ROOT = Path(__file__).resolve().parent.parent
+DATABASE_ROOT = PROJECT_ROOT / 'database'
+
+L1_DIR = DATABASE_ROOT / 'L1'
+L2_DIR = DATABASE_ROOT / 'L2'
+L3_DIR = DATABASE_ROOT / 'L3'
+WEB_DIR = DATABASE_ROOT / 'Web'
+
+L1_DB = L1_DIR / 'L1.db'
+L2_DB = L2_DIR / 'L2.db'
+L3_DB = L3_DIR / 'L3.db'
+WEB_DB = WEB_DIR / 'Web_App.sqlite'
+
+L2_SCHEMA = L2_DIR / 'schema.sql'
+L3_SCHEMA = L3_DIR / 'schema.sql'
+WEB_SCHEMA = WEB_DIR / 'schema.sql'
+
+OUTPUT_ARENA = PROJECT_ROOT / 'output_arena'
+BACKUP_ROOT = DATABASE_ROOT / 'backups'
+PIPELINE_LOCK = DATABASE_ROOT / '.pipeline.lock'
+
+
+def ensure_runtime_directories():
+    for path in (
+        L1_DIR,
+        L2_DIR,
+        L3_DIR,
+        WEB_DIR,
+        OUTPUT_ARENA,
+        BACKUP_ROOT,
+    ):
+        path.mkdir(parents=True, exist_ok=True)
+
@@ -0,0 +1,212 @@
+import argparse
+import fcntl
+import os
+import sqlite3
+import subprocess
+import sys
+import time
+from pathlib import Path
+
+PROJECT_ROOT = Path(__file__).resolve().parent.parent
+if str(PROJECT_ROOT) not in sys.path:
+    sys.path.insert(0, str(PROJECT_ROOT))
+
+from database.job_store import JobStore
+from database.maintenance import (
+    check_managed_databases,
+    create_backup_set,
+    prune_backup_sets,
+    restore_backup_set,
+)
+from database.paths import L1_DB, L2_DB, PIPELINE_LOCK
+
+
+STAGES = (
+    ('l1', 15, Path('database/L1/L1_Builder.py')),
+    ('l2', 55, Path('database/L2/L2_Builder.py')),
+    ('l3', 85, Path('database/L3/L3_Builder.py')),
+)
+
+
+class PipelineError(RuntimeError):
+    pass
+
+
+def _run_stage(store, job_id, stage, progress, script_path, replace=False):
+    store.update_progress(
+        job_id,
+        stage,
+        progress,
+        f'Running {stage.upper()} builder',
+    )
+    command = [sys.executable, str(PROJECT_ROOT / script_path)]
+    if stage == 'l1' and replace:
+        command.append('--force')
+    if stage == 'l3':
+        command.append('--no-backup')
+
+    started = time.monotonic()
+    result = subprocess.run(
+        command,
+        cwd=str(PROJECT_ROOT),
+        capture_output=True,
+        text=True,
+        timeout=1200,
+    )
+    duration = time.monotonic() - started
+    store.append_log(
+        job_id,
+        (
+            f'\n===== {stage.upper()} ({duration:.2f}s) =====\n'
+            f'{result.stdout}\n{result.stderr}'
+        ),
+    )
+    if result.returncode != 0:
+        raise PipelineError(
+            f'{stage.upper()} builder exited with code {result.returncode}'
+        )
+
+
+def _validate_pipeline_output(match_id=None):
+    database_status = check_managed_databases()
+    failures = [
+        f"{name}: {status['quick_check']}"
+        for name, status in database_status.items()
+        if status['quick_check'] != 'ok'
+    ]
+    if failures:
+        raise PipelineError(
+            'Database quick_check failed: ' + ', '.join(failures)
+        )
+
+    if not match_id:
+        return
+
+    l1 = sqlite3.connect(str(L1_DB))
+    l2 = sqlite3.connect(str(L2_DB))
+    try:
+        raw_count = l1.execute(
+            'SELECT COUNT(*) FROM raw_iframe_network WHERE match_id = ?',
+            [match_id],
+        ).fetchone()[0]
+        match_count = l2.execute(
+            'SELECT COUNT(*) FROM fact_matches WHERE match_id = ?',
+            [match_id],
+        ).fetchone()[0]
+        player_count = l2.execute(
+            'SELECT COUNT(*) FROM fact_match_players WHERE match_id = ?',
+            [match_id],
+        ).fetchone()[0]
+        round_count = l2.execute(
+            'SELECT COUNT(*) FROM fact_rounds WHERE match_id = ?',
+            [match_id],
+        ).fetchone()[0]
+    finally:
+        l1.close()
+        l2.close()
+
+    if raw_count != 1:
+        raise PipelineError(f'L1 does not contain imported match {match_id}')
+    if match_count != 1:
+        raise PipelineError(f'L2 does not contain imported match {match_id}')
+    if player_count != 10:
+        raise PipelineError(
+            f'Imported match has {player_count} players; expected 10'
+        )
+    if round_count <= 0:
+        raise PipelineError('Imported match has no round facts')
+
+
+def run_pipeline(job_id, match_id=None, replace=False):
+    store = JobStore()
+    job = store.get_job(job_id)
+    if not job:
+        raise PipelineError(f'Unknown ETL job: {job_id}')
+
+    started = time.monotonic()
+    backup_dir = None
+    PIPELINE_LOCK.parent.mkdir(parents=True, exist_ok=True)
+    lock_file = PIPELINE_LOCK.open('w')
+    try:
+        try:
+            fcntl.flock(lock_file.fileno(), fcntl.LOCK_EX | fcntl.LOCK_NB)
+        except BlockingIOError as exc:
+            raise PipelineError('Another database pipeline is already running') from exc
+
+        store.start_job(job_id, 'backup', 'Creating rollback snapshot')
+        backup_dir = create_backup_set(f'job-{job_id}')
+        store.append_log(job_id, f'Backup created: {backup_dir}\n')
+
+        for stage, progress, script_path in STAGES:
+            _run_stage(
+                store,
+                job_id,
+                stage,
+                progress,
+                script_path,
+                replace=replace,
+            )
+
+        store.update_progress(
+            job_id,
+            'validation',
+            95,
+            'Validating imported data',
+        )
+        _validate_pipeline_output(match_id)
+        removed_backups = prune_backup_sets(keep=3)
+        if removed_backups:
+            store.append_log(
+                job_id,
+                f"Pruned old backups: {', '.join(removed_backups)}\n",
+            )
+        duration = time.monotonic() - started
+        store.finish_job(
+            job_id,
+            True,
+            'Pipeline completed and validated',
+            duration,
+        )
+        return True
+    except Exception as exc:
+        store.append_log(job_id, f'\nPIPELINE FAILED: {exc}\n')
+        if backup_dir:
+            try:
+                restore_backup_set(backup_dir)
+                store.append_log(job_id, 'Rollback snapshot restored successfully\n')
+            except Exception as restore_exc:
+                store.append_log(
+                    job_id,
+                    f'ROLLBACK FAILED: {restore_exc}\n',
+                )
+                exc = PipelineError(f'{exc}; rollback also failed: {restore_exc}')
+        store.finish_job(
+            job_id,
+            False,
+            str(exc),
+            time.monotonic() - started,
+        )
+        return False
+    finally:
+        try:
+            fcntl.flock(lock_file.fileno(), fcntl.LOCK_UN)
+        finally:
+            lock_file.close()
+
+
+def _parse_args():
+    parser = argparse.ArgumentParser()
+    parser.add_argument('--job-id', type=int, required=True)
+    parser.add_argument('--match-id')
+    parser.add_argument('--replace', action='store_true')
+    return parser.parse_args()
+
+
+if __name__ == '__main__':
+    args = _parse_args()
+    succeeded = run_pipeline(
+        args.job_id,
+        match_id=args.match_id,
+        replace=args.replace,
+    )
+    raise SystemExit(0 if succeeded else 1)
@@ -1,7 +1,3 @@
-Flask
-pandas
-numpy
-playwright
-gunicorn
-gevent
-matplotlib
+Flask>=3.0,<4
+gunicorn>=21,<24
+pytest>=8,<9
@@ -0,0 +1 @@
+
@@ -0,0 +1,620 @@
+import json
+import io
+import os
+import shutil
+import sqlite3
+import tempfile
+import unittest
+from pathlib import Path
+
+from web.config import Config
+
+
+class ApplicationIntegrationTests(unittest.TestCase):
+    @classmethod
+    def setUpClass(cls):
+        cls.temp_dir = tempfile.TemporaryDirectory()
+        cls.original_web_path = Config.DB_WEB_PATH
+        Config.DB_WEB_PATH = os.path.join(cls.temp_dir.name, 'Web_App.sqlite')
+        shutil.copy2(cls.original_web_path, Config.DB_WEB_PATH)
+
+        from web.app import create_app
+
+        cls.app = create_app()
+        cls.app.config.update(TESTING=True)
+        cls.client = cls.app.test_client()
+
+        with sqlite3.connect(Config.DB_WEB_PATH) as db:
+            raw_ids = db.execute(
+                """
+                SELECT player_ids_json
+                FROM team_lineups
+                WHERE is_active = 1
+                LIMIT 1
+                """
+            ).fetchone()[0]
+        cls.roster_ids = [str(value) for value in json.loads(raw_ids)]
+
+    @classmethod
+    def tearDownClass(cls):
+        Config.DB_WEB_PATH = cls.original_web_path
+        cls.temp_dir.cleanup()
+
+    def test_primary_pages_render(self):
+        paths = [
+            '/',
+            '/matches/',
+            '/players/',
+            f'/players/{self.roster_ids[0]}',
+            '/teams/',
+            '/tactics/',
+            '/opponents/',
+        ]
+        for path in paths:
+            with self.subTest(path=path):
+                response = self.client.get(path)
+                self.assertEqual(response.status_code, 200)
+                self.assertGreater(len(response.data), 100)
+
+    def test_admin_integrity_page_and_json_render(self):
+        with self.client.session_transaction() as session:
+            session['is_admin'] = True
+
+        response = self.client.get('/admin/data-integrity')
+        self.assertEqual(response.status_code, 200)
+        self.assertIn('数据完整性中心'.encode('utf-8'), response.data)
+
+        response = self.client.get('/admin/data-integrity?format=json')
+        self.assertEqual(response.status_code, 200)
+        report = response.get_json()
+        self.assertIn(report['overall_status'], {'pass', 'warn', 'fail'})
+        self.assertGreater(report['counts']['matches'], 0)
+        self.assertEqual(
+            report['counts']['web_schema_version'],
+            Config.WEB_SCHEMA_VERSION,
+        )
+
+        response = self.client.get('/admin/import-match')
+        self.assertEqual(response.status_code, 200)
+        self.assertIn('比赛数据导入'.encode('utf-8'), response.data)
+
+    def test_duplicate_match_upload_is_rejected_without_starting_job(self):
+        from database.paths import L1_DB
+
+        with self.client.session_transaction() as session:
+            session['is_admin'] = True
+        with sqlite3.connect(str(L1_DB)) as db:
+            raw = db.execute(
+                """
+                SELECT content
+                FROM raw_iframe_network
+                ORDER BY match_id
+                LIMIT 1
+                """
+            ).fetchone()[0]
+
+        response = self.client.post(
+            '/admin/import-match',
+            data={
+                'capture': (
+                    io.BytesIO(raw.encode('utf-8')),
+                    'iframe_network.json',
+                ),
+            },
+            content_type='multipart/form-data',
+            follow_redirects=True,
+        )
+        self.assertEqual(response.status_code, 200)
+        self.assertIn(b'already imported with identical data', response.data)
+
+    def test_player_search_works_across_l2_and_l3(self):
+        response = self.client.get('/players/?search=jAck')
+        self.assertEqual(response.status_code, 200)
+        self.assertIn(b'jAckY0987', response.data)
+
+    def test_profile_keeps_all_primary_sections(self):
+        response = self.client.get(f'/players/{self.roster_ids[0]}')
+        self.assertEqual(response.status_code, 200)
+        for label in (
+            '近期表现走势',
+            '能力八维图',
+            'CORE (核心表现)',
+            '阶段表现',
+            '职业纪录',
+            '比赛记录',
+            '地图数据',
+            '留言板',
+        ):
+            with self.subTest(label=label):
+                self.assertIn(label.encode('utf-8'), response.data)
+
+    def test_profile_period_api_and_trend_window(self):
+        steam_id = self.roster_ids[0]
+        response = self.client.get(
+            f'/players/{steam_id}/period_stats?period=last_20'
+        )
+        self.assertEqual(response.status_code, 200)
+        period = response.get_json()
+        self.assertEqual(period['period_key'], 'last_20')
+        self.assertEqual(period['matches'], 20)
+        self.assertEqual(period['sample_reliable'], 1)
+
+        response = self.client.get(
+            f'/players/{steam_id}/charts_data?period=last_10'
+        )
+        self.assertEqual(response.status_code, 200)
+        chart = response.get_json()
+        self.assertEqual(chart['period']['period_key'], 'last_10')
+        self.assertLessEqual(len(chart['trend']['labels']), 10)
+
+    def test_date_filter_uses_unix_timestamp_conversion(self):
+        from web.services.stats_service import StatsService
+
+        with self.app.app_context():
+            matches, total = StatsService.get_matches(
+                page=1,
+                per_page=20,
+                date_from='2025-01-01',
+                date_to='2026-12-31',
+            )
+        self.assertGreater(total, 0)
+        self.assertGreater(len(matches), 0)
+
+    def test_shared_matches_require_same_team(self):
+        from web.services.stats_service import StatsService
+
+        selected_ids = self.roster_ids[:2]
+        with self.app.app_context():
+            matches = StatsService.get_shared_matches(selected_ids)
+        self.assertGreater(len(matches), 0)
+
+        l2 = sqlite3.connect(Config.DB_L2_PATH)
+        try:
+            for match in matches:
+                placeholders = ','.join('?' for _ in selected_ids)
+                team_count = l2.execute(
+                    f"""
+                    SELECT COUNT(DISTINCT team_id)
+                    FROM fact_match_players
+                    WHERE match_id = ?
+                      AND steam_id_64 IN ({placeholders})
+                    """,
+                    [match['match_id']] + selected_ids,
+                ).fetchone()[0]
+                self.assertEqual(team_count, 1)
+        finally:
+            l2.close()
+
+    def test_opponent_list_only_contains_actual_opponents(self):
+        from web.services.opponent_service import OpponentService
+
+        with self.app.app_context():
+            opponents, total = OpponentService.get_opponent_list(
+                page=1,
+                per_page=20,
+            )
+        self.assertGreater(total, 0)
+        self.assertGreater(len(opponents), 0)
+
+        l2 = sqlite3.connect(Config.DB_L2_PATH)
+        try:
+            roster_ph = ','.join('?' for _ in self.roster_ids)
+            for opponent in opponents:
+                faced = l2.execute(
+                    f"""
+                    SELECT COUNT(*)
+                    FROM fact_match_players opponent
+                    JOIN fact_match_players roster
+                      ON roster.match_id = opponent.match_id
+                     AND roster.team_id != opponent.team_id
+                    WHERE opponent.steam_id_64 = ?
+                      AND roster.steam_id_64 IN ({roster_ph})
+                    """,
+                    [opponent['steam_id_64']] + self.roster_ids,
+                ).fetchone()[0]
+                self.assertGreater(faced, 0)
+        finally:
+            l2.close()
+
+
+class L3MartBuilderTests(unittest.TestCase):
+    def test_auxiliary_marts_build_on_database_copy(self):
+        from database.L3.L3_Builder import (
+            _get_team_players,
+            _rebuild_auxiliary_marts,
+        )
+
+        with tempfile.TemporaryDirectory() as temp_dir:
+            l3_path = os.path.join(temp_dir, 'L3.db')
+            shutil.copy2(Config.DB_L3_PATH, l3_path)
+            l2 = sqlite3.connect(Config.DB_L2_PATH)
+            l2.row_factory = sqlite3.Row
+            l3 = sqlite3.connect(l3_path)
+            try:
+                _rebuild_auxiliary_marts(
+                    l2,
+                    l3,
+                    sorted(_get_team_players()),
+                )
+                l3.commit()
+                for table in (
+                    'dm_player_match_history',
+                    'dm_player_map_stats',
+                    'dm_player_period_stats',
+                    'dm_player_records',
+                    'dm_player_weapon_stats',
+                ):
+                    count = l3.execute(
+                        f'SELECT COUNT(*) FROM {table}'
+                    ).fetchone()[0]
+                    self.assertGreater(count, 0)
+                self.assertEqual(
+                    l3.execute('PRAGMA quick_check').fetchone()[0],
+                    'ok',
+                )
+            finally:
+                l2.close()
+                l3.close()
+
+    def test_spatial_processor_does_not_emit_fake_geometry_metrics(self):
+        from database.L3.L3_Builder import _get_team_players
+        from database.L3.processors.intelligence_processor import IntelligenceProcessor
+
+        roster_ids = sorted(_get_team_players())
+        placeholders = ','.join('?' for _ in roster_ids)
+        l2 = sqlite3.connect(Config.DB_L2_PATH)
+        try:
+            row = l2.execute(
+                f"""
+                SELECT attacker_steam_id
+                FROM fact_round_events
+                WHERE attacker_steam_id IN ({placeholders})
+                  AND attacker_pos_x IS NOT NULL
+                GROUP BY attacker_steam_id
+                ORDER BY COUNT(*) DESC
+                LIMIT 1
+                """,
+                roster_ids,
+            ).fetchone()
+            self.assertIsNotNone(row)
+            features = IntelligenceProcessor._calculate_position_mastery(
+                str(row[0]),
+                l2,
+            )
+        finally:
+            l2.close()
+
+        for key in (
+            'int_pos_site_a_control_rate',
+            'int_pos_site_b_control_rate',
+            'int_pos_mid_control_rate',
+            'int_pos_rotation_speed',
+            'int_pos_lurk_tendency',
+            'int_pos_site_anchor_score',
+            'int_pos_retake_positioning',
+            'int_pos_postplant_positioning',
+            'int_pos_avg_distance_from_teammates',
+        ):
+            with self.subTest(key=key):
+                self.assertIsNone(features[key])
+        self.assertIsNotNone(features['int_pos_position_diversity'])
+
+    def test_percentiles_are_calculated_from_peer_scores(self):
+        from database.L3.L3_Builder import _update_percentiles
+
+        with tempfile.TemporaryDirectory() as temp_dir:
+            l3_path = os.path.join(temp_dir, 'L3.db')
+            shutil.copy2(Config.DB_L3_PATH, l3_path)
+            l3 = sqlite3.connect(l3_path)
+            try:
+                player_ids = [
+                    row[0] for row in l3.execute(
+                        """
+                        SELECT steam_id_64
+                        FROM dm_player_features
+                        ORDER BY steam_id_64
+                        LIMIT 3
+                        """
+                    )
+                ]
+                for steam_id, score in zip(player_ids, (10.0, 20.0, 30.0)):
+                    l3.execute(
+                        """
+                        UPDATE dm_player_features
+                        SET score_overall = ?
+                        WHERE steam_id_64 = ?
+                        """,
+                        [score, steam_id],
+                    )
+                _update_percentiles(l3, player_ids)
+                values = [
+                    row[0] for row in l3.execute(
+                        f"""
+                        SELECT tier_percentile
+                        FROM dm_player_features
+                        WHERE steam_id_64 IN ({','.join('?' for _ in player_ids)})
+                        ORDER BY score_overall
+                        """,
+                        player_ids,
+                    )
+                ]
+                self.assertEqual(values, [33.33, 66.67, 100.0])
+            finally:
+                l3.close()
+
+    def test_l3_backup_is_a_valid_sqlite_database(self):
+        from database.L3.L3_Builder import _backup_l3_database
+
+        with tempfile.TemporaryDirectory() as temp_dir:
+            source = os.path.join(temp_dir, 'source.db')
+            backup = os.path.join(temp_dir, 'backup.db')
+            shutil.copy2(Config.DB_L3_PATH, source)
+            result_path = _backup_l3_database(source, backup)
+            self.assertEqual(result_path, backup)
+            with sqlite3.connect(backup) as db:
+                self.assertEqual(db.execute('PRAGMA quick_check').fetchone()[0], 'ok')
+
+
+class FeatureFormulaTests(unittest.TestCase):
+    def test_empty_pace_inputs_do_not_receive_free_points(self):
+        from database.L3.processors.composite_processor import CompositeProcessor
+
+        self.assertEqual(CompositeProcessor._calculate_pace_score({}), 0.0)
+
+    def test_lower_map_and_elo_volatility_improves_stability_score(self):
+        from database.L3.processors.composite_processor import CompositeProcessor
+
+        common = {
+            'meta_rating_volatility': 0.2,
+            'meta_loss_rating': 1.0,
+            'meta_rating_consistency': 70,
+            'int_pressure_tilt_resistance': 0.8,
+            'meta_recent_form_rating': 1.15,
+        }
+        stable = dict(common, meta_map_stability=0.05, meta_elo_tier_stability=0.05)
+        volatile = dict(common, meta_map_stability=0.25, meta_elo_tier_stability=0.48)
+        self.assertGreater(
+            CompositeProcessor._calculate_stability_score(stable),
+            CompositeProcessor._calculate_stability_score(volatile),
+        )
+
+    def test_recent_form_uses_match_time_and_elo_stability_is_calculated(self):
+        from database.L3.L3_Builder import _get_team_players
+        from database.L3.processors.meta_processor import MetaProcessor
+
+        steam_id = sorted(_get_team_players())[0]
+        l2 = sqlite3.connect(Config.DB_L2_PATH)
+        try:
+            expected_rows = l2.execute(
+                """
+                SELECT p.rating
+                FROM fact_match_players p
+                JOIN fact_matches m ON m.match_id = p.match_id
+                WHERE p.steam_id_64 = ?
+                ORDER BY m.start_time DESC, p.match_id DESC
+                LIMIT 10
+                """,
+                [steam_id],
+            ).fetchall()
+            expected = sum(row[0] for row in expected_rows) / len(expected_rows)
+            features = MetaProcessor._calculate_stability(steam_id, l2)
+        finally:
+            l2.close()
+
+        self.assertAlmostEqual(
+            features['meta_recent_form_rating'],
+            round(expected, 3),
+            places=3,
+        )
+        self.assertGreaterEqual(features['meta_elo_tier_stability'], 0)
+        self.assertNotEqual(
+            features['meta_elo_tier_stability'],
+            features['meta_rating_volatility'],
+        )
+
+
+class DatabaseGovernanceTests(unittest.TestCase):
+    def test_database_paths_are_absolute_and_exist(self):
+        from database.paths import L1_DB, L2_DB, L3_DB, WEB_DB
+
+        for path in (L1_DB, L2_DB, L3_DB, WEB_DB):
+            with self.subTest(path=path):
+                self.assertTrue(path.is_absolute())
+                self.assertTrue(path.exists())
+
+    def test_valid_capture_is_identified_from_network_urls(self):
+        from database.paths import L1_DB
+        from web.services.import_service import MatchImportService
+
+        with sqlite3.connect(str(L1_DB)) as db:
+            match_id, raw = db.execute(
+                """
+                SELECT match_id, content
+                FROM raw_iframe_network
+                ORDER BY match_id
+                LIMIT 1
+                """
+            ).fetchone()
+        result = MatchImportService.validate_capture(raw.encode('utf-8'))
+        self.assertEqual(result['match_id'], match_id)
+        self.assertGreaterEqual(result['successful_responses'], 2)
+
+    def test_prepare_import_is_atomic_and_rejects_duplicate_queue(self):
+        import web.services.import_service as import_module
+        from database.paths import L1_DB
+        from web.services.import_service import (
+            DuplicateMatchError,
+            MatchImportService,
+        )
+
+        with sqlite3.connect(str(L1_DB)) as db:
+            original_id, raw = db.execute(
+                """
+                SELECT match_id, content
+                FROM raw_iframe_network
+                ORDER BY match_id
+                LIMIT 1
+                """
+            ).fetchone()
+        new_id = 'g161-99999999999999999999999'
+        raw = raw.replace(original_id, new_id).encode('utf-8')
+
+        with tempfile.TemporaryDirectory() as temp_dir:
+            old_l1 = import_module.L1_DB
+            old_arena = import_module.OUTPUT_ARENA
+            old_web = Config.DB_WEB_PATH
+            fake_l1 = Path(temp_dir) / 'L1.db'
+            fake_web = Path(temp_dir) / 'Web_App.sqlite'
+            with sqlite3.connect(str(fake_l1)) as db:
+                db.execute(
+                    """
+                    CREATE TABLE raw_iframe_network (
+                        match_id TEXT PRIMARY KEY,
+                        content TEXT
+                    )
+                    """
+                )
+            with sqlite3.connect(str(fake_web)) as db:
+                with open(Config.DB_WEB_SCHEMA_PATH, 'r', encoding='utf-8') as schema:
+                    db.executescript(schema.read())
+            import_module.L1_DB = fake_l1
+            import_module.OUTPUT_ARENA = Path(temp_dir) / 'output_arena'
+            Config.DB_WEB_PATH = str(fake_web)
+            try:
+                prepared = MatchImportService.prepare_import(
+                    raw,
+                    'iframe_network.json',
+                    created_by='test',
+                )
+                self.assertEqual(prepared['match_id'], new_id)
+                self.assertTrue(Path(prepared['source_path']).exists())
+                with self.assertRaises(DuplicateMatchError):
+                    MatchImportService.prepare_import(
+                        raw,
+                        'iframe_network.json',
+                        created_by='test',
+                    )
+            finally:
+                import_module.L1_DB = old_l1
+                import_module.OUTPUT_ARENA = old_arena
+                Config.DB_WEB_PATH = old_web
+
+    def test_pipeline_post_validation_accepts_current_databases(self):
+        from database.paths import L1_DB
+        from database.pipeline import _validate_pipeline_output
+
+        with sqlite3.connect(str(L1_DB)) as db:
+            match_id = db.execute(
+                'SELECT match_id FROM raw_iframe_network ORDER BY match_id LIMIT 1'
+            ).fetchone()[0]
+        _validate_pipeline_output(match_id)
+
+    def test_job_store_tracks_progress_logs_and_completion(self):
+        from database.job_store import JobStore
+
+        with tempfile.TemporaryDirectory() as temp_dir:
+            web_path = Path(temp_dir) / 'Web.sqlite'
+            with sqlite3.connect(str(web_path)) as db:
+                with open(Config.DB_WEB_SCHEMA_PATH, 'r', encoding='utf-8') as schema:
+                    db.executescript(schema.read())
+
+            store = JobStore(web_path)
+            job_id = store.create_job(
+                'test_pipeline',
+                match_id='g161-99999999999999999999999',
+                created_by='test',
+            )
+            store.start_job(job_id, 'backup', 'Starting')
+            store.update_progress(job_id, 'l2', 55, 'Building L2')
+            store.append_log(job_id, 'line one\n')
+            store.finish_job(job_id, True, 'Done', 1.25)
+
+            job = store.get_job(job_id)
+            self.assertEqual(job['status'], 'succeeded')
+            self.assertEqual(job['progress'], 100)
+            self.assertEqual(job['current_stage'], 'complete')
+            self.assertIn('line one', job['log_text'])
+            self.assertEqual(job['duration_seconds'], 1.25)
+
+    def test_database_backup_and_restore_round_trip(self):
+        from database.maintenance import backup_database, restore_database
+
+        with tempfile.TemporaryDirectory() as temp_dir:
+            source = Path(temp_dir) / 'source.db'
+            backup = Path(temp_dir) / 'backup.db'
+            with sqlite3.connect(str(source)) as db:
+                db.execute('CREATE TABLE values_table (value INTEGER)')
+                db.execute('INSERT INTO values_table VALUES (1)')
+
+            backup_database(source, backup)
+            with sqlite3.connect(str(source)) as db:
+                db.execute('UPDATE values_table SET value = 2')
+            restore_database(backup, source)
+
+            with sqlite3.connect(str(source)) as db:
+                value = db.execute(
+                    'SELECT value FROM values_table'
+                ).fetchone()[0]
+            self.assertEqual(value, 1)
+
+    def test_high_frequency_queries_use_operational_indexes(self):
+        l2 = sqlite3.connect(Config.DB_L2_PATH)
+        try:
+            plans = {
+                'player': l2.execute(
+                    """
+                    EXPLAIN QUERY PLAN
+                    SELECT * FROM fact_match_players
+                    WHERE steam_id_64 = ?
+                    """,
+                    ['76561198330488905'],
+                ).fetchone()[3],
+                'party': l2.execute(
+                    """
+                    EXPLAIN QUERY PLAN
+                    SELECT * FROM fact_match_players
+                    WHERE match_id = ? AND match_team_id = ?
+                    """,
+                    ['match', 1],
+                ).fetchone()[3],
+                'victim': l2.execute(
+                    """
+                    EXPLAIN QUERY PLAN
+                    SELECT * FROM fact_round_events
+                    WHERE victim_steam_id = ?
+                    """,
+                    ['player'],
+                ).fetchone()[3],
+            }
+        finally:
+            l2.close()
+
+        self.assertIn('idx_match_players_player_match', plans['player'])
+        self.assertIn('idx_match_players_party', plans['party'])
+        self.assertIn('idx_round_events_victim', plans['victim'])
+
+    def test_player_records_reference_real_matches(self):
+        l3 = sqlite3.connect(Config.DB_L3_PATH)
+        l2 = sqlite3.connect(Config.DB_L2_PATH)
+        try:
+            records = l3.execute(
+                """
+                SELECT match_id
+                FROM dm_player_records
+                WHERE match_id IS NOT NULL
+                """
+            ).fetchall()
+            self.assertGreater(len(records), 0)
+            for (match_id,) in records:
+                exists = l2.execute(
+                    'SELECT 1 FROM fact_matches WHERE match_id = ?',
+                    [match_id],
+                ).fetchone()
+                self.assertIsNotNone(exists)
+        finally:
+            l3.close()
+            l2.close()
+
+
+if __name__ == '__main__':
+    unittest.main()
@@ -1,20 +1,20 @@
 import sys
 import os

-# Add the project root directory to sys.path
 sys.path.append(os.path.abspath(os.path.join(os.path.dirname(__file__), '..')))

-from flask import Flask, render_template
+from flask import Flask
 from web.config import Config
-from web.database import close_dbs
+from web.database import close_dbs, initialize_web_db

-def create_app():
+
+def create_app(config_object=Config):
    app = Flask(__name__)
-    app.config.from_object(Config)
-    
+    app.config.from_object(config_object)
+
+    initialize_web_db()
    app.teardown_appcontext(close_dbs)
-    
-    # Register Blueprints
+
    from web.routes import main, matches, players, teams, tactics, admin, wiki, opponents
    app.register_blueprint(main.bp)
    app.register_blueprint(matches.bp)
@@ -24,13 +24,10 @@ def create_app():
    app.register_blueprint(admin.bp)
    app.register_blueprint(wiki.bp)
    app.register_blueprint(opponents.bp)
-    
-    @app.route('/')
-    def index():
-        return render_template('home/index.html')
-        
+
    return app

+
 if __name__ == '__main__':
    app = create_app()
    app.run(debug=True, port=5000)
@@ -1,14 +1,25 @@
 import os

+from database.paths import L2_DB, L3_DB, WEB_DB, WEB_SCHEMA
+
+
 class Config:
-    SECRET_KEY = os.environ.get('SECRET_KEY') or 'yrtv-secret-key-dev'
    BASE_DIR = os.path.abspath(os.path.join(os.path.dirname(__file__), '..'))
-    
-    DB_L2_PATH = os.path.join(BASE_DIR, 'database', 'L2', 'L2.db')
-    DB_L3_PATH = os.path.join(BASE_DIR, 'database', 'L3', 'L3.db')
-    DB_WEB_PATH = os.path.join(BASE_DIR, 'database', 'Web', 'Web_App.sqlite')
-    
-    ADMIN_TOKEN = 'jackyyang0929'
-    
+
+    SECRET_KEY = os.environ.get('SECRET_KEY', 'yrtv-dev-only-change-me')
+    ADMIN_TOKEN = os.environ.get('ADMIN_TOKEN', 'yrtv-admin-dev')
+
+    DB_L2_PATH = str(L2_DB)
+    DB_L3_PATH = str(L3_DB)
+    DB_WEB_PATH = str(WEB_DB)
+    DB_WEB_SCHEMA_PATH = str(WEB_SCHEMA)
+    WEB_SCHEMA_VERSION = 2
+
+    MAX_CONTENT_LENGTH = 5 * 1024 * 1024
+    SQLITE_TIMEOUT_SECONDS = 15
+    SLOW_QUERY_THRESHOLD_SECONDS = float(
+        os.environ.get('SLOW_QUERY_THRESHOLD_SECONDS', '0.25')
+    )
+
    # Pagination
    ITEMS_PER_PAGE = 20
@@ -1,7 +1,96 @@
+import os
+import logging
 import sqlite3
+import time
+
 from flask import g
 from web.config import Config

+logger = logging.getLogger(__name__)
+
+
+def _database_path(db_name):
+    paths = {
+        'l2': Config.DB_L2_PATH,
+        'l3': Config.DB_L3_PATH,
+        'web': Config.DB_WEB_PATH,
+    }
+    try:
+        return paths[db_name]
+    except KeyError as exc:
+        raise ValueError(f"Unknown database: {db_name}") from exc
+
+
+def initialize_web_db():
+    """Create and migrate the small application-owned database."""
+    os.makedirs(os.path.dirname(Config.DB_WEB_PATH), exist_ok=True)
+    db = sqlite3.connect(
+        Config.DB_WEB_PATH,
+        timeout=Config.SQLITE_TIMEOUT_SECONDS,
+    )
+    try:
+        table_exists = db.execute(
+            "SELECT 1 FROM sqlite_master WHERE type='table' AND name='team_lineups'"
+        ).fetchone()
+        if table_exists:
+            columns = {
+                row[1] for row in db.execute("PRAGMA table_info(team_lineups)")
+            }
+            if 'is_active' not in columns:
+                db.execute(
+                    "ALTER TABLE team_lineups "
+                    "ADD COLUMN is_active INTEGER NOT NULL DEFAULT 0"
+                )
+                first_id = db.execute(
+                    "SELECT id FROM team_lineups "
+                    "ORDER BY created_at DESC, id DESC LIMIT 1"
+                ).fetchone()
+                if first_id:
+                    db.execute(
+                        "UPDATE team_lineups SET is_active = 1 WHERE id = ?",
+                        first_id,
+                    )
+            else:
+                active_ids = [
+                    row[0] for row in db.execute(
+                        "SELECT id FROM team_lineups WHERE is_active = 1 "
+                        "ORDER BY created_at DESC, id DESC"
+                    )
+                ]
+                if not active_ids:
+                    latest_id = db.execute(
+                        "SELECT id FROM team_lineups "
+                        "ORDER BY created_at DESC, id DESC LIMIT 1"
+                    ).fetchone()
+                    if latest_id:
+                        db.execute(
+                            "UPDATE team_lineups SET is_active = 1 WHERE id = ?",
+                            latest_id,
+                        )
+                elif len(active_ids) > 1:
+                    db.execute("UPDATE team_lineups SET is_active = 0")
+                    db.execute(
+                        "UPDATE team_lineups SET is_active = 1 WHERE id = ?",
+                        [active_ids[0]],
+                    )
+
+        with open(Config.DB_WEB_SCHEMA_PATH, 'r', encoding='utf-8') as schema_file:
+            db.executescript(schema_file.read())
+        db.execute(
+            """
+            INSERT OR IGNORE INTO schema_migrations (version, description)
+            VALUES (?, ?)
+            """,
+            [
+                Config.WEB_SCHEMA_VERSION,
+                'ETL jobs, match imports and active lineup governance',
+            ],
+        )
+        db.commit()
+    finally:
+        db.close()
+
+
 def get_db(db_name):
    """
    db_name: 'l2', 'l3', or 'web'
@@ -10,18 +99,20 @@ def get_db(db_name):
    db = getattr(g, db_attr, None)
    
    if db is None:
-        if db_name == 'l2':
-            path = Config.DB_L2_PATH
-        elif db_name == 'l3':
-            path = Config.DB_L3_PATH
-        elif db_name == 'web':
-            path = Config.DB_WEB_PATH
-        else:
-            raise ValueError(f"Unknown database: {db_name}")
-            
-        # Connect with check_same_thread=False if needed for dev, but default is safer per thread
-        db = sqlite3.connect(path)
+        path = _database_path(db_name)
+        if db_name != 'web' and not os.path.exists(path):
+            raise RuntimeError(
+                f"{db_name.upper()} database does not exist: {path}. "
+                "Run the corresponding data builder first."
+            )
+        db = sqlite3.connect(
+            path,
+            timeout=Config.SQLITE_TIMEOUT_SECONDS,
+        )
        db.row_factory = sqlite3.Row
+        db.execute("PRAGMA busy_timeout = 15000")
+        if db_name != 'l3':
+            db.execute("PRAGMA foreign_keys = ON")
        setattr(g, db_attr, db)
        
    return db
@@ -34,14 +125,34 @@ def close_dbs(e=None):
            db.close()

 def query_db(db_name, query, args=(), one=False):
+    started = time.perf_counter()
    cur = get_db(db_name).execute(query, args)
-    rv = cur.fetchall()
-    cur.close()
+    try:
+        rv = cur.fetchall()
+    finally:
+        cur.close()
+    duration = time.perf_counter() - started
+    if duration >= Config.SLOW_QUERY_THRESHOLD_SECONDS:
+        logger.warning(
+            "Slow query db=%s duration=%.3fs sql=%s",
+            db_name,
+            duration,
+            " ".join(query.split())[:500],
+        )
    return (rv[0] if rv else None) if one else rv

 def execute_db(db_name, query, args=()):
    db = get_db(db_name)
+    started = time.perf_counter()
    cur = db.execute(query, args)
    db.commit()
+    duration = time.perf_counter() - started
+    if duration >= Config.SLOW_QUERY_THRESHOLD_SECONDS:
+        logger.warning(
+            "Slow write db=%s duration=%.3fs sql=%s",
+            db_name,
+            duration,
+            " ".join(query.split())[:500],
+        )
    cur.close()
    return cur.lastrowid
@@ -1,16 +1,17 @@
-from flask import Blueprint, render_template, request, redirect, url_for, session, flash
+from flask import Blueprint, render_template, request, redirect, url_for, session, flash, jsonify
 from web.config import Config
 from web.auth import admin_required
 from web.database import query_db
-import os
+from web.services.etl_service import EtlService
+import hmac

 bp = Blueprint('admin', __name__, url_prefix='/admin')

@bp.route('/login', methods=['GET', 'POST'])
 def login():
    if request.method == 'POST':
-        token = request.form.get('token')
-        if token == Config.ADMIN_TOKEN:
+        token = request.form.get('token') or ''
+        if hmac.compare_digest(token, Config.ADMIN_TOKEN):
            session['is_admin'] = True
            return redirect(url_for('admin.dashboard'))
        else:
@@ -27,19 +28,104 @@ def logout():
 def dashboard():
    return render_template('admin/dashboard.html')

-from web.services.etl_service import EtlService
+@bp.route('/data-integrity')
+@admin_required
+def data_integrity():
+    from web.services.integrity_service import IntegrityService
+
+    report = IntegrityService.build_report()
+    if request.args.get('format') == 'json':
+        return jsonify(report)
+    return render_template('admin/data_integrity.html', report=report)

@bp.route('/trigger_etl', methods=['POST'])
@admin_required
 def trigger_etl():
-    script_name = request.form.get('script')
-    allowed = ['L1A.py', 'L2_Builder.py', 'L3_Builder.py']
-    if script_name not in allowed:
-         return "Invalid script", 400
-         
-    success, message = EtlService.run_script(script_name)
-    status_code = 200 if success else 500
-    return message, status_code
+    from database.job_store import JobStore
+
+    store = JobStore(Config.DB_WEB_PATH)
+    job_id = store.create_job(
+        'manual_pipeline',
+        created_by='admin',
+    )
+    try:
+        pid = EtlService.start_pipeline(job_id)
+    except Exception as exc:
+        store.finish_job(job_id, False, str(exc), 0)
+        return jsonify({'success': False, 'error': str(exc)}), 500
+    return jsonify({'success': True, 'job_id': job_id, 'pid': pid}), 202
+
+
+@bp.route('/import-match', methods=['GET', 'POST'])
+@admin_required
+def import_match():
+    from database.job_store import JobStore
+    from web.services.import_service import (
+        DuplicateMatchError,
+        ImportValidationError,
+        MatchImportService,
+    )
+
+    store = JobStore(Config.DB_WEB_PATH)
+    if request.method == 'POST':
+        upload = request.files.get('capture')
+        if not upload or not upload.filename:
+            flash('请选择 iframe_network.json 文件。', 'error')
+            return redirect(url_for('admin.import_match'))
+
+        prepared = None
+        try:
+            prepared = MatchImportService.prepare_import(
+                upload.read(),
+                upload.filename,
+                created_by='admin',
+                replace=request.form.get('replace') == '1',
+            )
+            EtlService.start_pipeline(
+                prepared['job_id'],
+                match_id=prepared['match_id'],
+                replace=prepared['replace'],
+            )
+            flash(
+                f"比赛 {prepared['match_id']} 已进入导入队列。",
+                'success',
+            )
+            return redirect(url_for(
+                'admin.import_match',
+                job_id=prepared['job_id'],
+            ))
+        except (DuplicateMatchError, ImportValidationError) as exc:
+            flash(str(exc), 'warning')
+        except Exception as exc:
+            if prepared:
+                store.finish_job(
+                    prepared['job_id'],
+                    False,
+                    f'Failed to start pipeline: {exc}',
+                    0,
+                )
+            flash(f'启动导入失败：{exc}', 'error')
+
+    selected_job = None
+    selected_job_id = request.args.get('job_id', type=int)
+    if selected_job_id:
+        selected_job = store.get_job(selected_job_id)
+    return render_template(
+        'admin/import_match.html',
+        jobs=store.list_jobs(30),
+        selected_job=selected_job,
+    )
+
+
+@bp.route('/api/jobs/<int:job_id>')
+@admin_required
+def api_job(job_id):
+    from database.job_store import JobStore
+
+    job = JobStore(Config.DB_WEB_PATH).get_job(job_id)
+    if not job:
+        return jsonify({'error': 'Job not found'}), 404
+    return jsonify(job)

@bp.route('/sql', methods=['GET', 'POST'])
@admin_required
@@ -50,18 +136,21 @@ def sql_runner():
    db_name = "l2"
    
    if request.method == 'POST':
-        query = request.form.get('query')
+        query = (request.form.get('query') or '').strip()
        db_name = request.form.get('db_name', 'l2')
-        
-        # Safety check
-        forbidden = ['DELETE', 'DROP', 'UPDATE', 'INSERT', 'ALTER', 'GRANT', 'REVOKE']
-        if any(x in query.upper() for x in forbidden):
-             error = "Only SELECT queries allowed in Web Runner."
+
+        statement = query.rstrip(';').strip()
+        if db_name not in {'l2', 'l3', 'web'}:
+            error = "Unknown database."
+        elif not statement.upper().startswith('SELECT '):
+            error = "Only SELECT queries are allowed."
+        elif ';' in statement:
+            error = "Only one SQL statement is allowed."
        else:
            try:
-                # Enforce limit if not present
-                if 'LIMIT' not in query.upper():
-                    query += " LIMIT 50"
+                query = statement
+                if 'LIMIT' not in statement.upper():
+                    query = f"{statement} LIMIT 50"
                
                rows = query_db(db_name, query)
                if rows:
@@ -1,6 +1,5 @@
 from flask import Blueprint, render_template, request, jsonify
 from web.services.stats_service import StatsService
-import time

 bp = Blueprint('main', __name__)

@@ -18,18 +17,17 @@ def index():
            
    return render_template('home/index.html', recent_matches=recent_matches, heatmap_data=heatmap_data, live_matches=live_matches)

-from web.services.etl_service import EtlService
-
@bp.route('/parse_match', methods=['POST'])
 def parse_match():
    url = request.form.get('url')
    if not url or '5eplay.com' not in url:
-        return jsonify({'success': False, 'message': 'Invalid 5EPlay URL'})
-    
-    # Trigger L1A.py with URL argument
-    success, msg = EtlService.run_script('L1A.py', args=[url])
-    
-    if success:
-        return jsonify({'success': True, 'message': 'Match parsing completed successfully!'})
-    else:
-        return jsonify({'success': False, 'message': f'Error: {msg}'})
+        return jsonify({'success': False, 'message': 'Invalid 5EPlay URL'}), 400
+
+    return jsonify({
+        'success': False,
+        'message': (
+            'URL downloader is not included in this repository. '
+            'Place iframe_network.json under output_arena/<match_id>/ '
+            'and run the L1/L2/L3 builders from Admin.'
+        ),
+    }), 501
@@ -33,19 +33,9 @@ def detail(match_id):
    
    rounds = StatsService.get_match_rounds(match_id)
    
-    # --- Roster Identification ---
-    # Fetch active roster to identify "Our Team" players
-    from web.services.web_service import WebService
-    lineups = WebService.get_lineups()
-    # Assume we use the first/active lineup
-    active_roster_ids = []
-    if lineups:
-        try:
-            active_roster_ids = json.loads(lineups[0]['player_ids_json'])
-        except:
-            pass
-            
-    # Mark roster players (Ensure strict string comparison)
+    from web.services.team_context_service import TeamContextService
+
+    active_roster_ids = TeamContextService.get_active_roster_ids()
    roster_set = set(str(uid) for uid in active_roster_ids)
    for p in players:
        p['is_in_roster'] = str(p['steam_id_64']) in roster_set
@@ -1,15 +1,16 @@
 from flask import Blueprint, render_template, request, jsonify, redirect, url_for, flash, current_app, session
 from web.services.stats_service import StatsService
 from web.services.feature_service import FeatureService
+from web.services.player_profile_service import PlayerProfileService
 from web.services.web_service import WebService
 from web.database import execute_db, query_db
 from web.config import Config
 from datetime import datetime
 import os
 import json
-from werkzeug.utils import secure_filename

 bp = Blueprint('players', __name__, url_prefix='/players')
+ALLOWED_AVATAR_EXTENSIONS = {'.jpg', '.jpeg', '.png', '.webp'}

@bp.route('/')
 def index():
@@ -41,7 +42,12 @@ def detail(steam_id):
                            # Use steam_id as filename to ensure uniqueness per player
                            # Preserve extension
                            ext = os.path.splitext(file.filename)[1].lower()
-                            if not ext: ext = '.jpg'
+                            if (
+                                ext not in ALLOWED_AVATAR_EXTENSIONS
+                                or not (file.mimetype or '').startswith('image/')
+                            ):
+                                flash('Avatar must be a JPG, PNG, or WebP image.', 'error')
+                                return redirect(url_for('players.detail', steam_id=steam_id))
                            
                            filename = f"{steam_id}{ext}"
                            upload_folder = os.path.join(current_app.root_path, 'static', 'avatars')
@@ -177,33 +183,9 @@ def detail(steam_id):
    history_asc = StatsService.get_player_trend(steam_id, limit=1000) 
    history = history_asc[::-1] if history_asc else []
    
-    # Calculate Map Stats
-    map_stats = {}
-    for match in history:
-        m_name = match['map_name']
-        if m_name not in map_stats:
-            map_stats[m_name] = {'matches': 0, 'wins': 0, 'adr_sum': 0, 'rating_sum': 0}
-        
-        map_stats[m_name]['matches'] += 1
-        if match['is_win']:
-            map_stats[m_name]['wins'] += 1
-        map_stats[m_name]['adr_sum'] += (match['adr'] or 0)
-        map_stats[m_name]['rating_sum'] += (match['rating'] or 0)
-    
-    map_stats_list = []
-    for m_name, data in map_stats.items():
-        cnt = data['matches']
-        map_stats_list.append({
-            'map_name': m_name,
-            'matches': cnt,
-            'win_rate': data['wins'] / cnt,
-            'adr': data['adr_sum'] / cnt,
-            'rating': data['rating_sum'] / cnt
-        })
-    map_stats_list.sort(key=lambda x: x['matches'], reverse=True)
-
-    # --- New: Recent Performance Stats ---
-    # recent_stats = StatsService.get_recent_performance_stats(steam_id)
+    map_stats_list = PlayerProfileService.get_map_stats(steam_id)
+    period_stats = PlayerProfileService.get_period_stats(steam_id)
+    records = PlayerProfileService.get_records(steam_id)

    return render_template('players/profile.html',
                         player=player,
@@ -213,6 +195,8 @@ def detail(steam_id):
                         history=history,
                         distribution=distribution,
                         map_stats=map_stats_list,
+                         period_stats=period_stats,
+                         records=records,
                         l2_stats=l2_stats,
                         side_stats=side_stats)

@@ -223,9 +207,13 @@ def like_comment(comment_id):

@bp.route('/<steam_id>/charts_data')
 def charts_data(steam_id):
-    # ... (existing code) ...
-    # Trend Data
-    trends = StatsService.get_player_trend(steam_id, limit=1000)
+    period_key = request.args.get('period', 'last_20')
+    if period_key not in PlayerProfileService.PERIOD_KEYS:
+        period_key = 'last_20'
+    trends = PlayerProfileService.get_period_history(steam_id, period_key)
+    if not trends:
+        trends = StatsService.get_player_trend(steam_id, limit=20)
+    period_summary = PlayerProfileService.get_period(steam_id, period_key)
    
    # Radar Data (Construct from features)
    features = FeatureService.get_player_features(steam_id)
@@ -234,17 +222,11 @@ def charts_data(steam_id):
    
    # Task 1: Strict Team Average Calculation
    team_avg_radar = None
-    lineups = WebService.get_lineups()
-    if lineups:
-        target_lineup = None
-        try:
-            p_ids = [str(i) for i in json.loads(lineups[0].get("player_ids_json") or "[]")]
-            if str(steam_id) in p_ids:
-                target_lineup = p_ids
-        except:
-            target_lineup = None
-        
-        if target_lineup:
+    from web.services.team_context_service import TeamContextService
+
+    active_roster_ids = TeamContextService.get_active_roster_ids()
+    target_lineup = active_roster_ids if str(steam_id) in active_roster_ids else None
+    if target_lineup:
            # Calculate strict average for this lineup
            team_sums = {
                'score_aim': 0.0, 'score_defense': 0.0, 'score_utility': 0.0,
@@ -303,9 +285,19 @@ def charts_data(steam_id):
        'trend': {'labels': trend_labels, 'values': trend_values},
        'radar': radar_data,
        'radar_dist': radar_dist,
-        'team_avg_radar': team_avg_radar
+        'team_avg_radar': team_avg_radar,
+        'period': period_summary,
    })

+
+@bp.route('/<steam_id>/period_stats')
+def period_stats(steam_id):
+    period_key = request.args.get('period', 'last_20')
+    period = PlayerProfileService.get_period(steam_id, period_key)
+    if not period:
+        return jsonify({'error': 'Unknown period or player'}), 404
+    return jsonify(period)
+
 # --- API for Comparison ---
@bp.route('/api/search')
 def api_search():
@@ -67,14 +67,12 @@ def api_search():

@bp.route('/api/roster', methods=['GET', 'POST'])
 def api_roster():
-    # Assume single team mode, always operating on ID=1 or the first lineup
-    lineups = WebService.get_lineups()
-    if not lineups:
-        # Auto-create default team if none exists
+    target_team = WebService.get_active_lineup()
+    if not target_team:
        WebService.save_lineup("My Team", "Default Roster", [])
-        lineups = WebService.get_lineups()
-    
-    target_team = dict(lineups[0]) # Get the latest one
+        target_team = WebService.get_active_lineup()
+
+    target_team = dict(target_team)
    
    if request.method == 'POST':
        # Admin Check
@@ -4,13 +4,51 @@ import sys
 from web.config import Config

 class EtlService:
+    SCRIPT_PATHS = {
+        'L1A.py': os.path.join('database', 'L1', 'L1_Builder.py'),
+        'L2_Builder.py': os.path.join('database', 'L2', 'L2_Builder.py'),
+        'L3_Builder.py': os.path.join('database', 'L3', 'L3_Builder.py'),
+    }
+
+    @staticmethod
+    def start_pipeline(job_id, match_id=None, replace=False):
+        script_path = os.path.join(
+            Config.BASE_DIR,
+            'database',
+            'pipeline.py',
+        )
+        command = [
+            sys.executable,
+            script_path,
+            '--job-id',
+            str(int(job_id)),
+        ]
+        if match_id:
+            command.extend(['--match-id', str(match_id)])
+        if replace:
+            command.append('--replace')
+
+        process = subprocess.Popen(
+            command,
+            cwd=Config.BASE_DIR,
+            stdin=subprocess.DEVNULL,
+            stdout=subprocess.DEVNULL,
+            stderr=subprocess.DEVNULL,
+            start_new_session=True,
+        )
+        return process.pid
+
    @staticmethod
    def run_script(script_name, args=None):
        """
-        Executes an ETL script located in the ETL directory.
+        Executes an allow-listed data builder from its actual repository path.
        Returns (success, message)
        """
-        script_path = os.path.join(Config.BASE_DIR, 'ETL', script_name)
+        relative_path = EtlService.SCRIPT_PATHS.get(script_name)
+        if not relative_path:
+            return False, f"Unsupported data script: {script_name}"
+
+        script_path = os.path.join(Config.BASE_DIR, relative_path)
        
        if not os.path.exists(script_path):
            return False, f"Script not found: {script_path}"
@@ -28,7 +66,7 @@ class EtlService:
                cwd=Config.BASE_DIR,
                capture_output=True,
                text=True,
-                timeout=300 # 5 min timeout
+                timeout=900
            )
            
            if result.returncode == 0:
@@ -1,6 +1,6 @@
 from __future__ import annotations

-from typing import Any, Iterable
+from typing import Any

 from web.database import query_db

@@ -138,20 +138,47 @@ class FeatureService:
        }
        order_col = sort_map.get(sort_by, "core_avg_rating")

-        where = []
-        args: list[Any] = []
        if search:
-            where.append("steam_id_64 IN (SELECT steam_id_64 FROM dim_players WHERE username LIKE ?)")
-            args.append(f"%{search}%")
-        where_sql = f"WHERE {' AND '.join(where)}" if where else ""
-
-        rows = query_db(
-            "l3",
-            f"SELECT * FROM dm_player_features {where_sql} ORDER BY {order_col} DESC LIMIT ? OFFSET ?",
-            args + [per_page, offset],
-        )
-        total_row = query_db("l3", f"SELECT COUNT(*) as cnt FROM dm_player_features {where_sql}", args, one=True)
-        total = int(total_row["cnt"]) if total_row else 0
+            dim_rows = query_db(
+                "l2",
+                """
+                SELECT steam_id_64
+                FROM dim_players
+                WHERE LOWER(username) LIKE LOWER(?) OR steam_id_64 LIKE ?
+                """,
+                [f"%{search}%", f"%{search}%"],
+            )
+            matching_ids = [str(row["steam_id_64"]) for row in dim_rows]
+            rows = []
+            for start in range(0, len(matching_ids), 500):
+                chunk = matching_ids[start:start + 500]
+                placeholders = ",".join("?" for _ in chunk)
+                rows.extend(query_db(
+                    "l3",
+                    f"SELECT * FROM dm_player_features "
+                    f"WHERE steam_id_64 IN ({placeholders})",
+                    chunk,
+                ))
+            rows = sorted(
+                rows,
+                key=lambda row: row[order_col] if row[order_col] is not None else float("-inf"),
+                reverse=True,
+            )
+            total = len(rows)
+            rows = rows[offset:offset + per_page]
+        else:
+            rows = query_db(
+                "l3",
+                f"SELECT * FROM dm_player_features "
+                f"ORDER BY {order_col} DESC LIMIT ? OFFSET ?",
+                [per_page, offset],
+            )
+            total_row = query_db(
+                "l3",
+                "SELECT COUNT(*) as cnt FROM dm_player_features",
+                one=True,
+            )
+            total = int(total_row["cnt"]) if total_row else 0

        players = [FeatureService._normalize_features(dict(r)) for r in rows] if rows else []
        players = [p for p in players if p]
@@ -160,19 +187,11 @@ class FeatureService:

    @staticmethod
    def get_roster_features_distribution(target_steam_id: str):
-        from web.services.web_service import WebService
-        import json
+        from web.services.team_context_service import TeamContextService

-        lineups = WebService.get_lineups()
-        roster_ids: list[str] = []
-        
-        if lineups:
-            try:
-                p_ids = [str(i) for i in json.loads(lineups[0].get("player_ids_json") or "[]")]
-                if str(target_steam_id) in p_ids:
-                    roster_ids = p_ids
-            except Exception:
-                roster_ids = []
+        roster_ids = TeamContextService.get_active_roster_ids()
+        if str(target_steam_id) not in roster_ids:
+            roster_ids = []

        if not roster_ids:
            return None
@@ -202,7 +221,17 @@ class FeatureService:
                sample_keys = list(p.keys())
                break
        
-        lower_is_better = {"int_timing_first_contact_time", "tac_avg_fd", "core_avg_match_duration"}
+        lower_is_better = {
+            "int_timing_first_contact_time",
+            "int_trade_response_time",
+            "tac_avg_fd",
+            "tac_fd_rate",
+            "core_avg_match_duration",
+            "core_dpr",
+            "meta_rating_volatility",
+            "meta_map_stability",
+            "meta_elo_tier_stability",
+        }

        result: dict[str, Any] = {}
        for m in sample_keys:
@@ -224,16 +253,22 @@ class FeatureService:
            values = []
            for p in stats_map.values():
                v = (p or {}).get(m)
+                if v is None:
+                    continue
                try:
-                    values.append(float(v) if v is not None else 0.0)
+                    values.append(float(v))
                except (ValueError, TypeError):
-                    values.append(0.0)
+                    continue

            target_val_raw = (stats_map.get(target_steam_id) or {}).get(m)
+            if target_val_raw is None or not values:
+                result[m] = None
+                continue
            try:
-                target_val = float(target_val_raw) if target_val_raw is not None else 0.0
+                target_val = float(target_val_raw)
            except (ValueError, TypeError):
-                target_val = 0.0
+                result[m] = None
+                continue

            is_reverse = m not in lower_is_better
            # Sort values. For standard metrics, higher is better (reverse=True).
@@ -251,9 +286,9 @@ class FeatureService:
                "val": target_val,
                "rank": rank,
                "total": len(values_sorted),
-                "min": min(values_sorted) if values_sorted else 0,
-                "max": max(values_sorted) if values_sorted else 0,
-                "avg": (sum(values_sorted) / len(values_sorted)) if values_sorted else 0,
+                "min": min(values_sorted),
+                "max": max(values_sorted),
+                "avg": sum(values_sorted) / len(values_sorted),
                "inverted": not is_reverse,
            }
        return result
@@ -0,0 +1,186 @@
+import hashlib
+import json
+import os
+from pathlib import Path
+import re
+import sqlite3
+from typing import Any, Dict
+
+from database.job_store import JobStore
+from database.paths import L1_DB, OUTPUT_ARENA
+from web.config import Config
+
+
+MATCH_ID_PATTERN = re.compile(r'\bg161-[0-9]{10,}\b')
+
+
+class ImportValidationError(ValueError):
+    pass
+
+
+class DuplicateMatchError(ImportValidationError):
+    pass
+
+
+class MatchImportService:
+    @staticmethod
+    def validate_capture(raw_bytes: bytes) -> Dict[str, Any]:
+        if not raw_bytes:
+            raise ImportValidationError('Uploaded file is empty')
+
+        try:
+            text = raw_bytes.decode('utf-8-sig')
+        except UnicodeDecodeError as exc:
+            raise ImportValidationError('Capture must be UTF-8 JSON') from exc
+
+        try:
+            capture = json.loads(text)
+        except json.JSONDecodeError as exc:
+            raise ImportValidationError(
+                f'Invalid JSON at line {exc.lineno}, column {exc.colno}'
+            ) from exc
+
+        if not isinstance(capture, list) or not capture:
+            raise ImportValidationError(
+                'Capture root must be a non-empty list of network responses'
+            )
+
+        urls = []
+        successful_responses = 0
+        for index, item in enumerate(capture):
+            if not isinstance(item, dict):
+                raise ImportValidationError(
+                    f'Capture item {index} must be an object'
+                )
+            url = item.get('url')
+            if not isinstance(url, str) or not url:
+                raise ImportValidationError(
+                    f'Capture item {index} has no URL'
+                )
+            urls.append(url)
+            if item.get('status') == 200 and item.get('body') is not None:
+                successful_responses += 1
+
+        match_ids = sorted({
+            match.group(0)
+            for url in urls
+            for match in MATCH_ID_PATTERN.finditer(url)
+        })
+        if len(match_ids) != 1:
+            raise ImportValidationError(
+                f'Capture must reference exactly one match ID; found {match_ids}'
+            )
+        if successful_responses < 2:
+            raise ImportValidationError(
+                'Capture does not contain enough successful API responses'
+            )
+
+        match_id = match_ids[0]
+        has_match_data = any(
+            f'/api/data/match/{match_id}' in url for url in urls
+        )
+        has_round_data = any(
+            f'/api/match/round/{match_id}' in url for url in urls
+        )
+        if not has_match_data or not has_round_data:
+            missing = []
+            if not has_match_data:
+                missing.append('match data')
+            if not has_round_data:
+                missing.append('round data')
+            raise ImportValidationError(
+                f"Capture is missing required endpoint(s): {', '.join(missing)}"
+            )
+
+        return {
+            'match_id': match_id,
+            'content_sha256': hashlib.sha256(raw_bytes).hexdigest(),
+            'response_count': len(capture),
+            'successful_responses': successful_responses,
+            'text': text,
+        }
+
+    @staticmethod
+    def _existing_l1_content(match_id: str):
+        if not L1_DB.exists():
+            return None
+        db = sqlite3.connect(str(L1_DB))
+        try:
+            row = db.execute(
+                """
+                SELECT content
+                FROM raw_iframe_network
+                WHERE match_id = ?
+                """,
+                [match_id],
+            ).fetchone()
+            return row[0] if row else None
+        finally:
+            db.close()
+
+    @staticmethod
+    def prepare_import(
+        raw_bytes: bytes,
+        original_filename: str,
+        created_by: str,
+        replace: bool = False,
+    ):
+        validation = MatchImportService.validate_capture(raw_bytes)
+        match_id = validation['match_id']
+        content_hash = validation['content_sha256']
+
+        existing_content = MatchImportService._existing_l1_content(match_id)
+        if existing_content is not None:
+            existing_hash = hashlib.sha256(
+                existing_content.encode('utf-8')
+            ).hexdigest()
+            if existing_hash == content_hash:
+                raise DuplicateMatchError(
+                    f'Match {match_id} is already imported with identical data'
+                )
+            if not replace:
+                raise DuplicateMatchError(
+                    f'Match {match_id} already exists with different data; '
+                    'explicit replacement is required'
+                )
+
+        match_dir = OUTPUT_ARENA / match_id
+        match_dir.mkdir(parents=True, exist_ok=True)
+        destination = match_dir / 'iframe_network.json'
+        if destination.exists() and not replace:
+            current_hash = hashlib.sha256(destination.read_bytes()).hexdigest()
+            if current_hash == content_hash:
+                raise DuplicateMatchError(
+                    f'Match {match_id} is already queued with identical data'
+                )
+            raise DuplicateMatchError(
+                f'Pending capture already exists for {match_id}'
+            )
+
+        temporary = destination.with_suffix('.json.tmp')
+        temporary.write_bytes(raw_bytes)
+        os.replace(str(temporary), str(destination))
+
+        store = JobStore(Config.DB_WEB_PATH)
+        job_id = store.create_job(
+            'match_import',
+            match_id=match_id,
+            input_path=str(destination),
+            created_by=created_by,
+        )
+        store.upsert_match_import(
+            match_id,
+            content_hash,
+            str(destination),
+            'queued',
+            job_id,
+        )
+        return {
+            'job_id': job_id,
+            'match_id': match_id,
+            'content_sha256': content_hash,
+            'response_count': validation['response_count'],
+            'source_path': str(destination),
+            'original_filename': Path(original_filename or '').name,
+            'replace': bool(replace),
+        }
@@ -0,0 +1,511 @@
+from datetime import datetime, timezone
+import json
+import os
+import sqlite3
+
+from database.maintenance import backup_storage_status
+from web.config import Config
+from web.services.team_context_service import TeamContextService
+
+
+class IntegrityService:
+    DATABASES = {
+        'L2': Config.DB_L2_PATH,
+        'L3': Config.DB_L3_PATH,
+        'Web': Config.DB_WEB_PATH,
+    }
+
+    @staticmethod
+    def _check(checks, name, status, detail, value=None):
+        checks.append({
+            'name': name,
+            'status': status,
+            'detail': detail,
+            'value': value,
+        })
+
+    @staticmethod
+    def _connect(path):
+        db = sqlite3.connect(path, timeout=Config.SQLITE_TIMEOUT_SECONDS)
+        db.row_factory = sqlite3.Row
+        return db
+
+    @staticmethod
+    def build_report():
+        checks = []
+        counts = {}
+        connections = {}
+
+        try:
+            for name, path in IntegrityService.DATABASES.items():
+                if not os.path.exists(path):
+                    IntegrityService._check(
+                        checks,
+                        f'{name} database',
+                        'fail',
+                        f'Missing file: {path}',
+                    )
+                    continue
+                try:
+                    db = IntegrityService._connect(path)
+                    connections[name] = db
+                    result = db.execute('PRAGMA quick_check').fetchone()[0]
+                    IntegrityService._check(
+                        checks,
+                        f'{name} database',
+                        'pass' if result == 'ok' else 'fail',
+                        f'quick_check: {result}',
+                        os.path.getsize(path),
+                    )
+                except sqlite3.Error as exc:
+                    IntegrityService._check(
+                        checks,
+                        f'{name} database',
+                        'fail',
+                        str(exc),
+                    )
+
+            l2 = connections.get('L2')
+            if l2:
+                IntegrityService._check_l2(l2, checks, counts)
+
+            l3 = connections.get('L3')
+            roster_ids = TeamContextService.get_active_roster_ids()
+            counts['active_roster'] = len(roster_ids)
+            if l3:
+                IntegrityService._check_l3(l3, roster_ids, checks, counts)
+
+            web = connections.get('Web')
+            if web:
+                IntegrityService._check_web(web, checks, counts)
+
+            backup_status = backup_storage_status()
+            counts['backup_sets'] = backup_status['sets']
+            counts['backup_bytes'] = backup_status['total_bytes']
+            IntegrityService._check(
+                checks,
+                'Backup retention',
+                'warn' if backup_status['sets'] > 3 else 'pass',
+                (
+                    f"{backup_status['sets']} backup sets, "
+                    f"{backup_status['total_bytes']:,} bytes"
+                ),
+                backup_status['sets'],
+            )
+        finally:
+            for db in connections.values():
+                db.close()
+
+        status_order = {'pass': 0, 'warn': 1, 'fail': 2}
+        overall_status = max(
+            (check['status'] for check in checks),
+            key=lambda status: status_order[status],
+            default='fail',
+        )
+        return {
+            'generated_at': datetime.now(timezone.utc).isoformat(),
+            'overall_status': overall_status,
+            'counts': counts,
+            'checks': checks,
+            'totals': {
+                status: sum(1 for check in checks if check['status'] == status)
+                for status in ('pass', 'warn', 'fail')
+            },
+        }
+
+    @staticmethod
+    def _table_names(db):
+        return {
+            row[0]
+            for row in db.execute(
+                "SELECT name FROM sqlite_master WHERE type='table'"
+            )
+        }
+
+    @staticmethod
+    def _check_l2(db, checks, counts):
+        required_tables = {
+            'dim_players',
+            'fact_matches',
+            'fact_match_teams',
+            'fact_match_players',
+            'fact_rounds',
+            'fact_round_events',
+            'fact_round_player_economy',
+        }
+        missing = sorted(required_tables - IntegrityService._table_names(db))
+        IntegrityService._check(
+            checks,
+            'L2 required tables',
+            'fail' if missing else 'pass',
+            f"Missing: {', '.join(missing)}" if missing else 'All required tables exist',
+        )
+        if missing:
+            return
+
+        table_count_map = {
+            'matches': 'fact_matches',
+            'players': 'dim_players',
+            'player_match_rows': 'fact_match_players',
+            'rounds': 'fact_rounds',
+            'events': 'fact_round_events',
+            'economy_rows': 'fact_round_player_economy',
+        }
+        for key, table in table_count_map.items():
+            counts[key] = db.execute(f'SELECT COUNT(*) FROM {table}').fetchone()[0]
+
+        orphan_players = db.execute(
+            """
+            SELECT COUNT(*)
+            FROM fact_match_players mp
+            LEFT JOIN fact_matches m ON m.match_id = mp.match_id
+            WHERE m.match_id IS NULL
+            """
+        ).fetchone()[0]
+        IntegrityService._check(
+            checks,
+            'Player-match referential integrity',
+            'fail' if orphan_players else 'pass',
+            f'{orphan_players} player rows reference missing matches',
+            orphan_players,
+        )
+
+        orphan_events = db.execute(
+            """
+            SELECT COUNT(*)
+            FROM fact_round_events e
+            LEFT JOIN fact_rounds r
+              ON r.match_id = e.match_id AND r.round_num = e.round_num
+            WHERE r.match_id IS NULL
+            """
+        ).fetchone()[0]
+        IntegrityService._check(
+            checks,
+            'Round-event referential integrity',
+            'fail' if orphan_events else 'pass',
+            f'{orphan_events} events reference missing rounds',
+            orphan_events,
+        )
+
+        unusual_rosters = db.execute(
+            """
+            SELECT COUNT(*)
+            FROM (
+                SELECT match_id, COUNT(*) AS player_count
+                FROM fact_match_players
+                GROUP BY match_id
+                HAVING player_count != 10
+            )
+            """
+        ).fetchone()[0]
+        IntegrityService._check(
+            checks,
+            'Match player cardinality',
+            'warn' if unusual_rosters else 'pass',
+            f'{unusual_rosters} matches do not contain exactly 10 players',
+            unusual_rosters,
+        )
+
+        missing_names = db.execute(
+            "SELECT COUNT(*) FROM dim_players WHERE username IS NULL OR TRIM(username) = ''"
+        ).fetchone()[0]
+        IntegrityService._check(
+            checks,
+            'Player identity coverage',
+            'warn' if missing_names else 'pass',
+            f'{missing_names} players have no username',
+            missing_names,
+        )
+
+        required_indexes = {
+            'idx_match_players_player_match',
+            'idx_match_players_match_team',
+            'idx_match_players_party',
+            'idx_round_events_victim',
+            'idx_economy_player_match',
+            'idx_matches_map_time',
+        }
+        existing_indexes = {
+            row[0] for row in db.execute(
+                "SELECT name FROM sqlite_master WHERE type = 'index'"
+            )
+        }
+        missing_indexes = sorted(required_indexes - existing_indexes)
+        IntegrityService._check(
+            checks,
+            'L2 operational indexes',
+            'fail' if missing_indexes else 'pass',
+            (
+                f"Missing: {', '.join(missing_indexes)}"
+                if missing_indexes else
+                'All high-frequency query indexes exist'
+            ),
+            len(required_indexes) - len(missing_indexes),
+        )
+
+    @staticmethod
+    def _check_l3(db, roster_ids, checks, counts):
+        required_tables = {
+            'dm_player_features',
+            'dm_player_match_history',
+            'dm_player_map_stats',
+            'dm_player_period_stats',
+            'dm_player_records',
+            'dm_player_weapon_stats',
+        }
+        missing = sorted(required_tables - IntegrityService._table_names(db))
+        IntegrityService._check(
+            checks,
+            'L3 required tables',
+            'fail' if missing else 'pass',
+            f"Missing: {', '.join(missing)}" if missing else 'All required tables exist',
+        )
+        if missing:
+            return
+
+        counts['l3_features'] = db.execute(
+            'SELECT COUNT(*) FROM dm_player_features'
+        ).fetchone()[0]
+        counts['l3_history'] = db.execute(
+            'SELECT COUNT(*) FROM dm_player_match_history'
+        ).fetchone()[0]
+        counts['l3_maps'] = db.execute(
+            'SELECT COUNT(*) FROM dm_player_map_stats'
+        ).fetchone()[0]
+        counts['l3_weapons'] = db.execute(
+            'SELECT COUNT(*) FROM dm_player_weapon_stats'
+        ).fetchone()[0]
+        counts['l3_periods'] = db.execute(
+            'SELECT COUNT(*) FROM dm_player_period_stats'
+        ).fetchone()[0]
+        counts['l3_records'] = db.execute(
+            'SELECT COUNT(*) FROM dm_player_records'
+        ).fetchone()[0]
+
+        if roster_ids:
+            placeholders = ','.join('?' for _ in roster_ids)
+            covered = db.execute(
+                f"""
+                SELECT COUNT(DISTINCT steam_id_64)
+                FROM dm_player_features
+                WHERE steam_id_64 IN ({placeholders})
+                """,
+                roster_ids,
+            ).fetchone()[0]
+            IntegrityService._check(
+                checks,
+                'Active roster feature coverage',
+                'pass' if covered == len(roster_ids) else 'fail',
+                f'{covered}/{len(roster_ids)} roster players have L3 features',
+                covered,
+            )
+
+            expected_history = db.execute(
+                f"""
+                SELECT COALESCE(SUM(total_matches), 0)
+                FROM dm_player_features
+                WHERE steam_id_64 IN ({placeholders})
+                """,
+                roster_ids,
+            ).fetchone()[0]
+            actual_history = db.execute(
+                f"""
+                SELECT COUNT(*)
+                FROM dm_player_match_history
+                WHERE steam_id_64 IN ({placeholders})
+                """,
+                roster_ids,
+            ).fetchone()[0]
+            IntegrityService._check(
+                checks,
+                'Roster history completeness',
+                'pass' if actual_history == expected_history else 'fail',
+                f'{actual_history}/{expected_history} player-match rows materialized',
+                actual_history,
+            )
+
+            score_rows = db.execute(
+                f"""
+                SELECT steam_id_64, score_overall, tier_percentile
+                FROM dm_player_features
+                WHERE steam_id_64 IN ({placeholders})
+                  AND score_overall > 0
+                """,
+                roster_ids,
+            ).fetchall()
+            scores = [float(row['score_overall']) for row in score_rows]
+            invalid_percentiles = 0
+            for row in score_rows:
+                expected = (
+                    sum(value <= float(row['score_overall']) for value in scores)
+                    / len(scores)
+                    * 100
+                )
+                actual = row['tier_percentile']
+                if actual is None or abs(float(actual) - expected) > 0.011:
+                    invalid_percentiles += 1
+            IntegrityService._check(
+                checks,
+                'Roster percentile correctness',
+                'pass' if invalid_percentiles == 0 else 'warn',
+                f'{invalid_percentiles} eligible players have stale percentiles',
+                invalid_percentiles,
+            )
+
+            for table, label in (
+                ('dm_player_period_stats', 'Roster period-stat coverage'),
+                ('dm_player_records', 'Roster record coverage'),
+            ):
+                covered = db.execute(
+                    f"""
+                    SELECT COUNT(DISTINCT steam_id_64)
+                    FROM {table}
+                    WHERE steam_id_64 IN ({placeholders})
+                    """,
+                    roster_ids,
+                ).fetchone()[0]
+                IntegrityService._check(
+                    checks,
+                    label,
+                    'pass' if covered == len(roster_ids) else 'fail',
+                    f'{covered}/{len(roster_ids)} roster players covered',
+                    covered,
+                )
+
+        for key, label in (
+            ('l3_history', 'Player match history mart'),
+            ('l3_maps', 'Player map stats mart'),
+            ('l3_weapons', 'Player weapon stats mart'),
+            ('l3_periods', 'Player period stats mart'),
+            ('l3_records', 'Player records mart'),
+        ):
+            value = counts[key]
+            IntegrityService._check(
+                checks,
+                label,
+                'pass' if value else 'warn',
+                f'{value} rows',
+                value,
+            )
+
+        if roster_ids:
+            placeholders = ','.join('?' for _ in roster_ids)
+            scope_sql = f"AND steam_id_64 IN ({placeholders})"
+            scope_args = roster_ids
+        else:
+            scope_sql = ''
+            scope_args = []
+        placeholder_rows = db.execute(
+            f"""
+            SELECT COUNT(*)
+            FROM dm_player_features
+            WHERE int_pos_site_a_control_rate = 0.33
+              AND int_pos_site_b_control_rate = 0.33
+              AND int_pos_mid_control_rate = 0.34
+              {scope_sql}
+            """,
+            scope_args,
+        ).fetchone()[0]
+        IntegrityService._check(
+            checks,
+            'Experimental spatial metrics',
+            'warn' if placeholder_rows else 'pass',
+            f'{placeholder_rows} active-roster rows contain placeholder site-control values',
+            placeholder_rows,
+        )
+
+    @staticmethod
+    def _check_web(db, checks, counts):
+        required_tables = {
+            'comments',
+            'etl_jobs',
+            'match_imports',
+            'player_metadata',
+            'schema_migrations',
+            'strategy_boards',
+            'team_lineups',
+            'wiki_pages',
+        }
+        missing = sorted(required_tables - IntegrityService._table_names(db))
+        IntegrityService._check(
+            checks,
+            'Web required tables',
+            'fail' if missing else 'pass',
+            f"Missing: {', '.join(missing)}" if missing else 'All required tables exist',
+        )
+        if missing:
+            return
+
+        counts['etl_jobs'] = db.execute(
+            'SELECT COUNT(*) FROM etl_jobs'
+        ).fetchone()[0]
+        counts['match_imports'] = db.execute(
+            'SELECT COUNT(*) FROM match_imports'
+        ).fetchone()[0]
+
+        schema_version = db.execute(
+            'SELECT COALESCE(MAX(version), 0) FROM schema_migrations'
+        ).fetchone()[0]
+        counts['web_schema_version'] = schema_version
+        IntegrityService._check(
+            checks,
+            'Web schema version',
+            'pass' if schema_version == Config.WEB_SCHEMA_VERSION else 'fail',
+            f'{schema_version}/{Config.WEB_SCHEMA_VERSION}',
+            schema_version,
+        )
+
+        foreign_key_errors = db.execute(
+            'PRAGMA foreign_key_check'
+        ).fetchall()
+        IntegrityService._check(
+            checks,
+            'Web foreign key integrity',
+            'fail' if foreign_key_errors else 'pass',
+            f'{len(foreign_key_errors)} foreign key violations',
+            len(foreign_key_errors),
+        )
+
+        running_jobs = db.execute(
+            """
+            SELECT COUNT(*)
+            FROM etl_jobs
+            WHERE status = 'running'
+            """
+        ).fetchone()[0]
+        IntegrityService._check(
+            checks,
+            'Pipeline concurrency',
+            'warn' if running_jobs > 1 else 'pass',
+            f'{running_jobs} running pipeline jobs',
+            running_jobs,
+        )
+
+        lineups = db.execute(
+            'SELECT id, player_ids_json, is_active FROM team_lineups'
+        ).fetchall()
+        counts['lineups'] = len(lineups)
+        invalid_lineups = 0
+        for lineup in lineups:
+            try:
+                player_ids = json.loads(lineup['player_ids_json'] or '[]')
+                if not isinstance(player_ids, list):
+                    invalid_lineups += 1
+            except (TypeError, json.JSONDecodeError):
+                invalid_lineups += 1
+        IntegrityService._check(
+            checks,
+            'Lineup JSON validity',
+            'fail' if invalid_lineups else 'pass',
+            f'{invalid_lineups} lineups contain invalid player ID JSON',
+            invalid_lineups,
+        )
+
+        active_count = sum(1 for lineup in lineups if lineup['is_active'] == 1)
+        IntegrityService._check(
+            checks,
+            'Active lineup',
+            'pass' if active_count == 1 else 'warn',
+            f'{active_count} active lineups configured',
+            active_count,
+        )
@@ -1,19 +1,10 @@
 from web.database import query_db
-from web.services.web_service import WebService
-import json
+from web.services.team_context_service import TeamContextService

 class OpponentService:
    @staticmethod
    def _get_active_roster_ids():
-        lineups = WebService.get_lineups()
-        active_roster_ids = []
-        if lineups:
-            try:
-                raw_ids = json.loads(lineups[0]['player_ids_json'])
-                active_roster_ids = [str(uid) for uid in raw_ids]
-            except:
-                pass
-        return active_roster_ids
+        return TeamContextService.get_active_roster_ids()

    @staticmethod
    def get_opponent_list(page=1, per_page=20, sort_by='matches', search=None):
@@ -21,30 +12,21 @@ class OpponentService:
        if not roster_ids:
            return [], 0

-        # Placeholders
        roster_ph = ','.join('?' for _ in roster_ids)
-        
-        # 1. Identify Matches involving our roster (at least 1 member? usually 2 for 'team' match)
-        # Let's say at least 1 for broader coverage as requested ("1 match sample")
-        # But "Our Team" usually implies the entity. Let's stick to matches where we can identify "Us".
-        # If we use >=1, we catch solo Q matches of roster members. The user said "Non-team members or 1 match sample", 
-        # but implied "facing different our team lineups".
-        # Let's use the standard "candidate matches" logic (>=2 roster members) to represent "The Team".
-        # OR, if user wants "Opponent Analysis" for even 1 match, maybe they mean ANY match in DB?
-        # "Left Top add Opponent Analysis... (non-team member or 1 sample)"
-        # This implies we analyze PLAYERS who are NOT us.
-        # Let's stick to matches where >= 1 roster member played, to define "Us" vs "Them".
-        
-        # Actually, let's look at ALL matches in DB, and any player NOT in active roster is an "Opponent".
-        # This covers "1 sample".
-        
-        # Query:
-        # Select all players who are NOT in active roster.
-        # Group by steam_id.
-        # Aggregate stats.
-        
-        where_clauses = [f"CAST(mp.steam_id_64 AS TEXT) NOT IN ({roster_ph})"]
-        args = list(roster_ids)
+
+        where_clauses = [
+            f"CAST(mp.steam_id_64 AS TEXT) NOT IN ({roster_ph})",
+            f"""
+            EXISTS (
+                SELECT 1
+                FROM fact_match_players roster_mp
+                WHERE roster_mp.match_id = mp.match_id
+                  AND CAST(roster_mp.steam_id_64 AS TEXT) IN ({roster_ph})
+                  AND roster_mp.team_id != mp.team_id
+            )
+            """,
+        ]
+        args = list(roster_ids) + list(roster_ids)
        
        if search:
            where_clauses.append("(LOWER(p.username) LIKE LOWER(?) OR mp.steam_id_64 LIKE ?)")
@@ -61,16 +43,6 @@ class OpponentService:
        elif sort_by == 'win_rate':
            sort_sql = "win_rate DESC"
        
-        # Main Aggregation Query
-        # We need to join fact_matches to get match info (win/loss, elo) if needed, 
-        # but fact_match_players has is_win (boolean) usually? No, it has team_id.
-        # We need to determine if THEY won.
-        # fact_match_players doesn't store is_win directly in schema (I should check schema, but stats_service calculates it).
-        # Wait, stats_service.get_player_trend uses `mp.is_win`? 
-        # Let's check schema. `fact_match_players` usually has `match_id`, `team_id`.
-        # `fact_matches` has `winner_team`.
-        # So we join.
-        
        offset = (page - 1) * per_page
        
        sql = f"""
@@ -151,10 +123,17 @@ class OpponentService:
            LEFT JOIN fact_match_teams fmt_gid ON mp.match_id = fmt_gid.match_id AND fmt_gid.group_id = mp.team_id
            LEFT JOIN fact_match_teams fmt_tid ON mp.match_id = fmt_tid.match_id AND fmt_tid.group_tid = mp.match_team_id
            WHERE CAST(mp.steam_id_64 AS TEXT) NOT IN ({roster_ph})
+              AND EXISTS (
+                  SELECT 1
+                  FROM fact_match_players roster_mp
+                  WHERE roster_mp.match_id = mp.match_id
+                    AND CAST(roster_mp.steam_id_64 AS TEXT) IN ({roster_ph})
+                    AND roster_mp.team_id != mp.team_id
+              )
            GROUP BY mp.steam_id_64
        """
        
-        rows = query_db('l2', sql, roster_ids)
+        rows = query_db('l2', sql, roster_ids + roster_ids)
        
        # Initialize Buckets
        elo_buckets = {'<1000': 0, '1000-1200': 0, '1200-1400': 0, '1400-1600': 0, '1600-1800': 0, '1800-2000': 0, '>2000': 0}
@@ -216,13 +195,12 @@ class OpponentService:
        player = dict(info)
        player['avatar_url'] = StatsService.resolve_avatar_url(steam_id, player.get('avatar_url'))
            
-        # 2. Match History vs Us (All matches this player played)
-        # We define "Us" as matches where this player is an opponent.
-        # But actually, we just show ALL their matches in our DB, assuming our DB only contains matches relevant to us?
-        # Usually yes, but if we have a huge DB, we might want to filter by "Contains Roster Member".
-        # For now, show all matches in DB for this player.
-        
-        sql_history = """
+        roster_ids = OpponentService._get_active_roster_ids()
+        if not roster_ids:
+            return None
+        roster_ph = ','.join('?' for _ in roster_ids)
+
+        sql_history = f"""
            SELECT 
                m.match_id, m.start_time, m.map_name, m.score_team1, m.score_team2, m.winner_team,
                mp.team_id, mp.match_team_id, mp.rating, mp.kd_ratio, mp.adr, mp.kills, mp.deaths,
@@ -236,9 +214,16 @@ class OpponentService:
            LEFT JOIN fact_match_teams fmt_gid ON mp.match_id = fmt_gid.match_id AND fmt_gid.group_id = mp.team_id
            LEFT JOIN fact_match_teams fmt_tid ON mp.match_id = fmt_tid.match_id AND fmt_tid.group_tid = mp.match_team_id
            WHERE mp.steam_id_64 = ?
+              AND EXISTS (
+                  SELECT 1
+                  FROM fact_match_players roster_mp
+                  WHERE roster_mp.match_id = mp.match_id
+                    AND CAST(roster_mp.steam_id_64 AS TEXT) IN ({roster_ph})
+                    AND roster_mp.team_id != mp.team_id
+              )
            ORDER BY m.start_time DESC
        """
-        history = query_db('l2', sql_history, [steam_id])
+        history = query_db('l2', sql_history, [steam_id] + roster_ids)
        
        # 3. Aggregation by ELO
        elo_buckets = {
@@ -389,11 +374,18 @@ class OpponentService:
            LEFT JOIN fact_match_teams fmt_gid ON mp.match_id = fmt_gid.match_id AND fmt_gid.group_id = mp.team_id
            LEFT JOIN fact_match_teams fmt_tid ON mp.match_id = fmt_tid.match_id AND fmt_tid.group_tid = mp.match_team_id
            WHERE CAST(mp.steam_id_64 AS TEXT) NOT IN ({roster_ph})
+              AND EXISTS (
+                  SELECT 1
+                  FROM fact_match_players roster_mp
+                  WHERE roster_mp.match_id = mp.match_id
+                    AND CAST(roster_mp.steam_id_64 AS TEXT) IN ({roster_ph})
+                    AND roster_mp.team_id != mp.team_id
+              )
              AND m.map_name IS NOT NULL AND m.map_name <> ''
            GROUP BY m.map_name
            ORDER BY matches DESC
        """
-        rows = query_db('l2', sql, roster_ids)
+        rows = query_db('l2', sql, roster_ids + roster_ids)
        results = []
        for r in rows:
            d = dict(r)
@@ -0,0 +1,123 @@
+from web.database import query_db
+
+
+class PlayerProfileService:
+    PERIOD_KEYS = (
+        'career',
+        'last_10',
+        'last_20',
+        'last_30',
+        'days_30',
+        'days_90',
+    )
+
+    @staticmethod
+    def get_period_stats(steam_id):
+        rows = query_db(
+            'l3',
+            """
+            SELECT *
+            FROM dm_player_period_stats
+            WHERE steam_id_64 = ?
+            ORDER BY CASE period_key
+                WHEN 'career' THEN 1
+                WHEN 'last_10' THEN 2
+                WHEN 'last_20' THEN 3
+                WHEN 'last_30' THEN 4
+                WHEN 'days_30' THEN 5
+                WHEN 'days_90' THEN 6
+                ELSE 99
+            END
+            """,
+            [steam_id],
+        )
+        return [dict(row) for row in rows]
+
+    @staticmethod
+    def get_period(steam_id, period_key):
+        if period_key not in PlayerProfileService.PERIOD_KEYS:
+            return None
+        row = query_db(
+            'l3',
+            """
+            SELECT *
+            FROM dm_player_period_stats
+            WHERE steam_id_64 = ? AND period_key = ?
+            """,
+            [steam_id, period_key],
+            one=True,
+        )
+        return dict(row) if row else None
+
+    @staticmethod
+    def get_records(steam_id):
+        rows = query_db(
+            'l3',
+            """
+            SELECT *
+            FROM dm_player_records
+            WHERE steam_id_64 = ?
+            ORDER BY CASE record_key
+                WHEN 'highest_rating' THEN 1
+                WHEN 'most_kills' THEN 2
+                WHEN 'highest_adr' THEN 3
+                WHEN 'highest_kd' THEN 4
+                WHEN 'most_headshots' THEN 5
+                WHEN 'longest_win_streak' THEN 6
+                ELSE 99
+            END
+            """,
+            [steam_id],
+        )
+        return [dict(row) for row in rows]
+
+    @staticmethod
+    def get_period_history(steam_id, period_key):
+        period = PlayerProfileService.get_period(steam_id, period_key)
+        if not period:
+            return []
+        rows = query_db(
+            'l3',
+            """
+            SELECT
+                match_date AS start_time,
+                rating,
+                kd_ratio,
+                adr,
+                kast,
+                match_id,
+                map_name,
+                is_win,
+                match_sequence AS match_index
+            FROM dm_player_match_history
+            WHERE steam_id_64 = ?
+              AND match_date BETWEEN ? AND ?
+            ORDER BY match_date, match_id
+            """,
+            [steam_id, period['period_start'], period['period_end']],
+        )
+        return [dict(row) for row in rows]
+
+    @staticmethod
+    def get_map_stats(steam_id):
+        rows = query_db(
+            'l3',
+            """
+            SELECT
+                map_name,
+                matches,
+                wins,
+                win_rate,
+                avg_rating AS rating,
+                avg_kd AS kd,
+                avg_adr AS adr,
+                avg_kast AS kast,
+                best_rating,
+                worst_rating
+            FROM dm_player_map_stats
+            WHERE steam_id_64 = ?
+            ORDER BY matches DESC, map_name
+            """,
+            [steam_id],
+        )
+        return [dict(row) for row in rows]
@@ -1,4 +1,4 @@
-from web.database import query_db, execute_db
+from web.database import query_db
 from flask import current_app, url_for
 import os

@@ -13,7 +13,7 @@ class StatsService:
        try:
            # Check local file first (User Request: "directly associate if exists")
            base = os.path.join(current_app.root_path, 'static', 'avatars')
-            for ext in ('.jpg', '.png', '.jpeg'):
+            for ext in ('.jpg', '.png', '.jpeg', '.webp'):
                fname = f"{steam_id}{ext}"
                fpath = os.path.join(base, fname)
                if os.path.exists(fpath):
@@ -38,18 +38,9 @@ class StatsService:
                'round_stats': [{'type', 'count', 'wins', 'win_rate'}]
            }
        """
-        # 1. Get Active Roster
-        from web.services.web_service import WebService
-        import json
-        
-        lineups = WebService.get_lineups()
-        active_roster_ids = []
-        if lineups:
-            try:
-                raw_ids = json.loads(lineups[0]['player_ids_json'])
-                active_roster_ids = [str(uid) for uid in raw_ids]
-            except:
-                pass
+        from web.services.team_context_service import TeamContextService
+
+        active_roster_ids = TeamContextService.get_active_roster_ids()
        
        if not active_roster_ids:
            return {}
@@ -60,21 +51,23 @@ class StatsService:
        
        placeholders = ','.join('?' for _ in active_roster_ids)
        
-        # Step A: Get Candidate Match IDs (matches with >= 2 roster players)
-        # Also get the team_id of our players in that match to determine win
        candidate_sql = f"""
-            SELECT mp.match_id, MAX(mp.team_id) as our_team_id
+            SELECT mp.match_id, mp.team_id as our_team_id,
+                   COUNT(DISTINCT mp.steam_id_64) as roster_count
            FROM fact_match_players mp
            WHERE CAST(mp.steam_id_64 AS TEXT) IN ({placeholders})
-            GROUP BY mp.match_id
+            GROUP BY mp.match_id, mp.team_id
            HAVING COUNT(DISTINCT mp.steam_id_64) >= 2
+            ORDER BY mp.match_id, roster_count DESC, mp.team_id
        """
        candidate_rows = query_db('l2', candidate_sql, active_roster_ids)
        
        if not candidate_rows:
            return {}
            
-        candidate_map = {row['match_id']: row['our_team_id'] for row in candidate_rows}
+        candidate_map = {}
+        for row in candidate_rows:
+            candidate_map.setdefault(row['match_id'], row['our_team_id'])
        match_ids = list(candidate_map.keys())
        match_placeholders = ','.join('?' for _ in match_ids)
        
@@ -221,11 +214,15 @@ class StatsService:
            args.append(map_name)
        
        if date_from:
-            where_clauses.append("start_time >= ?")
+            where_clauses.append(
+                "start_time >= CAST(strftime('%s', ?) AS INTEGER)"
+            )
            args.append(date_from)
            
        if date_to:
-            where_clauses.append("start_time <= ?")
+            where_clauses.append(
+                "start_time < CAST(strftime('%s', date(?, '+1 day')) AS INTEGER)"
+            )
            args.append(date_to)
            
        where_str = " AND ".join(where_clauses)
@@ -270,109 +267,51 @@ class StatsService:
            party_rows = query_db('l2', party_sql, match_ids)
            party_map = {row['match_id']: row['max_party'] for row in party_rows}
            
-            # --- New: Determine "Our Team" Result ---
-            # Logic: Check if any player from `active_roster` played in these matches.
-            # Use WebService to get the active roster
-            from web.services.web_service import WebService
-            import json
-            
-            lineups = WebService.get_lineups()
-            active_roster_ids = []
-            if lineups:
-                try:
-                    # Load IDs and ensure they are all strings for DB comparison consistency
-                    raw_ids = json.loads(lineups[0]['player_ids_json'])
-                    active_roster_ids = [str(uid) for uid in raw_ids]
-                except:
-                    pass
+            from web.services.team_context_service import TeamContextService
+
+            active_roster_ids = TeamContextService.get_active_roster_ids()
            
            # If no roster, we can't determine "Our Result"
            if not active_roster_ids:
                result_map = {}
            else:
-                # 1. Get UIDs for Roster Members involved in these matches
-                # We query fact_match_players to ensure we get the UIDs actually used in these matches
                roster_placeholders = ','.join('?' for _ in active_roster_ids)
-                uid_sql = f"""
-                    SELECT DISTINCT steam_id_64, uid 
+                roster_team_sql = f"""
+                    SELECT match_id, team_id,
+                           COUNT(DISTINCT steam_id_64) as roster_count
                    FROM fact_match_players 
                    WHERE match_id IN ({placeholders})
                    AND CAST(steam_id_64 AS TEXT) IN ({roster_placeholders})
+                    GROUP BY match_id, team_id
                """
-                combined_args_uid = match_ids + active_roster_ids
-                uid_rows = query_db('l2', uid_sql, combined_args_uid)
-                
-                # Set of "Our UIDs" (as strings)
-                our_uids = set()
-                for r in uid_rows:
-                    if r['uid']:
-                        our_uids.add(str(r['uid']))
-                
-                # 2. Get Group UIDs and Winner info from fact_match_teams
-                # We need to know which group contains our UIDs
-                teams_sql = f"""
-                    SELECT fmt.match_id, fmt.group_id, fmt.group_uids, m.winner_team
-                    FROM fact_match_teams fmt
-                    JOIN fact_matches m ON fmt.match_id = m.match_id
-                    WHERE fmt.match_id IN ({placeholders})
-                """
-                teams_rows = query_db('l2', teams_sql, match_ids)
-                
-                # 3. Determine Result per Match
+                roster_team_rows = query_db(
+                    'l2',
+                    roster_team_sql,
+                    match_ids + active_roster_ids,
+                )
+                winner_by_match = {
+                    str(match['match_id']): match['winner_team']
+                    for match in matches
+                }
+                teams_by_match = {}
+                for row in roster_team_rows:
+                    teams_by_match.setdefault(str(row['match_id']), []).append(
+                        row['team_id']
+                    )
+
                result_map = {}
-                
-                # Group data by match
-                match_groups = {} # match_id -> {group_id: [uids...], winner: int}
-                
-                for r in teams_rows:
-                    mid = r['match_id']
-                    gid = r['group_id']
-                    uids_str = r['group_uids'] or ""
-                    # Split and clean UIDs
-                    uids = set(str(u).strip() for u in uids_str.split(',') if u.strip())
-                    
-                    if mid not in match_groups:
-                        match_groups[mid] = {'groups': {}, 'winner': r['winner_team']}
-                    
-                    match_groups[mid]['groups'][gid] = uids
-                
-                # Analyze
-                for mid, data in match_groups.items():
-                    winner_gid = data['winner']
-                    groups = data['groups']
-                    
-                    our_in_winner = False
-                    our_in_loser = False
-                    
-                    # Check each group
-                    for gid, uids in groups.items():
-                        # Intersection of Our UIDs and Group UIDs
-                        common = our_uids.intersection(uids)
-                        if common:
-                            if gid == winner_gid:
-                                our_in_winner = True
-                            else:
-                                our_in_loser = True
-                    
-                    if our_in_winner and not our_in_loser:
-                        result_map[mid] = 'win'
-                    elif our_in_loser and not our_in_winner:
-                        result_map[mid] = 'loss'
-                    elif our_in_winner and our_in_loser:
-                        result_map[mid] = 'mixed'
-                    else:
-                        # Fallback: If UID matching failed (maybe missing UIDs), try old team_id method?
-                        # Or just leave it as None (safe)
-                        pass
+                for match_id, team_ids in teams_by_match.items():
+                    unique_team_ids = set(team_ids)
+                    if len(unique_team_ids) > 1:
+                        result_map[match_id] = 'mixed'
+                        continue
+                    our_team_id = next(iter(unique_team_ids))
+                    result_map[match_id] = (
+                        'win'
+                        if str(our_team_id) == str(winner_by_match.get(match_id))
+                        else 'loss'
+                    )

-            # Convert to dict to modify
-            matches = [dict(m) for m in matches]
-            for m in matches:
-                m['avg_elo'] = elo_map.get(m['match_id'], 0)
-                m['max_party'] = party_map.get(m['match_id'], 1)
-                m['our_result'] = result_map.get(m['match_id'])
-            
-            # Convert to dict to modify
            matches = [dict(m) for m in matches]
            for m in matches:
                m['avg_elo'] = elo_map.get(m['match_id'], 0)
@@ -542,33 +481,20 @@ class StatsService:

    @staticmethod
    def get_shared_matches(steam_ids):
-        # Find matches where ALL steam_ids were present
        if not steam_ids or len(steam_ids) < 1:
            return []
-            
+
+        steam_ids = list(dict.fromkeys(str(steam_id) for steam_id in steam_ids))
        placeholders = ','.join('?' for _ in steam_ids)
        count = len(steam_ids)
-        
-        # We need to know which team the players were on to determine win/loss
-        # Assuming they were on the SAME team for "shared experience"
-        # If count=1, it's just match history
-        
-        # Query: Get matches where all steam_ids are present
-        # Also join to get team_id to check if they were on the same team (optional but better)
-        # For simplicity in v1: Just check presence in the match.
-        # AND check if the player won.
-        
-        # We need to return: match_id, map_name, score, result (Win/Loss)
-        # "Result" is relative to the lineup.
-        # If they were on the winning team, it's a Win.
-        
+
        sql = f"""
            SELECT m.match_id, m.start_time, m.map_name, m.score_team1, m.score_team2, m.winner_team,
-                   MAX(mp.team_id) as player_team_id -- Just take one team_id (assuming same)
+                   mp.team_id as player_team_id
            FROM fact_matches m
            JOIN fact_match_players mp ON m.match_id = mp.match_id
            WHERE mp.steam_id_64 IN ({placeholders})
-            GROUP BY m.match_id
+            GROUP BY m.match_id, mp.team_id
            HAVING COUNT(DISTINCT mp.steam_id_64) = ?
            ORDER BY m.start_time DESC
        """
@@ -580,14 +506,7 @@ class StatsService:
        
        results = []
        for r in rows:
-            # Determine if Win
-            # winner_team in DB is 'Team 1' or 'Team 2' usually, or the team name.
-            # fact_matches.winner_team stores the NAME of the winner? Or 'team1'/'team2'?
-            # Let's check how L2_Builder stores it. Usually it stores the name.
-            # But fact_match_players.team_id stores the name too.
-            
-            # Logic: If m.winner_team == mp.team_id, then Win.
-            is_win = (r['winner_team'] == r['player_team_id'])
+            is_win = str(r['winner_team']) == str(r['player_team_id'])
            
            # If winner_team is NULL or empty, it's a draw?
            if not r['winner_team']:
@@ -628,7 +547,31 @@ class StatsService:
        """
        l3_rows = query_db("l3", l3_sql, [steam_id, limit])
        if l3_rows:
-            return l3_rows
+            history = [dict(row) for row in l3_rows]
+            match_ids = [row['match_id'] for row in history]
+            placeholders = ','.join('?' for _ in match_ids)
+            party_rows = query_db(
+                "l2",
+                f"""
+                SELECT me.match_id, COUNT(p.steam_id_64) AS party_size
+                FROM fact_match_players me
+                LEFT JOIN fact_match_players p
+                  ON p.match_id = me.match_id
+                 AND p.match_team_id = me.match_team_id
+                 AND me.match_team_id > 0
+                WHERE me.steam_id_64 = ?
+                  AND me.match_id IN ({placeholders})
+                GROUP BY me.match_id
+                """,
+                [steam_id] + match_ids,
+            )
+            party_map = {
+                row['match_id']: max(int(row['party_size'] or 0), 1)
+                for row in party_rows
+            }
+            for row in history:
+                row['party_size'] = party_map.get(row['match_id'], 1)
+            return history

        sql = """
            SELECT * FROM (
@@ -729,19 +672,10 @@ class StatsService:
        Calculates rank and distribution of the target player within the active roster.
        Now covers all L3 Basic Features for Detailed Panel.
        """
-        from web.services.web_service import WebService
        from web.services.feature_service import FeatureService
-        import json
-        
-        # 1. Get Active Roster IDs
-        lineups = WebService.get_lineups()
-        active_roster_ids = []
-        if lineups:
-            try:
-                raw_ids = json.loads(lineups[0]['player_ids_json'])
-                active_roster_ids = [str(uid) for uid in raw_ids]
-            except:
-                pass
+        from web.services.team_context_service import TeamContextService
+
+        active_roster_ids = TeamContextService.get_active_roster_ids()
        
        if not active_roster_ids:
            return None
@@ -851,33 +785,38 @@ class StatsService:
            "basic_avg_rating", "basic_avg_kd", "basic_avg_adr", "basic_avg_kast", "basic_avg_rws",
        ]
        
-        lower_is_better = []
+        lower_is_better = {
+            "int_timing_first_contact_time",
+            "int_trade_response_time",
+            "tac_avg_fd",
+            "tac_fd_rate",
+            "core_avg_match_duration",
+            "core_dpr",
+            "meta_rating_volatility",
+            "meta_map_stability",
+            "meta_elo_tier_stability",
+        }

        result = {}
        
        for m in metrics:
            values = []
-            non_numeric = False
            for p in stats_map.values():
                raw = (p or {}).get(m)
                if raw is None:
-                    raw = 0
+                    continue
                try:
                    values.append(float(raw))
-                except Exception:
-                    non_numeric = True
-                    break
+                except (TypeError, ValueError):
+                    continue

            raw_target = (stats_map.get(target_steam_id) or {}).get(m)
            if raw_target is None:
-                raw_target = 0
+                result[m] = None
+                continue
            try:
                target_val = float(raw_target)
-            except Exception:
-                non_numeric = True
-                target_val = 0
-            
-            if non_numeric:
+            except (TypeError, ValueError):
                result[m] = None
                continue

@@ -0,0 +1,36 @@
+import json
+
+from web.services.web_service import WebService
+
+
+class TeamContextService:
+    """Single source of truth for the private team's active roster."""
+
+    @staticmethod
+    def get_active_lineup():
+        lineup = WebService.get_active_lineup()
+        return dict(lineup) if lineup else None
+
+    @staticmethod
+    def get_active_roster_ids():
+        lineup = TeamContextService.get_active_lineup()
+        if not lineup:
+            return []
+
+        try:
+            raw_ids = json.loads(lineup.get('player_ids_json') or '[]')
+        except (TypeError, json.JSONDecodeError):
+            return []
+
+        if not isinstance(raw_ids, list):
+            return []
+
+        seen = set()
+        roster_ids = []
+        for raw_id in raw_ids:
+            steam_id = str(raw_id).strip()
+            if steam_id and steam_id not in seen:
+                seen.add(steam_id)
+                roster_ids.append(steam_id)
+        return roster_ids
+
@@ -53,17 +53,39 @@ class WebService:
            sql = "UPDATE team_lineups SET name=?, description=?, player_ids_json=? WHERE id=?"
            return execute_db('web', sql, [name, description, ids_json, lineup_id])
        else:
-            sql = "INSERT INTO team_lineups (name, description, player_ids_json) VALUES (?, ?, ?)"
-            return execute_db('web', sql, [name, description, ids_json])
+            active = 0 if WebService.get_active_lineup() else 1
+            sql = """
+                INSERT INTO team_lineups
+                    (name, description, player_ids_json, is_active)
+                VALUES (?, ?, ?, ?)
+            """
+            return execute_db('web', sql, [name, description, ids_json, active])

    @staticmethod
    def get_lineups():
-        return query_db('web', "SELECT * FROM team_lineups ORDER BY created_at DESC")
+        return query_db(
+            'web',
+            "SELECT * FROM team_lineups ORDER BY is_active DESC, created_at DESC, id DESC",
+        )

    @staticmethod
    def get_lineup(lineup_id):
        return query_db('web', "SELECT * FROM team_lineups WHERE id = ?", [lineup_id], one=True)

+    @staticmethod
+    def get_active_lineup():
+        lineup = query_db(
+            'web',
+            "SELECT * FROM team_lineups WHERE is_active = 1 ORDER BY id LIMIT 1",
+            one=True,
+        )
+        if lineup:
+            return lineup
+        return query_db(
+            'web',
+            "SELECT * FROM team_lineups ORDER BY created_at DESC, id DESC LIMIT 1",
+            one=True,
+        )

    # --- Users / Auth ---
    @staticmethod
@@ -12,9 +12,8 @@
        <div class="border border-gray-200 dark:border-gray-700 rounded-lg p-4">
            <h3 class="text-lg font-bold text-gray-900 dark:text-white mb-4">数据管线 (ETL)</h3>
            <div class="space-y-2">
-                <button onclick="triggerEtl('L1A.py')" class="w-full bg-blue-600 text-white py-2 px-4 rounded hover:bg-blue-700">Trigger L1A (Ingest)</button>
-                <button onclick="triggerEtl('L2_Builder.py')" class="w-full bg-blue-600 text-white py-2 px-4 rounded hover:bg-blue-700">Trigger L2 Builder</button>
-                <button onclick="triggerEtl('L3_Builder.py')" class="w-full bg-blue-600 text-white py-2 px-4 rounded hover:bg-blue-700">Trigger L3 Builder</button>
+                <a href="{{ url_for('admin.import_match') }}" class="block w-full text-center bg-yrtv-600 text-white py-2 px-4 rounded hover:bg-yrtv-500">上传并导入比赛</a>
+                <button onclick="triggerEtl()" class="w-full bg-blue-600 text-white py-2 px-4 rounded hover:bg-blue-700">运行完整 L1 → L2 → L3</button>
            </div>
            <div id="etlResult" class="mt-4 text-sm text-gray-600 dark:text-gray-400"></div>
        </div>
@@ -23,6 +22,7 @@
        <div class="border border-gray-200 dark:border-gray-700 rounded-lg p-4">
            <h3 class="text-lg font-bold text-gray-900 dark:text-white mb-4">工具箱</h3>
            <div class="space-y-2">
+                <a href="{{ url_for('admin.data_integrity') }}" class="block w-full text-center bg-emerald-600 text-white py-2 px-4 rounded hover:bg-emerald-700">数据完整性中心</a>
                <a href="{{ url_for('admin.sql_runner') }}" class="block w-full text-center bg-gray-600 text-white py-2 px-4 rounded hover:bg-gray-700">SQL Runner</a>
                <a href="{{ url_for('wiki.index') }}" class="block w-full text-center bg-gray-600 text-white py-2 px-4 rounded hover:bg-gray-700">Manage Wiki</a>
            </div>
@@ -31,20 +31,23 @@
 </div>

 <script>
-function triggerEtl(scriptName) {
+function triggerEtl() {
    const resultDiv = document.getElementById('etlResult');
-    resultDiv.innerText = "Triggering " + scriptName + "...";
+    resultDiv.innerText = "正在创建后台流水线...";
    
    fetch("{{ url_for('admin.trigger_etl') }}", {
        method: 'POST',
        headers: {
            'Content-Type': 'application/x-www-form-urlencoded',
        },
-        body: 'script=' + scriptName
    })
-    .then(response => response.text())
-    .then(text => {
-        resultDiv.innerText = text;
+    .then(response => response.json())
+    .then(data => {
+        if (data.success) {
+            window.location.href = "{{ url_for('admin.import_match') }}?job_id=" + data.job_id;
+        } else {
+            resultDiv.innerText = data.error || "启动失败";
+        }
    })
    .catch(err => {
        resultDiv.innerText = "Error: " + err;
@@ -0,0 +1,84 @@
+{% extends "base.html" %}
+
+{% block title %}数据完整性 - YRTV{% endblock %}
+
+{% block content %}
+{% set status_styles = {
+    'pass': 'bg-emerald-100 text-emerald-800 dark:bg-emerald-900/40 dark:text-emerald-300',
+    'warn': 'bg-amber-100 text-amber-800 dark:bg-amber-900/40 dark:text-amber-300',
+    'fail': 'bg-red-100 text-red-800 dark:bg-red-900/40 dark:text-red-300'
+} %}
+<div class="space-y-6 px-4 sm:px-0">
+    <div class="flex flex-col gap-4 sm:flex-row sm:items-center sm:justify-between">
+        <div>
+            <div class="flex items-center gap-3">
+                <h1 class="text-2xl font-bold text-slate-900 dark:text-white">数据完整性中心</h1>
+                <span class="rounded-full px-3 py-1 text-xs font-semibold uppercase {{ status_styles[report.overall_status] }}">
+                    {{ report.overall_status }}
+                </span>
+            </div>
+            <p class="mt-1 text-sm text-slate-500 dark:text-slate-400">
+                校验时间：{{ report.generated_at }}
+            </p>
+        </div>
+        <div class="flex gap-2">
+            <a href="{{ url_for('admin.data_integrity', format='json') }}"
+               class="rounded-lg border border-slate-300 px-4 py-2 text-sm font-medium text-slate-700 hover:bg-slate-50 dark:border-slate-600 dark:text-slate-200 dark:hover:bg-slate-800">
+                JSON
+            </a>
+            <a href="{{ url_for('admin.data_integrity') }}"
+               class="rounded-lg bg-yrtv-600 px-4 py-2 text-sm font-medium text-white hover:bg-yrtv-500">
+                重新校验
+            </a>
+        </div>
+    </div>
+
+    <div class="grid grid-cols-3 gap-3">
+        {% for status, label in [('pass', '通过'), ('warn', '警告'), ('fail', '失败')] %}
+        <div class="rounded-xl bg-white p-4 shadow dark:bg-slate-800">
+            <div class="text-sm text-slate-500 dark:text-slate-400">{{ label }}</div>
+            <div class="mt-1 text-3xl font-bold {% if status == 'pass' %}text-emerald-600{% elif status == 'warn' %}text-amber-600{% else %}text-red-600{% endif %}">
+                {{ report.totals[status] }}
+            </div>
+        </div>
+        {% endfor %}
+    </div>
+
+    <div class="rounded-xl bg-white p-5 shadow dark:bg-slate-800">
+        <h2 class="mb-4 text-lg font-semibold text-slate-900 dark:text-white">数据规模</h2>
+        <div class="grid grid-cols-2 gap-3 sm:grid-cols-3 lg:grid-cols-5">
+            {% for key, value in report.counts.items() %}
+            <div class="rounded-lg bg-slate-50 p-3 dark:bg-slate-900/60">
+                <div class="truncate text-xs uppercase tracking-wide text-slate-500">{{ key|replace('_', ' ') }}</div>
+                <div class="mt-1 text-xl font-semibold text-slate-900 dark:text-white">{{ "{:,}".format(value) }}</div>
+            </div>
+            {% endfor %}
+        </div>
+    </div>
+
+    <div class="overflow-hidden rounded-xl bg-white shadow dark:bg-slate-800">
+        <div class="border-b border-slate-200 px-5 py-4 dark:border-slate-700">
+            <h2 class="text-lg font-semibold text-slate-900 dark:text-white">校验项目</h2>
+        </div>
+        <div class="divide-y divide-slate-100 dark:divide-slate-700">
+            {% for check in report.checks %}
+            <div class="flex flex-col gap-2 px-5 py-4 sm:flex-row sm:items-center sm:justify-between">
+                <div>
+                    <div class="font-medium text-slate-900 dark:text-white">{{ check.name }}</div>
+                    <div class="mt-1 text-sm text-slate-500 dark:text-slate-400">{{ check.detail }}</div>
+                </div>
+                <span class="self-start rounded-full px-3 py-1 text-xs font-semibold uppercase sm:self-center {{ status_styles[check.status] }}">
+                    {{ check.status }}
+                </span>
+            </div>
+            {% endfor %}
+        </div>
+    </div>
+
+    <div>
+        <a href="{{ url_for('admin.dashboard') }}" class="text-sm font-medium text-yrtv-600 hover:text-yrtv-500">
+            返回管理后台
+        </a>
+    </div>
+</div>
+{% endblock %}
@@ -0,0 +1,153 @@
+{% extends "base.html" %}
+
+{% block title %}比赛导入 - YRTV{% endblock %}
+
+{% block content %}
+<div class="space-y-6 px-4 sm:px-0">
+    {% with messages = get_flashed_messages(with_categories=true) %}
+    {% if messages %}
+    <div class="space-y-2">
+        {% for category, message in messages %}
+        <div class="rounded-lg px-4 py-3 text-sm {% if category == 'success' %}bg-emerald-100 text-emerald-800{% elif category == 'warning' %}bg-amber-100 text-amber-800{% else %}bg-red-100 text-red-800{% endif %}">
+            {{ message }}
+        </div>
+        {% endfor %}
+    </div>
+    {% endif %}
+    {% endwith %}
+
+    <div class="flex items-center justify-between">
+        <div>
+            <h1 class="text-2xl font-bold text-slate-900 dark:text-white">比赛数据导入</h1>
+            <p class="mt-1 text-sm text-slate-500 dark:text-slate-400">
+                上传完整的 iframe_network.json，系统会自动识别比赛 ID，并执行 L1 → L2 → L3。
+            </p>
+        </div>
+        <a href="{{ url_for('admin.dashboard') }}" class="text-sm font-medium text-yrtv-600 hover:text-yrtv-500">
+            返回管理后台
+        </a>
+    </div>
+
+    <div class="grid grid-cols-1 gap-6 lg:grid-cols-3">
+        <div class="rounded-xl bg-white p-6 shadow dark:bg-slate-800">
+            <h2 class="text-lg font-semibold text-slate-900 dark:text-white">上传抓包</h2>
+            <form method="POST" enctype="multipart/form-data" class="mt-5 space-y-4">
+                <div>
+                    <label class="block text-sm font-medium text-slate-700 dark:text-slate-300">
+                        iframe_network.json
+                    </label>
+                    <input type="file" name="capture" accept=".json,application/json" required
+                           class="mt-2 block w-full text-sm text-slate-600 file:mr-4 file:rounded-lg file:border-0 file:bg-yrtv-50 file:px-4 file:py-2 file:font-medium file:text-yrtv-700 hover:file:bg-yrtv-100 dark:text-slate-300">
+                </div>
+                <label class="flex items-start gap-2 text-sm text-slate-600 dark:text-slate-300">
+                    <input type="checkbox" name="replace" value="1" class="mt-1 rounded border-slate-300 text-yrtv-600">
+                    <span>允许替换已存在但内容不同的比赛。流水线失败时会自动恢复数据库。</span>
+                </label>
+                <button type="submit"
+                        class="w-full rounded-lg bg-yrtv-600 px-4 py-2.5 font-medium text-white hover:bg-yrtv-500">
+                    校验并开始导入
+                </button>
+            </form>
+            <div class="mt-5 rounded-lg bg-amber-50 p-3 text-xs text-amber-800 dark:bg-amber-900/30 dark:text-amber-200">
+                直接重复上传会被拒绝。每次流水线运行前都会备份 L1、L2、L3，并执行后置完整性检查。
+            </div>
+        </div>
+
+        <div class="rounded-xl bg-white p-6 shadow dark:bg-slate-800 lg:col-span-2">
+            <div class="flex items-center justify-between">
+                <h2 class="text-lg font-semibold text-slate-900 dark:text-white">作业状态</h2>
+                {% if selected_job %}
+                <span id="job-status" class="rounded-full bg-slate-100 px-3 py-1 text-xs font-semibold uppercase text-slate-700 dark:bg-slate-700 dark:text-slate-200">
+                    {{ selected_job.status }}
+                </span>
+                {% endif %}
+            </div>
+
+            {% if selected_job %}
+            <div class="mt-5">
+                <div class="flex justify-between text-sm text-slate-600 dark:text-slate-300">
+                    <span id="job-stage">{{ selected_job.current_stage or 'queued' }}</span>
+                    <span id="job-progress-label">{{ selected_job.progress }}%</span>
+                </div>
+                <div class="mt-2 h-2 overflow-hidden rounded-full bg-slate-200 dark:bg-slate-700">
+                    <div id="job-progress" class="h-full bg-yrtv-500 transition-all" style="width: {{ selected_job.progress }}%"></div>
+                </div>
+                <p id="job-message" class="mt-3 text-sm text-slate-600 dark:text-slate-300">
+                    {{ selected_job.message or '等待执行' }}
+                </p>
+                <pre id="job-log" class="mt-4 max-h-96 overflow-auto whitespace-pre-wrap rounded-lg bg-slate-950 p-4 text-xs text-slate-200">{{ selected_job.log_text }}</pre>
+            </div>
+            {% else %}
+            <div class="mt-8 rounded-lg border border-dashed border-slate-300 p-8 text-center text-sm text-slate-500 dark:border-slate-600">
+                上传比赛或从下方选择历史作业。
+            </div>
+            {% endif %}
+        </div>
+    </div>
+
+    <div class="overflow-hidden rounded-xl bg-white shadow dark:bg-slate-800">
+        <div class="border-b border-slate-200 px-5 py-4 dark:border-slate-700">
+            <h2 class="font-semibold text-slate-900 dark:text-white">最近作业</h2>
+        </div>
+        <div class="overflow-x-auto">
+            <table class="min-w-full divide-y divide-slate-200 dark:divide-slate-700">
+                <thead class="bg-slate-50 dark:bg-slate-900/50">
+                    <tr>
+                        <th class="px-5 py-3 text-left text-xs font-semibold uppercase text-slate-500">ID</th>
+                        <th class="px-5 py-3 text-left text-xs font-semibold uppercase text-slate-500">类型</th>
+                        <th class="px-5 py-3 text-left text-xs font-semibold uppercase text-slate-500">比赛</th>
+                        <th class="px-5 py-3 text-left text-xs font-semibold uppercase text-slate-500">状态</th>
+                        <th class="px-5 py-3 text-left text-xs font-semibold uppercase text-slate-500">阶段</th>
+                        <th class="px-5 py-3 text-left text-xs font-semibold uppercase text-slate-500">耗时</th>
+                    </tr>
+                </thead>
+                <tbody class="divide-y divide-slate-100 dark:divide-slate-700">
+                    {% for job in jobs %}
+                    <tr class="hover:bg-slate-50 dark:hover:bg-slate-700/40">
+                        <td class="px-5 py-3 text-sm">
+                            <a href="{{ url_for('admin.import_match', job_id=job.id) }}" class="font-medium text-yrtv-600">#{{ job.id }}</a>
+                        </td>
+                        <td class="px-5 py-3 text-sm text-slate-600 dark:text-slate-300">{{ job.job_type }}</td>
+                        <td class="px-5 py-3 text-sm font-mono text-slate-600 dark:text-slate-300">{{ job.match_id or '-' }}</td>
+                        <td class="px-5 py-3 text-sm text-slate-600 dark:text-slate-300">{{ job.status }}</td>
+                        <td class="px-5 py-3 text-sm text-slate-600 dark:text-slate-300">{{ job.current_stage or '-' }}</td>
+                        <td class="px-5 py-3 text-sm text-slate-600 dark:text-slate-300">{{ '%.2fs'|format(job.duration_seconds) if job.duration_seconds is not none else '-' }}</td>
+                    </tr>
+                    {% else %}
+                    <tr><td colspan="6" class="px-5 py-8 text-center text-sm text-slate-500">暂无作业</td></tr>
+                    {% endfor %}
+                </tbody>
+            </table>
+        </div>
+    </div>
+</div>
+{% endblock %}
+
+{% block scripts %}
+{% if selected_job %}
+<script>
+const jobId = {{ selected_job.id }};
+let pollTimer = null;
+
+async function refreshJob() {
+    const response = await fetch(`/admin/api/jobs/${jobId}`);
+    if (!response.ok) return;
+    const job = await response.json();
+    document.getElementById('job-status').textContent = job.status;
+    document.getElementById('job-stage').textContent = job.current_stage || job.status;
+    document.getElementById('job-progress-label').textContent = `${job.progress}%`;
+    document.getElementById('job-progress').style.width = `${job.progress}%`;
+    document.getElementById('job-message').textContent = job.message || '';
+    const log = document.getElementById('job-log');
+    log.textContent = job.log_text || '';
+    log.scrollTop = log.scrollHeight;
+    if (job.status === 'succeeded' || job.status === 'failed') {
+        clearInterval(pollTimer);
+    }
+}
+
+pollTimer = setInterval(refreshJob, 1500);
+refreshJob();
+</script>
+{% endif %}
+{% endblock %}
@@ -0,0 +1,88 @@
+<div class="grid grid-cols-1 gap-8 xl:grid-cols-2">
+    <section class="rounded-2xl border border-gray-100 bg-white p-6 shadow-lg dark:border-slate-700 dark:bg-slate-800"
+             x-data="{ selected: 'last_20', periods: {{ period_stats|tojson }} }">
+        <div class="flex flex-col gap-4 sm:flex-row sm:items-center sm:justify-between">
+            <div>
+                <h3 class="text-lg font-bold text-gray-900 dark:text-white">阶段表现</h3>
+                <p class="text-xs text-gray-500">窗口按该玩家最新一场比赛向前计算</p>
+            </div>
+            <select x-model="selected"
+                    class="rounded-lg border-gray-200 bg-gray-50 text-sm dark:border-slate-600 dark:bg-slate-700 dark:text-white">
+                {% for period in period_stats %}
+                <option value="{{ period.period_key }}">{{ period.period_label }}</option>
+                {% endfor %}
+            </select>
+        </div>
+
+        {% for period in period_stats %}
+        <div x-show="selected === '{{ period.period_key }}'"
+             {% if period.period_key != 'last_20' %}style="display:none"{% endif %}
+             class="mt-6">
+            <div class="grid grid-cols-2 gap-3 sm:grid-cols-3">
+                <div class="rounded-xl bg-gray-50 p-3 dark:bg-slate-700/40">
+                    <div class="text-xs font-bold uppercase text-gray-400">Matches</div>
+                    <div class="mt-1 text-2xl font-black text-gray-900 dark:text-white">{{ period.matches }}</div>
+                </div>
+                <div class="rounded-xl bg-gray-50 p-3 dark:bg-slate-700/40">
+                    <div class="text-xs font-bold uppercase text-gray-400">Rating</div>
+                    <div class="mt-1 text-2xl font-black text-yrtv-600">{{ '%.2f'|format(period.avg_rating or 0) }}</div>
+                </div>
+                <div class="rounded-xl bg-gray-50 p-3 dark:bg-slate-700/40">
+                    <div class="text-xs font-bold uppercase text-gray-400">K/D</div>
+                    <div class="mt-1 text-2xl font-black text-gray-900 dark:text-white">{{ '%.2f'|format(period.avg_kd or 0) }}</div>
+                </div>
+                <div class="rounded-xl bg-gray-50 p-3 dark:bg-slate-700/40">
+                    <div class="text-xs font-bold uppercase text-gray-400">ADR</div>
+                    <div class="mt-1 text-2xl font-black text-gray-900 dark:text-white">{{ '%.1f'|format(period.avg_adr or 0) }}</div>
+                </div>
+                <div class="rounded-xl bg-gray-50 p-3 dark:bg-slate-700/40">
+                    <div class="text-xs font-bold uppercase text-gray-400">KAST</div>
+                    <div class="mt-1 text-2xl font-black text-gray-900 dark:text-white">{{ '%.1f%%'|format((period.avg_kast or 0) * 100) }}</div>
+                </div>
+                <div class="rounded-xl bg-gray-50 p-3 dark:bg-slate-700/40">
+                    <div class="text-xs font-bold uppercase text-gray-400">Win Rate</div>
+                    <div class="mt-1 text-2xl font-black {% if period.win_rate >= 0.5 %}text-green-600{% else %}text-red-500{% endif %}">
+                        {{ '%.0f%%'|format((period.win_rate or 0) * 100) }}
+                    </div>
+                </div>
+            </div>
+            <div class="mt-3 text-xs text-gray-400">
+                {% if period.sample_reliable %}
+                    样本充足
+                {% else %}
+                    样本不足，仅供参考
+                {% endif %}
+            </div>
+        </div>
+        {% else %}
+        <div class="mt-8 text-center text-sm text-gray-400">暂无阶段统计</div>
+        {% endfor %}
+    </section>
+
+    <section class="rounded-2xl border border-gray-100 bg-white p-6 shadow-lg dark:border-slate-700 dark:bg-slate-800">
+        <div>
+            <h3 class="text-lg font-bold text-gray-900 dark:text-white">职业纪录</h3>
+            <p class="text-xs text-gray-500">每项纪录都可追溯到具体比赛</p>
+        </div>
+        <div class="mt-6 grid grid-cols-2 gap-3 sm:grid-cols-3">
+            {% for record in records %}
+            <a href="{{ url_for('matches.detail', match_id=record.match_id) }}"
+               class="rounded-xl border border-gray-100 bg-gray-50 p-3 transition hover:border-yrtv-300 hover:bg-yrtv-50 dark:border-slate-600 dark:bg-slate-700/40 dark:hover:bg-slate-700">
+                <div class="truncate text-xs font-bold uppercase text-gray-400">{{ record.record_label }}</div>
+                <div class="mt-1 text-2xl font-black text-gray-900 dark:text-white">
+                    {% if record.record_key in ['most_kills', 'most_headshots', 'longest_win_streak'] %}
+                        {{ record.record_value|int }}
+                    {% elif record.record_key == 'highest_adr' %}
+                        {{ '%.1f'|format(record.record_value) }}
+                    {% else %}
+                        {{ '%.2f'|format(record.record_value) }}
+                    {% endif %}
+                </div>
+                <div class="mt-2 truncate text-[10px] font-mono text-gray-400">{{ record.map_name }}</div>
+            </a>
+            {% else %}
+            <div class="col-span-full py-8 text-center text-sm text-gray-400">暂无职业纪录</div>
+            {% endfor %}
+        </div>
+    </section>
+</div>
@@ -213,6 +213,8 @@
        </div>
    </div>

+    {% include "players/_career_dashboard.html" %}
+
    <!-- 2. Charts Section (Middle) -->
    <div class="grid grid-cols-1 lg:grid-cols-3 gap-8">
        <!-- Trend Chart -->
@@ -221,8 +223,11 @@
                <h3 class="text-lg font-bold text-gray-900 dark:text-white flex items-center gap-2">
                    <span>📈</span> 近期表现走势 (Performance Trend)
                </h3>
-                <div class="flex bg-gray-100 dark:bg-slate-700 rounded-lg p-1">
-                    <button class="px-3 py-1 text-xs font-bold rounded-md bg-white dark:bg-slate-600 shadow-sm text-gray-800 dark:text-white">Recent 20</button>
+                <div id="trend-period-buttons" class="flex flex-wrap bg-gray-100 dark:bg-slate-700 rounded-lg p-1">
+                    <button onclick="loadTrendPeriod('last_10', this)" class="px-3 py-1 text-xs font-bold rounded-md text-gray-500">10</button>
+                    <button onclick="loadTrendPeriod('last_20', this)" class="trend-active px-3 py-1 text-xs font-bold rounded-md bg-white dark:bg-slate-600 shadow-sm text-gray-800 dark:text-white">20</button>
+                    <button onclick="loadTrendPeriod('last_30', this)" class="px-3 py-1 text-xs font-bold rounded-md text-gray-500">30</button>
+                    <button onclick="loadTrendPeriod('career', this)" class="px-3 py-1 text-xs font-bold rounded-md text-gray-500">Career</button>
                </div>
            </div>
            <div class="relative h-80 w-full">
@@ -256,7 +261,7 @@
        </div>

        {% macro detail_item(label, value, key, format_str='{:.2f}', sublabel=None, count_label=None) %}
-            {% set dist = distribution[key] if distribution else None %}
+            {% set dist = distribution[key] if distribution and distribution[key] else None %}
            <div class="flex flex-col group relative h-full p-2 rounded hover:bg-gray-50 dark:hover:bg-slate-700/30 transition-colors">
                <div class="flex justify-between items-center mb-1">
                    <span class="text-xs font-bold text-gray-400 uppercase tracking-wider truncate max-w-[150px]" title="{{ label }}">{{ label }}</span>
@@ -273,7 +278,11 @@
                <div class="flex justify-between items-end mb-1">
                    <div class="flex items-baseline gap-1">
                        <span class="text-lg font-black text-gray-900 dark:text-white font-mono">
-                            {{ format_str.format(value if value is not none else 0) }}
+                            {% if value is none %}
+                                <span class="text-sm text-gray-400">N/A</span>
+                            {% else %}
+                                {{ format_str.format(value) }}
+                            {% endif %}
                        </span>
                        {% if sublabel %}
                        <span class="text-[10px] text-gray-400">{{ sublabel }}</span>
@@ -834,6 +843,7 @@
 {% block scripts %}
 <script>
 let trendChartInstance = null;
+const profileSteamId = "{{ player.steam_id_64 }}";

 function resetZoom() {
    if (trendChartInstance) {
@@ -853,8 +863,29 @@ function likeComment(commentId, btn) {
        });
 }

+function loadTrendPeriod(periodKey, button) {
+    fetch(`/players/${profileSteamId}/charts_data?period=${periodKey}`)
+        .then(response => response.json())
+        .then(data => {
+            if (!trendChartInstance) return;
+            trendChartInstance.data.labels = data.trend.labels;
+            trendChartInstance.data.datasets[0].data = data.trend.values;
+            trendChartInstance.data.datasets[1].data = Array(data.trend.labels.length).fill(1.5);
+            trendChartInstance.data.datasets[2].data = Array(data.trend.labels.length).fill(1.0);
+            trendChartInstance.data.datasets[3].data = Array(data.trend.labels.length).fill(0.6);
+            trendChartInstance.update();
+
+            document.querySelectorAll('#trend-period-buttons button').forEach(item => {
+                item.classList.remove('bg-white', 'dark:bg-slate-600', 'shadow-sm', 'text-gray-800', 'dark:text-white');
+                item.classList.add('text-gray-500');
+            });
+            button.classList.remove('text-gray-500');
+            button.classList.add('bg-white', 'dark:bg-slate-600', 'shadow-sm', 'text-gray-800', 'dark:text-white');
+        });
+}
+
 document.addEventListener('DOMContentLoaded', function() {
-    const steamId = "{{ player.steam_id_64 }}";
+    const steamId = profileSteamId;
    
    fetch(`/players/${steamId}/charts_data`)
        .then(response => response.json())
				`@@ -0,0 +1,2 @@`
				`"""Database builders, schemas, maintenance tools and local data stores."""`