feat: 添加回测框架

- 新增 src/llm/backtest.py: 回测核心逻辑
  - 查询历史已完赛比赛
  - 逐场预测(自动防未来信息泄漏)
  - 实际比分回填 + 统计
  - 1X2 准确率 / 比分 RMSE / 置信度校准
- 新增 POST /api/v1/backtest 路由
- 支持按联赛/日期范围/模式/模型筛选

src/api/app.py

@@ -41,11 +41,13 @@ def create_app() -> FastAPI:
     from src.api.routes.predict import router as predict_router
     from src.api.routes.ingest import router as ingest_router
     from src.api.routes.eval import router as eval_router
+    from src.api.routes.backtest import router as backtest_router
 
     app.include_router(matches_router)
     app.include_router(predict_router)
     app.include_router(ingest_router)
     app.include_router(eval_router)
+    app.include_router(backtest_router)
 
     @app.get("/health")
     async def health():

src/api/routes/backtest.py

@@ -0,0 +1,69 @@
+"""回测路由。"""
+from __future__ import annotations
+
+from fastapi import APIRouter, HTTPException
+from pydantic import BaseModel, Field
+
+from src.llm.backtest import run_backtest
+
+router = APIRouter(prefix="/api/v1", tags=["backtest"])
+
+
+class BacktestRequest(BaseModel):
+    league_id: int | None = Field(None, description="联赛 ID")
+    date_from: str | None = Field(None, description="起始日期 YYYY-MM-DD")
+    date_to: str | None = Field(None, description="结束日期 YYYY-MM-DD")
+    mode: str = Field("single", description="预测模式: single(快) / multi(多 agent)")
+    limit: int = Field(20, ge=1, le=200, description="最大回测场数")
+    model: str | None = Field(None, description="指定模型 (空=默认)")
+
+
+@router.post("/backtest")
+async def backtest(req: BacktestRequest):
+    """对历史比赛运行回测。
+
+    对每场已完赛比赛:
+      1. 用比赛之前的数据构建上下文 (防未来信息泄漏)
+      2. 调 LLM 预测
+      3. 用实际比分回填
+      4. 统计准确率 / RMSE / 校准度
+    """
+    try:
+        summary = await run_backtest(
+            league_id=req.league_id,
+            date_from=req.date_from,
+            date_to=req.date_to,
+            mode=req.mode,
+            limit=req.limit,
+            model=req.model,
+        )
+    except Exception as e:
+        raise HTTPException(500, f"backtest failed: {e}")
+
+    return {
+        "summary": {
+            "total": summary.total,
+            "scored": summary.scored,
+            "accuracy_1x2": summary.accuracy_1x2,
+            "avg_score_rmse": summary.avg_score_rmse,
+            "avg_confidence": summary.avg_confidence,
+            "calibration": summary.calibration,
+        },
+        "results": [
+            {
+                "match_id": r.match_id,
+                "league_code": r.league_code,
+                "home_team": r.home_team,
+                "away_team": r.away_team,
+                "match_date": r.match_date,
+                "actual_score": f"{r.actual_home}-{r.actual_away}",
+                "actual_1x2": r.actual_1x2,
+                "pred_home": r.pred_home,
+                "pred_away": r.pred_away,
+                "pred_1x2": r.pred_1x2,
+                "confidence": r.confidence,
+                "correct_1x2": r.correct_1x2,
+            }
+            for r in summary.results
+        ],
+    }