Require actual plan edge in Entry screening

training/tests/test_training_contract.py

@@ -50,6 +50,12 @@ class TrainingContractTest(unittest.TestCase):
         self.assertEqual("long_actual_plan_net_edge_bps", _screen_edge_column(dataset, "LONG"))
         self.assertEqual("short_actual_plan_net_edge_bps", _screen_edge_column(dataset, "SHORT"))
 
+    def test_entry_feature_screen_requires_actual_plan_edge(self) -> None:
+        dataset = pd.DataFrame({"long_expected_net_edge_bps": [20.0]})
+
+        with self.assertRaises(ValueError):
+            _screen_edge_column(dataset, "LONG")
+
     def test_entry_regression_heads_train_on_actual_plan_edge(self) -> None:
         heads = {head[0]: head[2] for head in TARGETS["ENTRY"]["heads"]}
 

training/trader_training/entry_feature_screen.py

@@ -17,7 +17,7 @@ ALL_SPLITS = (FIT_SPLIT, TUNE_SPLIT, VALIDATION_LOCKED_SPLIT, LATEST_STRESS_SPLI
 def screen_entry_features(args: Any) -> None:
     root = run_root(args)
     dataset = read_parquet(root / "dataset" / "entry_train.parquet")
-    required = {"split_id", *FEATURE_ORDER, "long_entry_target", "short_entry_target", "long_expected_net_edge_bps", "short_expected_net_edge_bps"}
+    required = {"split_id", *FEATURE_ORDER, "long_entry_target", "short_entry_target", "long_actual_plan_net_edge_bps", "short_actual_plan_net_edge_bps"}
     missing = sorted(required.difference(dataset.columns))
     if missing:
         raise ValueError(f"entry feature screen missing required columns: {missing}")
@@ -67,7 +67,7 @@ def _screen_edge_column(dataset: pd.DataFrame, side: str) -> str:
     actual_col = f"{prefix}_actual_plan_net_edge_bps"
     if actual_col in dataset.columns:
         return actual_col
-    return f"{prefix}_expected_net_edge_bps"
+    raise ValueError(f"entry feature screen requires actual plan edge column: {actual_col}")
 
 
 def _split_baselines(dataset: pd.DataFrame, target_col: str, edge_col: str) -> dict[str, dict[str, float]]:
@@ -247,7 +247,7 @@ def _markdown_report(result: dict[str, Any], candidates: pd.DataFrame) -> str:
         "",
         "这份报告只回答一个问题：历史数据里，单个特征的某些区间有没有稳定变好。",
         "",
-        "- 如果数据里有真实出场净收益，本报告用真实出场净收益；没有时才退回训练收益标签。",
+        "- 本报告只使用真实出场净收益；缺少真实收益列时直接失败。",
         "- `tune_inner` 用来挑候选区间。",
         "- `validation_locked` 和 `latest_stress` 用来检查这个区间是不是出了训练样本也还能站住。",
         "- `stable_positive_edge=true` 代表这个区间在三个检查集里的平均净收益都大于 0。",