Add Entry opportunity training diagnostics

training/scripts/11_train_small_models.py

@@ -11,6 +11,13 @@ def main() -> None:
     parser = argparse.ArgumentParser()
     add_common_args(parser)
     parser.add_argument("--max-rows", type=int, default=0)
+    parser.add_argument(
+        "--conditional-entry-source",
+        choices=("none", "direction_label", "side_opportunity"),
+        default="none",
+        help="Entry 训练样本人群来源：不筛选、按 Direction 标签筛选、或按本方向未来机会阈值筛选。",
+    )
+    parser.add_argument("--conditional-entry-opportunity-bps", type=float, default=40.0)
     parser.add_argument("--conditional-entry-direction-labels", action="store_true")
     parser.add_argument("--conditional-entry-min-fit-rows", type=int, default=1000)
     args = parser.parse_args()

training/scripts/30_diagnose_good_trade_structure.py

@@ -0,0 +1,26 @@
+from __future__ import annotations
+
+import argparse
+
+import _bootstrap  # noqa: F401
+from trader_training.good_trade_structure import diagnose_good_trade_structure
+from trader_training.io_utils import add_common_args, setup_logging
+
+
+def _float_tuple(value: str) -> tuple[float, ...]:
+    return tuple(float(item.strip()) for item in value.split(",") if item.strip())
+
+
+def main() -> None:
+    parser = argparse.ArgumentParser(description="Diagnose whether existing features separate good and bad Entry trades.")
+    add_common_args(parser)
+    parser.add_argument("--min-good-edge-bps", type=float, default=3.0)
+    parser.add_argument("--bad-edge-bps", type=float, default=-3.0)
+    parser.add_argument("--top-fractions", type=_float_tuple, default=(0.01, 0.05, 0.10))
+    args = parser.parse_args()
+    setup_logging()
+    diagnose_good_trade_structure(args)
+
+
+if __name__ == "__main__":
+    main()