fixes

adaptive_bot.py

@@ -331,7 +331,11 @@ class LearningState:
         trade_won = 1 if trade.pnl_usd > 0 else 0
         # Model predicts p(long wins), so short outcomes must be direction-adjusted.
         label_long = trade_won if trade.side == "long" else (1 - trade_won)
-        x_raw = np.asarray(entry_features, dtype=float).reshape(1, -1)
+        # Pad to NUM_FEATURES in case features were stored with an older (shorter) feature set.
+        feat_padded = list(entry_features[:NUM_FEATURES])
+        if len(feat_padded) < NUM_FEATURES:
+            feat_padded.extend([0.0] * (NUM_FEATURES - len(feat_padded)))
+        x_raw = np.asarray(feat_padded, dtype=float).reshape(1, -1)
         y = np.asarray([label_long], dtype=int)
 
         # Always update the scaler with new observations (even bootstrap trades).
@@ -440,7 +444,10 @@ class LearningState:
         """
         if not self.model_ready.get(coin):
             return
-        x_raw = np.asarray(features, dtype=float).reshape(1, -1)
+        feat_padded = list(features[:NUM_FEATURES])
+        if len(feat_padded) < NUM_FEATURES:
+            feat_padded.extend([0.0] * (NUM_FEATURES - len(feat_padded)))
+        x_raw = np.asarray(feat_padded, dtype=float).reshape(1, -1)
         x = self.scalers[coin].transform(x_raw) if self.scaler_ready.get(coin) else x_raw
         x = self.poly.transform(x)
         sw = np.array([_VIRTUAL_SAMPLE_WEIGHT])