Update backtest_engine.py

backtest_engine.py

@@ -1,9 +1,9 @@
 # ============================================================
-# 🧪 backtest_engine.py (V223.4 - GEM-Architect: Clean & Robust)
-#   FIXES APPLIED:
-#   1) Added Slope & MFI calculation (Critical for Oracle & Titan).
-#   2) Fixed Sniper prediction shape mismatch (Broadcasting Error).
-#   3) Removed all citation artifacts/syntax errors.
+# 🧪 backtest_engine.py (V223.5 - GEM-Architect: The Data Floodgates Open)
+#   FIXES:
+#   1) Relaxed Storage Thresholds (Solves "Signals: 0").
+#   2) Includes MFI/Slope/Aliases (Solves Warnings).
+#   3) Includes Sniper Shape Fix (Solves Crash).
 # ============================================================
 
 import asyncio
@@ -197,8 +197,7 @@ class HeavyDutyBacktester:
         }
 
         self.TARGET_COINS = [
-            "SOL/USDT", "XRP/USDT", "DOGE/USDT", "ADA/USDT", "AVAX/USDT", "LINK/USDT",
-                 ]
+            "SOL/USDT", "XRP/USDT", "DOGE/USDT"  ]
 
         self.USE_FIXED_DATES = False
         self.LOOKBACK_DAYS = 14
@@ -206,7 +205,7 @@ class HeavyDutyBacktester:
         self.force_end_date = "2024-02-01"
 
         self.required_timeframes = self._determine_required_timeframes()
-        print(f"🧪 [Backtest V223.4] IMMUTABLE TRUTH (Patched & Clean). TFs: {self.required_timeframes}")
+        print(f"🧪 [Backtest V223.5] IMMUTABLE TRUTH (Patched & Open Gates). TFs: {self.required_timeframes}")
 
     def _verify_system_integrity(self):
         errors = []
@@ -254,7 +253,7 @@ class HeavyDutyBacktester:
         return list(tfs)
 
     # --------------------------
-    # Indicator Hardening Layer (FIXED)
+    # Indicator Hardening Layer (FIXED: MFI, Slope, Aliases)
     # --------------------------
     @staticmethod
     def _safe_bbands(close: pd.Series, length=20, std=2.0):
@@ -279,7 +278,7 @@ class HeavyDutyBacktester:
         v = df["volume"].astype(np.float64) if "volume" in df.columns else pd.Series(np.zeros(len(df)), index=df.index)
 
         # ----------------------------------------------------
-        # ✅ [GEM-FIX] جسر التوافق مع Oracle & Titan
+        # ✅ [GEM-FIX] Compatibility Bridge (Oracle & Titan)
         # ----------------------------------------------------
         
         # 1. Oracle: Slope
@@ -291,24 +290,23 @@ class HeavyDutyBacktester:
         except: df['MFI'] = 50.0
 
         # 3. Oracle Mapping (LowerCase Aliases)
-        # يحسب المكتبة RSI، ونحن ننسخها إلى rsi لأن النموذج يطلبها صغيرة
         df["RSI"] = ta.rsi(c, length=14).fillna(50)
-        df["rsi"] = df["RSI"]  # Alias for Oracle
+        df["rsi"] = df["RSI"]  # Alias
 
         df["ATR"] = ta.atr(h, l, c, length=14).fillna(0)
         df["ATR_pct"] = (df["ATR"] / (c + 1e-12)) * 100
-        df["atr_pct"] = df["ATR_pct"] # Alias for Oracle
+        df["atr_pct"] = df["ATR_pct"] # Alias
 
         # 4. Oracle: Volume Z-Score (vol_z)
         vol_mean = v.rolling(20).mean()
         vol_std = v.rolling(20).std()
         df["vol_z"] = ((v - vol_mean) / (vol_std + 1e-9)).fillna(0) # For Oracle
 
-        # 5. Titan: Trend Strong (Approx for lower TFs, perfect for 1d)
+        # 5. Titan: Trend Strong (Approx)
         adx_df = ta.adx(h, l, c, length=14)
         if adx_df is not None and not adx_df.empty:
             df["ADX"] = adx_df.iloc[:, 0].fillna(0)
-            df["Trend_Strong"] = np.where(df["ADX"] > 25, 1, 0) # For Titan 1d
+            df["Trend_Strong"] = np.where(df["ADX"] > 25, 1, 0)
         else:
             df["ADX"] = 0.0
             df["Trend_Strong"] = 0
@@ -334,7 +332,7 @@ class HeavyDutyBacktester:
         for span in [9, 20, 21, 50, 200]:
             df[f"ema{span}"] = c.ewm(span=span, adjust=False).mean()
 
-        # Derived Features
+        # Derived
         df["EMA_9_dist"] = (c / (df["ema9"] + 1e-12)) - 1
         df["EMA_21_dist"] = (c / (df["ema21"] + 1e-12)) - 1
         df["EMA_50_dist"] = (c / (df["ema50"] + 1e-12)) - 1
@@ -344,6 +342,7 @@ class HeavyDutyBacktester:
         # BBANDS
         if len(df) < 30:
             df["lower_bb"] = c; df["upper_bb"] = c; df["bb_width"] = 0.0; df["bb_pct"] = 0.5
+            df["BB_w"] = 0.0; df["BB_p"] = 0.5
         else:
             bb = ta.bbands(c, length=20, std=2.0)
             if bb is not None and isinstance(bb, pd.DataFrame):
@@ -357,7 +356,6 @@ class HeavyDutyBacktester:
                 df["lower_bb"] = bb[col_l[0]] if col_l else c
                 df["upper_bb"] = bb[col_u[0]] if col_u else c
                 
-                # Aliases for Titan
                 df["BB_w"] = df["bb_width"]
                 df["BB_p"] = df["bb_pct"]
             else:
@@ -677,13 +675,9 @@ class HeavyDutyBacktester:
         
         for model in self.proc.sniper.models:
             raw_preds = model.predict(X_sniper)
-            
-            # 🛡️ Handle Multiclass Output (2D Array) vs Binary (1D Array)
             if len(raw_preds.shape) > 1 and raw_preds.shape[1] > 1:
-                # Assuming index 1 is the 'Buy' probability (Standard LightGBM multiclass)
                 preds_accum += raw_preds[:, 1].astype(np.float32)
             else:
-                # Regression or Binary Logic
                 preds_accum += raw_preds.astype(np.float32)
 
         global_sniper_scores = (preds_accum / max(1, len(self.proc.sniper.models))).astype(np.float32)
@@ -723,12 +717,14 @@ class HeavyDutyBacktester:
             ]
         ).astype(np.float32)
 
-        # SAVE
-        min_gov = float(self.GRID_RANGES["GOV_SCORE"][0])
-        min_oracle = float(self.GRID_RANGES["ORACLE"][0])
-        min_titan = float(self.GRID_RANGES["TITAN"][0])
-        min_sniper = float(self.GRID_RANGES["SNIPER"][0])
-        min_pattern = float(self.GRID_RANGES["PATTERN"][0])
+        # ==========================================
+        # 🟢 GEM-FIX: Relaxed Saving Thresholds
+        # ==========================================
+        min_gov = 0.01
+        min_oracle = 0.01
+        min_titan = 0.01
+        min_sniper = 0.01
+        min_pattern = 0.01
 
         filter_mask = (
             validity_mask