Update ml_engine/data_manager.py

ml_engine/data_manager.py

@@ -1,6 +1,6 @@
 # ============================================================
 # 📂 ml_engine/data_manager.py 
-# (V60.4 - GEM-Architect: Regime Gating + Robust Data)
+# (V61.0 - GEM-Architect: Dual-Opportunity Classification)
 # ============================================================
 
 import asyncio
@@ -44,7 +44,7 @@ class DataManager:
             'UP', 'DOWN', 'BEAR', 'BULL', '3S', '3L'
         ]
         
-        print(f"📦 [DataManager V60.4] Regime Gating (Range Protection) Active.")
+        print(f"📦 [DataManager V61.0] Dual Classification (Safe Bottoms & Momentum) Active.")
 
     async def initialize(self):
         print("   > [DataManager] Starting initialization...")
@@ -73,13 +73,13 @@ class DataManager:
         return self.contracts_db
 
     # ==================================================================
-    # 🧠 Layer 1: Screening + Diagnosis + Regime Gating
+    # 🧠 Layer 1: Classification (Bottom vs Momentum)
     # ==================================================================
     async def layer1_rapid_screening(self, adaptive_hub_ref=None) -> List[Dict[str, Any]]:
         self.adaptive_hub_ref = adaptive_hub_ref
-        print(f"🔍 [Layer 1] Screening with Regime Gating...")
+        print(f"🔍 [Layer 1] Screening for SAFE BOTTOMS & MOMENTUM LAUNCHES...")
         
-        # 1. فلتر السيولة
+        # 1. فلتر السيولة الأساسي
         initial_candidates = await self._stage0_universe_filter()
         if not initial_candidates:
             print("⚠️ [Layer 1] Stage 0 returned 0 candidates.")
@@ -92,41 +92,99 @@ class DataManager:
         final_list = []
 
         for item in enriched_data:
-            # 3. التصنيف الفني (Breakout vs Reversal)
-            classification = self._apply_strict_logic_tree(item)
+            # 3. التصنيف الجديد: قاع آمن أم انفجار وشيك؟
+            classification = self._classify_opportunity_type(item)
             
             if classification['type'] != 'NONE':
-                # 4. التشخيص (Diagnosis)
+                # تشخيص حالة السوق العامة (Regime) يبقى كما هو للأمان العام
                 regime_info = self._diagnose_asset_regime(item)
                 current_regime = regime_info['regime']
                 
-                # 🔥 5. Regime Gating (بوابة النظام - الحماية من المصيدة)
-                # إذا السوق عرضي (RANGE) أو ميت (DEAD)، نمنع الاختراقات (BREAKOUT)
-                # لأن الاختراقات تفشل في هذه الظروف وتصبح مصيدة ثيران.
-                if current_regime in ['RANGE', 'DEAD'] and classification['type'] == 'BREAKOUT':
-                    # تخطي بصمت (حماية)
-                    continue
+                # 🔥 Regime Gating: الحماية من المصائد
+                # إذا السوق عرضي (RANGE) أو ميت (DEAD)، نمنع صفقات الزخم (MOMENTUM) لأنها غالباً مصائد
+                # لكن نسمح بصفقات القيعان (SAFE_BOTTOM) لأنها آمنة في التذبذب
+                if current_regime in ['RANGE', 'DEAD'] and classification['type'] == 'MOMENTUM_LAUNCH':
+                    continue 
 
-                # إذا مر من البوابة، نتابع
+                # إضافة البيانات
                 item['asset_regime'] = current_regime
                 item['asset_regime_conf'] = regime_info['conf']
                 
-                # حقن العتبات
+                # 🏷️ الوسم الجديد للاستراتيجية
+                item['strategy_type'] = classification['type'] 
+                item['l1_sort_score'] = classification['score']
+                
+                # حقن العتبات الديناميكية
                 if self.adaptive_hub_ref:
                     dynamic_config = self.adaptive_hub_ref.get_regime_config(current_regime)
                     item['dynamic_limits'] = dynamic_config
                 
-                item['l1_sort_score'] = classification['score'] 
+                # حفظ التاج القديم للتوافق
                 item['strategy_tag'] = classification['type']
+                
                 final_list.append(item)
 
-        # 6. الترتيب النهائي
+        # 4. الترتيب النهائي
         final_list.sort(key=lambda x: x['l1_sort_score'], reverse=True)
         
         selection = final_list[:50]
-        print(f"✅ [Layer 1] Passed {len(selection)} candidates (Safe Strategies Only).")
+        print(f"✅ [Layer 1] Passed {len(selection)} candidates (Types: Bottom/Launch).")
         return selection
 
+    # ==================================================================
+    # ⚖️ The Dual-Classifier Logic
+    # ==================================================================
+    def _classify_opportunity_type(self, data: Dict[str, Any]) -> Dict[str, Any]:
+        """
+        تقسيم العملات إلى نوعين:
+        1. SAFE_BOTTOM: عملات في القاع (آمنة من الهبوط الحاد).
+        2. MOMENTUM_LAUNCH: عملات تجهز للانفجار (نسبة 90%).
+        """
+        try:
+            df_1h = self._calc_indicators(data['ohlcv_1h_raw'])
+            # يمكن استخدام فريم 15 دقيقة للتأكيد الإضافي مستقبلاً
+            # df_15m = self._calc_indicators(data['ohlcv_15m_raw']) 
+            curr = df_1h.iloc[-1]
+        except: return {'type': 'NONE', 'score': 0}
+
+        # --- المؤشرات الأساسية ---
+        rsi = curr['rsi']
+        close = curr['close']
+        ema50 = curr['ema50']
+        ema200 = curr['ema200'] if 'ema200' in curr else ema50 
+        
+        # Bollinger Bands Check
+        lower_bb = curr['lower_bb'] if 'lower_bb' in curr else (curr['ema20'] - (2*curr['atr']))
+        upper_bb = curr['upper_bb'] if 'upper_bb' in curr else (curr['ema20'] + (2*curr['atr']))
+        
+        # 🛡️ TYPE 1: SAFE_BOTTOM (The Safety Net)
+        # الشروط: RSI منخفض جداً، السعر عند أو تحت الحد السفلي، بعيد عن المتوسطات
+        # هذا النوع آمن لأن الهبوط الإضافي صعب إحصائياً (Mean Reversion)
+        if rsi < 35:
+            # التأكد من الابتعاد عن المتوسط (Oversold Extension)
+            dist_from_ema = (ema50 - close) / ema50
+            
+            # السعر عند الدعم السفلي أو تحته بقليل
+            if close <= lower_bb * 1.015 and dist_from_ema > 0.03:
+                # Score: كلما كان RSI أقل، كلما كان الارتداد أقوى
+                score = (50 - rsi) / 20.0 
+                return {'type': 'SAFE_BOTTOM', 'score': min(score, 1.0)}
+
+        # 🚀 TYPE 2: MOMENTUM_LAUNCH (The Rocket)
+        # الشروط: RSI صحي وقوي (فوق 55 وتحت 75)، السعر فوق المتوسطات، ضغط سعري (Squeeze)
+        # هذا النوع يجهز للانفجار للأعلى
+        elif 55 < rsi < 75:
+            if close > ema50 and close > ema200:
+                # السعر يضغط قرب الحد العلوي (Squeeze before breakout)
+                dist_to_upper = (upper_bb - close) / close
+                
+                # قريب جداً من الانفجار (أقل من 3% للحد العلوي)
+                if dist_to_upper < 0.03: 
+                    score = rsi / 100.0
+                    return {'type': 'MOMENTUM_LAUNCH', 'score': score}
+
+        return {'type': 'NONE', 'score': 0}
+
     # ==================================================================
     # 🔍 Stage 0: Universe Filter (Robust USD Calc)
     # ==================================================================
@@ -140,7 +198,6 @@ class DataManager:
             SOVEREIGN_COINS = ['BTC/USDT', 'ETH/USDT', 'SOL/USDT', 'BNB/USDT', 'XRP/USDT']
             
             reject_stats = {"volume": 0, "change": 0, "blacklist": 0}
-            debug_printed = False
 
             for symbol, ticker in tickers.items():
                 if not symbol.endswith('/USDT'): continue
@@ -157,22 +214,18 @@ class DataManager:
                 
                 is_sovereign = symbol in SOVEREIGN_COINS
                 
-                # طباعة فحص BTC مرة واحدة
-                if "BTC/USDT" in symbol and not debug_printed:
-                    print(f"   🐛 [DEBUG] BTC Vol: ${calc_quote_vol:,.0f}")
-                    debug_printed = True
-
                 # فلتر السيولة (500k) - نتجاوزه للعملات السيادية
                 if not is_sovereign:
                     if calc_quote_vol < 500000: 
                         reject_stats["volume"] += 1
                         continue
                 
-                # فلتر التغير (20%)
+                # فلتر التغير (20%) - نتجاوزه إذا كنا نبحث عن قيعان (قد تكون هابطة) لكن بحذر
+                # هنا نبقي الفلتر العام للحماية من الانهيارات التامة
                 change_pct = ticker.get('percentage')
                 if change_pct is None: change_pct = 0.0
                 
-                if abs(change_pct) > 20.0: 
+                if abs(change_pct) > 25.0: 
                     reject_stats["change"] += 1
                     continue
                 
@@ -183,9 +236,6 @@ class DataManager:
                     'change_24h': change_pct
                 })
             
-            print(f"   📊 [Filter Stats] Total: {len(tickers)} | Passed: {len(candidates)}")
-            print(f"      ❌ Rejected: Vol < 500k ({reject_stats['volume']}) | Change > 20% ({reject_stats['change']})")
-
             candidates.sort(key=lambda x: x['quote_volume'], reverse=True)
             return candidates
             
@@ -225,46 +275,7 @@ class DataManager:
         except Exception: return {'regime': 'RANGE', 'conf': 0.0}
 
     # ------------------------------------------------------------------
-    # 🛡️ The Logic Tree (Anti-FOMO Tuned)
-    # ------------------------------------------------------------------
-    def _apply_strict_logic_tree(self, data: Dict[str, Any]) -> Dict[str, Any]:
-        try:
-            df_1h = self._calc_indicators(data['ohlcv_1h_raw'])
-            df_15m = self._calc_indicators(data['ohlcv_15m_raw'])
-            data['df_1h'] = df_1h
-        except: return {'type': 'NONE', 'score': 0}
-
-        curr_1h = df_1h.iloc[-1]
-        curr_15m = df_15m.iloc[-1]
-        
-        try:
-            close_4h_ago = df_1h.iloc[-5]['close']
-            change_4h = ((curr_1h['close'] - close_4h_ago) / close_4h_ago) * 100
-        except: change_4h = 0.0
-
-        # Gates
-        if change_4h > 12.0: return {'type': 'NONE', 'score': 0}
-        if curr_1h['rsi'] > 75: return {'type': 'NONE', 'score': 0} 
-        dev = (curr_1h['close'] - curr_1h['ema20']) / curr_1h['atr'] if curr_1h['atr'] > 0 else 0
-        if dev > 2.2: return {'type': 'NONE', 'score': 0}
-
-        # A. Breakout
-        is_bullish = (curr_1h['ema20'] > curr_1h['ema50']) or (curr_1h['close'] > curr_1h['ema20'])
-        if is_bullish and (45 <= curr_1h['rsi'] <= 75): 
-            vol_ma = df_15m['volume'].rolling(20).mean().iloc[-1]
-            if curr_15m['volume'] >= 1.2 * vol_ma: 
-                score = curr_15m['volume'] / vol_ma if vol_ma > 0 else 1.0
-                return {'type': 'BREAKOUT', 'score': score}
-
-        # B. Reversal
-        if 20 <= curr_1h['rsi'] <= 40 and change_4h <= -2.0:
-            score = (100 - curr_1h['rsi'])
-            return {'type': 'REVERSAL', 'score': score}
-
-        return {'type': 'NONE', 'score': 0}
-
-    # ------------------------------------------------------------------
-    # Helpers
+    # Helpers & Indicators (Updated for BB & EMA200)
     # ------------------------------------------------------------------
     async def _fetch_technical_data_batch(self, candidates):
         chunk_size = 10; results = []
@@ -278,7 +289,7 @@ class DataManager:
 
     async def _fetch_single(self, c):
         try:
-            h1 = await self.exchange.fetch_ohlcv(c['symbol'], '1h', limit=60)
+            h1 = await self.exchange.fetch_ohlcv(c['symbol'], '1h', limit=210) # زاد العدد لحساب EMA200 بدقة
             m15 = await self.exchange.fetch_ohlcv(c['symbol'], '15m', limit=60)
             if not h1 or not m15: return None
             c['ohlcv'] = {'1h': h1, '15m': m15} 
@@ -294,10 +305,21 @@ class DataManager:
         loss = (-delta.where(delta<0, 0)).rolling(14).mean()
         rs = gain/loss
         df['rsi'] = 100 - (100/(1+rs))
+        
+        # EMAs
         df['ema20'] = df['c'].ewm(span=20).mean()
         df['ema50'] = df['c'].ewm(span=50).mean()
+        df['ema200'] = df['c'].ewm(span=200).mean() # مهم لنوع الزخم
+        
+        # ATR
         tr = np.maximum(df['h']-df['l'], np.maximum(abs(df['h']-df['c'].shift()), abs(df['l']-df['c'].shift())))
         df['atr'] = tr.rolling(14).mean()
+        
+        # Bollinger Bands (Standard: 20, 2)
+        std = df['c'].rolling(20).std()
+        df['upper_bb'] = df['ema20'] + (2 * std)
+        df['lower_bb'] = df['ema20'] - (2 * std)
+        
         df.rename(columns={'o':'open', 'h':'high', 'l':'low', 'c':'close', 'v':'volume'}, inplace=True)
         return df.fillna(0)