Riy777 commited on
Commit
943da29
·
verified ·
1 Parent(s): c383866

Update backtest_engine.py

Browse files
Files changed (1) hide show
  1. backtest_engine.py +139 -153
backtest_engine.py CHANGED
@@ -1,5 +1,5 @@
1
  # ============================================================
2
- # 🧪 backtest_engine.py (V112.0 - GEM-Architect: Matrix Batch Speed)
3
  # ============================================================
4
 
5
  import asyncio
@@ -55,7 +55,7 @@ class HeavyDutyBacktester:
55
  self.force_end_date = None
56
 
57
  if not os.path.exists(CACHE_DIR): os.makedirs(CACHE_DIR)
58
- print(f"🧪 [Backtest V112.0] Matrix-Batch Speed (No Loops Inside Signals).")
59
 
60
  def set_date_range(self, start_str, end_str):
61
  self.force_start_date = start_str
@@ -104,7 +104,7 @@ class HeavyDutyBacktester:
104
  return unique_candles
105
 
106
  # ==============================================================
107
- # 🏎️ VECTORIZED INDICATORS (ALL LAYERS + LAGS)
108
  # ==============================================================
109
  def _calculate_indicators_vectorized(self, df, timeframe='1m'):
110
  df['close'] = df['close'].astype(float)
@@ -118,7 +118,6 @@ class HeavyDutyBacktester:
118
  df['ema50'] = ta.ema(df['close'], length=50)
119
  df['atr'] = ta.atr(df['high'], df['low'], df['close'], length=14)
120
 
121
- # Hydra
122
  if timeframe == '1m':
123
  sma20 = df['close'].rolling(20).mean()
124
  std20 = df['close'].rolling(20).std()
@@ -126,14 +125,12 @@ class HeavyDutyBacktester:
126
  df['vol_ma50'] = df['volume'].rolling(50).mean()
127
  df['rel_vol'] = df['volume'] / (df['vol_ma50'] + 1e-9)
128
 
129
- # Oracle
130
  df['slope'] = ta.slope(df['close'], length=7)
131
  vol_mean = df['volume'].rolling(20).mean()
132
  vol_std = df['volume'].rolling(20).std()
133
  df['vol_z'] = (df['volume'] - vol_mean) / (vol_std + 1e-9)
134
  df['atr_pct'] = df['atr'] / df['close']
135
 
136
- # Sniper
137
  if timeframe == '1m':
138
  df['ret'] = df['close'].pct_change()
139
  df['dollar_vol'] = df['close'] * df['volume']
@@ -159,7 +156,6 @@ class HeavyDutyBacktester:
159
  s = df['volume'].rolling(500).std()
160
  df['vol_zscore_50'] = ((df['volume'] - r) / s).fillna(0)
161
 
162
- # Legacy Structure
163
  df['log_ret'] = np.log(df['close'] / df['close'].shift(1))
164
  roll_max = df['high'].rolling(50).max()
165
  roll_min = df['low'].rolling(50).min()
@@ -173,7 +169,7 @@ class HeavyDutyBacktester:
173
  df['ema200'] = ta.ema(df['close'], length=200)
174
  df['dist_ema200'] = (df['close'] - df['ema200']) / df['close']
175
 
176
- # PRE-CALCULATE LAGS FOR V2 (This enables Batch Processing)
177
  if timeframe == '1m':
178
  for lag in [1, 2, 3, 5, 10, 20]:
179
  df[f'log_ret_lag_{lag}'] = df['log_ret'].shift(lag).fillna(0)
@@ -185,7 +181,7 @@ class HeavyDutyBacktester:
185
  return df
186
 
187
  # ==============================================================
188
- # 🧠 CPU PROCESSING (Matrix Batch Mode)
189
  # ==============================================================
190
  async def _process_data_in_memory(self, sym, candles, start_ms, end_ms):
191
  safe_sym = sym.replace('/', '_')
@@ -196,7 +192,7 @@ class HeavyDutyBacktester:
196
  print(f" 📂 [{sym}] Data Exists -> Skipping.")
197
  return
198
 
199
- print(f" ⚙️ [CPU] Analyzing {sym} (Matrix Batch Mode)...", flush=True)
200
  t0 = time.time()
201
 
202
  df_1m = pd.DataFrame(candles, columns=['timestamp', 'open', 'high', 'low', 'close', 'volume'])
@@ -207,7 +203,7 @@ class HeavyDutyBacktester:
207
  frames = {}
208
  agg_dict = {'open': 'first', 'high': 'max', 'low': 'min', 'close': 'last', 'volume': 'sum'}
209
 
210
- # 1. Calc 1m with Lags
211
  frames['1m'] = self._calculate_indicators_vectorized(df_1m.copy(), timeframe='1m')
212
  frames['1m']['timestamp'] = frames['1m'].index.floor('1min').astype(np.int64) // 10**6
213
  fast_1m = {col: frames['1m'][col].values for col in frames['1m'].columns}
@@ -221,19 +217,93 @@ class HeavyDutyBacktester:
221
  frames[tf_str] = resampled
222
  numpy_htf[tf_str] = {col: resampled[col].values for col in resampled.columns}
223
 
224
- # 3. Create Global Index Maps (The Magic Step for Speed)
225
- # Allows instant mapping from 1m index -> 5m/15m index without searching inside loop
226
- # Using searchsorted on the whole array once
227
  map_1m_to_1h = np.searchsorted(numpy_htf['1h']['timestamp'], fast_1m['timestamp'])
228
  map_1m_to_5m = np.searchsorted(numpy_htf['5m']['timestamp'], fast_1m['timestamp'])
229
  map_1m_to_15m = np.searchsorted(numpy_htf['15m']['timestamp'], fast_1m['timestamp'])
230
 
231
- # Clamp indices to valid range
232
- map_1m_to_1h = np.clip(map_1m_to_1h, 0, len(numpy_htf['1h']['timestamp']) - 1)
233
- map_1m_to_5m = np.clip(map_1m_to_5m, 0, len(numpy_htf['5m']['timestamp']) - 1)
234
- map_1m_to_15m = np.clip(map_1m_to_15m, 0, len(numpy_htf['15m']['timestamp']) - 1)
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
235
 
236
- # 4. L1 Filter
237
  df_1h = frames['1h'].reindex(frames['5m'].index, method='ffill')
238
  df_5m = frames['5m'].copy()
239
  is_valid = (df_1h['rsi'] <= 70)
@@ -242,41 +312,28 @@ class HeavyDutyBacktester:
242
  final_valid_indices = [t for t in valid_indices if t >= start_dt]
243
 
244
  total_signals = len(final_valid_indices)
245
- print(f" 🎯 Candidates: {total_signals}. Running Matrix Models...", flush=True)
246
 
247
- # 5. Load Models
248
- hydra_models = getattr(self.proc.guardian_hydra, 'models', {}) if self.proc.guardian_hydra else {}
249
- hydra_cols = getattr(self.proc.guardian_hydra, 'feature_cols', []) if self.proc.guardian_hydra else []
250
-
251
- legacy_v2 = getattr(self.proc.guardian_legacy, 'model_v2', None)
252
- legacy_v3 = getattr(self.proc.guardian_legacy, 'model_v3', None)
253
- v3_feat_names = getattr(self.proc.guardian_legacy, 'v3_feature_names', [])
254
-
255
  oracle_dir_model = getattr(self.proc.oracle, 'model_direction', None)
256
  oracle_cols = getattr(self.proc.oracle, 'feature_cols', [])
257
  sniper_models = getattr(self.proc.sniper, 'models', [])
258
  sniper_cols = getattr(self.proc.sniper, 'feature_names', [])
259
 
260
- ai_results = []
 
 
 
261
 
262
- # --- 6. Main Simulation Loop (BATCH MODE) ---
263
  for i, current_time in enumerate(final_valid_indices):
264
- if i > 0 and i % 1000 == 0:
265
- print(f" ⏳ [{sym}] Processing... {i}/{total_signals}", flush=True)
266
-
267
  ts_val = int(current_time.timestamp() * 1000)
268
-
269
- # Find Entry Index
270
  idx_1m = np.searchsorted(fast_1m['timestamp'], ts_val)
271
 
272
- # Safety Check
273
  if idx_1m < 500 or idx_1m >= len(fast_1m['close']) - 245: continue
274
 
275
- # Determine Indices
276
  idx_1h = map_1m_to_1h[idx_1m]
277
- idx_5m = map_1m_to_5m[idx_1m]
278
  idx_15m = map_1m_to_15m[idx_1m]
279
- idx_4h = np.searchsorted(numpy_htf['4h']['timestamp'], ts_val) # Do this once per signal (rare)
280
  if idx_4h >= len(numpy_htf['4h']['close']): idx_4h = len(numpy_htf['4h']['close']) - 1
281
 
282
  # === Oracle (Single Call) ===
@@ -311,138 +368,73 @@ class HeavyDutyBacktester:
311
  sniper_score = np.mean(s_preds)
312
  except: pass
313
 
314
- # === RISK SIMULATION (MATRIX BATCH) ===
315
- # We construct a matrix of 240 rows (4 hours) at once and predict ONCE.
316
- # This replaces the minute-by-minute loop.
317
-
318
- future_len = 240
319
  start_idx = idx_1m + 1
320
- end_idx = start_idx + future_len
321
-
322
- # Slices for 1m data
323
- sl_close = fast_1m['close'][start_idx:end_idx]
324
- sl_ts = fast_1m['timestamp'][start_idx:end_idx]
325
-
326
- entry_price = fast_1m['close'][idx_1m]
327
 
328
- # Mapped Indices for HTF slices
329
- sl_map_5m = map_1m_to_5m[start_idx:end_idx]
330
- sl_map_15m = map_1m_to_15m[start_idx:end_idx]
331
-
332
- max_hydra_crash = 0.0; hydra_crash_time = 0
333
  max_legacy_v2 = 0.0; legacy_panic_time = 0
334
-
335
- # --- A. Hydra Batch ---
336
- if hydra_models:
337
- sl_atr = fast_1m['atr'][start_idx:end_idx]
338
- sl_rsi_1m = fast_1m['rsi'][start_idx:end_idx]
339
- sl_bb = fast_1m['bb_width'][start_idx:end_idx]
340
- sl_vol = fast_1m['rel_vol'][start_idx:end_idx]
 
 
 
 
 
 
341
 
342
- # HTF Lookups (Using integer array indexing - Fast)
343
- sl_rsi_5m = numpy_htf['5m']['rsi'][sl_map_5m]
344
- sl_rsi_15m = numpy_htf['15m']['rsi'][sl_map_15m]
345
 
346
- # Calc Features
347
  sl_dist = 1.5 * sl_atr
348
  sl_dist = np.where(sl_dist > 0, sl_dist, entry_price * 0.015)
349
 
350
- # PnL & Max PnL
351
  sl_pnl = sl_close - entry_price
352
  sl_norm_pnl = sl_pnl / sl_dist
353
 
354
- # Max PnL needs cumulative max (rolling max from start of trade)
355
  sl_cum_max = np.maximum.accumulate(sl_close)
356
- # Correction: cum max of trade needs to start from entry price
357
  sl_cum_max = np.maximum(sl_cum_max, entry_price)
358
  sl_max_pnl_r = (sl_cum_max - entry_price) / sl_dist
359
 
360
  sl_atr_pct = sl_atr / sl_close
361
- sl_time = np.arange(1, future_len + 1)
362
 
363
- # Stack Matrix: (240, N_Features)
364
- # Feature Order Must Match hydra_cols
365
- # Map cols manually to speed up
366
 
367
- # Create dict of vectors
368
- feat_vecs = {
369
- 'rsi_1m': sl_rsi_1m, 'rsi_5m': sl_rsi_5m, 'rsi_15m': sl_rsi_15m,
370
- 'bb_width': sl_bb, 'rel_vol': sl_vol,
371
- 'dist_ema20_1h': np.zeros(future_len),
372
- 'atr_pct': sl_atr_pct, 'norm_pnl_r': sl_norm_pnl, 'max_pnl_r': sl_max_pnl_r,
373
- 'dist_tp_atr': np.zeros(future_len), 'dist_sl_atr': np.zeros(future_len),
374
- 'time_in_trade': sl_time,
375
- 'entry_type': np.zeros(future_len), 'oracle_conf': np.full(future_len, oracle_conf),
376
- 'l2_score': np.full(future_len, 0.7), 'target_class': np.full(future_len, 3.0)
377
- }
378
 
379
- # Stack
380
- X_hydra = np.column_stack([feat_vecs.get(c, np.zeros(future_len)) for c in hydra_cols])
 
 
 
 
 
 
 
 
 
 
 
 
 
381
 
382
  try:
383
- # ONE PREDICTION FOR 240 ROWS
384
  probs_crash = hydra_models['crash'].predict_proba(X_hydra)[:, 1]
385
-
386
- # Find Max
387
  max_hydra_crash = np.max(probs_crash)
388
-
389
- # Find Time
390
  crash_indices = np.where(probs_crash > 0.6)[0]
391
  if len(crash_indices) > 0:
392
- hydra_crash_time = int(sl_ts[crash_indices[0]])
393
  except: pass
394
 
395
- # --- B. Legacy V2 Batch ---
396
- if legacy_v2:
397
- # 1m Feats
398
- l_log = fast_1m['log_ret'][start_idx:end_idx]
399
- l_rsi = fast_1m['rsi'][start_idx:end_idx] / 100.0
400
- l_fib = fast_1m['fib_pos'][start_idx:end_idx]
401
- l_vol = fast_1m['volatility'][start_idx:end_idx]
402
-
403
- # 5m Feats (Mapped)
404
- l5_log = numpy_htf['5m']['log_ret'][sl_map_5m]
405
- l5_rsi = numpy_htf['5m']['rsi'][sl_map_5m] / 100.0
406
- l5_fib = numpy_htf['5m']['fib_pos'][sl_map_5m]
407
- l5_trd = numpy_htf['5m']['trend_slope'][sl_map_5m]
408
-
409
- # 15m Feats (Mapped)
410
- l15_log = numpy_htf['15m']['log_ret'][sl_map_15m]
411
- l15_rsi = numpy_htf['15m']['rsi'][sl_map_15m] / 100.0
412
- l15_fib618 = numpy_htf['15m']['dist_fib618'][sl_map_15m]
413
- l15_trd = numpy_htf['15m']['trend_slope'][sl_map_15m]
414
-
415
- # Lags (Pre-calculated in _calculate_indicators_vectorized)
416
- # We just pull them from fast_1m
417
- lag_cols = []
418
- for lag in [1, 2, 3, 5, 10, 20]:
419
- lag_cols.append(fast_1m[f'log_ret_lag_{lag}'][start_idx:end_idx])
420
- lag_cols.append(fast_1m[f'rsi_lag_{lag}'][start_idx:end_idx])
421
- lag_cols.append(fast_1m[f'fib_pos_lag_{lag}'][start_idx:end_idx])
422
- lag_cols.append(fast_1m[f'volatility_lag_{lag}'][start_idx:end_idx])
423
-
424
- # Stack All
425
- X_v2 = np.column_stack([
426
- l_log, l_rsi, l_fib, l_vol,
427
- l5_log, l5_rsi, l5_fib, l5_trd,
428
- l15_log, l15_rsi, l15_fib618, l15_trd,
429
- *lag_cols
430
- ])
431
-
432
- try:
433
- # PREDICT BATCH
434
- dm_v2 = xgb.DMatrix(X_v2)
435
- preds_v2 = legacy_v2.predict(dm_v2)
436
- # Handle Multiclass
437
- probs_v2 = preds_v2[:, 2] if len(preds_v2.shape) > 1 else preds_v2
438
-
439
- max_legacy_v2 = np.max(probs_v2)
440
- panic_idx = np.where(probs_v2 > 0.8)[0]
441
- if len(panic_idx) > 0 and legacy_panic_time == 0:
442
- legacy_panic_time = int(sl_ts[panic_idx[0]])
443
- except: pass
444
-
445
- # --- Store Result ---
446
  ai_results.append({
447
  'timestamp': ts_val, 'symbol': sym, 'close': entry_price,
448
  'real_titan': 0.6,
@@ -459,11 +451,11 @@ class HeavyDutyBacktester:
459
  dt = time.time() - t0
460
  if ai_results:
461
  pd.DataFrame(ai_results).to_pickle(scores_file)
462
- print(f" ✅ [{sym}] Batch-Processed {len(ai_results)} signals in {dt:.2f} seconds.", flush=True)
463
  else:
464
  print(f" ⚠️ [{sym}] No valid signals. Time: {dt:.2f}s", flush=True)
465
 
466
- del frames, fast_1m, numpy_htf
467
  gc.collect()
468
 
469
  # ==============================================================
@@ -508,16 +500,13 @@ class HeavyDutyBacktester:
508
  for ts, group in grouped_by_time:
509
  active = list(wallet["positions"].keys())
510
  current_prices = {row['symbol']: row['close'] for _, row in group.iterrows()}
511
-
512
  for sym in active:
513
  if sym in current_prices:
514
  curr = current_prices[sym]
515
  pos = wallet["positions"][sym]
516
-
517
  h_risk = pos.get('risk_hydra_crash', 0)
518
  h_time = pos.get('time_hydra_crash', 0)
519
  is_crash = (h_risk > hydra_thresh) and (h_time > 0) and (ts >= h_time)
520
-
521
  pnl = (curr - pos['entry']) / pos['entry']
522
  if is_crash or pnl > 0.04 or pnl < -0.02:
523
  wallet['balance'] += pos['size'] * (1 + pnl - (fees_pct*2))
@@ -569,7 +558,6 @@ class HeavyDutyBacktester:
569
  'config': config, 'final_balance': final_bal, 'net_profit': net_profit,
570
  'total_trades': total_t, 'win_count': win_count, 'loss_count': loss_count,
571
  'win_rate': win_rate, 'max_single_win': max_win, 'max_single_loss': max_loss,
572
- 'max_win_streak': max_win_streak, 'max_loss_streak': max_loss_streak,
573
  'max_drawdown': max_drawdown * 100
574
  })
575
 
@@ -577,11 +565,9 @@ class HeavyDutyBacktester:
577
 
578
  async def run_optimization(self, target_regime="RANGE"):
579
  await self.generate_truth_data()
580
-
581
  oracle_range = [0.5, 0.6, 0.7]
582
  sniper_range = [0.4, 0.5, 0.6]
583
  hydra_range = [0.75, 0.85, 0.95]
584
-
585
  combinations = []
586
  for o, s, h in itertools.product(oracle_range, sniper_range, hydra_range):
587
  combinations.append({
@@ -595,7 +581,6 @@ class HeavyDutyBacktester:
595
 
596
  print(f"\n🧩 [Phase 2] Optimizing {len(combinations)} Configs (Full Stack) for {target_regime}...")
597
  best_res = self._worker_optimize(combinations, current_period_files, self.INITIAL_CAPITAL, self.TRADING_FEES, self.MAX_SLOTS)
598
-
599
  if not best_res: return None, None
600
  best = sorted(best_res, key=lambda x: x['final_balance'], reverse=True)[0]
601
 
@@ -618,6 +603,7 @@ class HeavyDutyBacktester:
618
  print("-" * 60)
619
  print(f" ⚙️ Oracle={best['config']['oracle_thresh']} | Sniper={best['config']['sniper_thresh']} | Hydra={best['config']['hydra_thresh']}")
620
  print("="*60)
 
621
 
622
  async def run_strategic_optimization_task():
623
  print("\n🧪 [STRATEGIC BACKTEST] Full Stack Mode...")
 
1
  # ============================================================
2
+ # 🧪 backtest_engine.py (V113.0 - GEM-Architect: Global Pre-Inference)
3
  # ============================================================
4
 
5
  import asyncio
 
55
  self.force_end_date = None
56
 
57
  if not os.path.exists(CACHE_DIR): os.makedirs(CACHE_DIR)
58
+ print(f"🧪 [Backtest V113.0] Pre-Inference Velocity Mode (Target: 60s).")
59
 
60
  def set_date_range(self, start_str, end_str):
61
  self.force_start_date = start_str
 
104
  return unique_candles
105
 
106
  # ==============================================================
107
+ # 🏎️ VECTORIZED INDICATORS
108
  # ==============================================================
109
  def _calculate_indicators_vectorized(self, df, timeframe='1m'):
110
  df['close'] = df['close'].astype(float)
 
118
  df['ema50'] = ta.ema(df['close'], length=50)
119
  df['atr'] = ta.atr(df['high'], df['low'], df['close'], length=14)
120
 
 
121
  if timeframe == '1m':
122
  sma20 = df['close'].rolling(20).mean()
123
  std20 = df['close'].rolling(20).std()
 
125
  df['vol_ma50'] = df['volume'].rolling(50).mean()
126
  df['rel_vol'] = df['volume'] / (df['vol_ma50'] + 1e-9)
127
 
 
128
  df['slope'] = ta.slope(df['close'], length=7)
129
  vol_mean = df['volume'].rolling(20).mean()
130
  vol_std = df['volume'].rolling(20).std()
131
  df['vol_z'] = (df['volume'] - vol_mean) / (vol_std + 1e-9)
132
  df['atr_pct'] = df['atr'] / df['close']
133
 
 
134
  if timeframe == '1m':
135
  df['ret'] = df['close'].pct_change()
136
  df['dollar_vol'] = df['close'] * df['volume']
 
156
  s = df['volume'].rolling(500).std()
157
  df['vol_zscore_50'] = ((df['volume'] - r) / s).fillna(0)
158
 
 
159
  df['log_ret'] = np.log(df['close'] / df['close'].shift(1))
160
  roll_max = df['high'].rolling(50).max()
161
  roll_min = df['low'].rolling(50).min()
 
169
  df['ema200'] = ta.ema(df['close'], length=200)
170
  df['dist_ema200'] = (df['close'] - df['ema200']) / df['close']
171
 
172
+ # Lags for V2
173
  if timeframe == '1m':
174
  for lag in [1, 2, 3, 5, 10, 20]:
175
  df[f'log_ret_lag_{lag}'] = df['log_ret'].shift(lag).fillna(0)
 
181
  return df
182
 
183
  # ==============================================================
184
+ # 🧠 CPU PROCESSING (PRE-INFERENCE OPTIMIZED)
185
  # ==============================================================
186
  async def _process_data_in_memory(self, sym, candles, start_ms, end_ms):
187
  safe_sym = sym.replace('/', '_')
 
192
  print(f" 📂 [{sym}] Data Exists -> Skipping.")
193
  return
194
 
195
+ print(f" ⚙️ [CPU] Analyzing {sym} (Global Pre-Inference)...", flush=True)
196
  t0 = time.time()
197
 
198
  df_1m = pd.DataFrame(candles, columns=['timestamp', 'open', 'high', 'low', 'close', 'volume'])
 
203
  frames = {}
204
  agg_dict = {'open': 'first', 'high': 'max', 'low': 'min', 'close': 'last', 'volume': 'sum'}
205
 
206
+ # 1. Calc 1m
207
  frames['1m'] = self._calculate_indicators_vectorized(df_1m.copy(), timeframe='1m')
208
  frames['1m']['timestamp'] = frames['1m'].index.floor('1min').astype(np.int64) // 10**6
209
  fast_1m = {col: frames['1m'][col].values for col in frames['1m'].columns}
 
217
  frames[tf_str] = resampled
218
  numpy_htf[tf_str] = {col: resampled[col].values for col in resampled.columns}
219
 
220
+ # 3. Global Index Maps
 
 
221
  map_1m_to_1h = np.searchsorted(numpy_htf['1h']['timestamp'], fast_1m['timestamp'])
222
  map_1m_to_5m = np.searchsorted(numpy_htf['5m']['timestamp'], fast_1m['timestamp'])
223
  map_1m_to_15m = np.searchsorted(numpy_htf['15m']['timestamp'], fast_1m['timestamp'])
224
 
225
+ # Clip
226
+ max_idx_1h = len(numpy_htf['1h']['timestamp']) - 1
227
+ max_idx_5m = len(numpy_htf['5m']['timestamp']) - 1
228
+ max_idx_15m = len(numpy_htf['15m']['timestamp']) - 1
229
+
230
+ map_1m_to_1h = np.clip(map_1m_to_1h, 0, max_idx_1h)
231
+ map_1m_to_5m = np.clip(map_1m_to_5m, 0, max_idx_5m)
232
+ map_1m_to_15m = np.clip(map_1m_to_15m, 0, max_idx_15m)
233
+
234
+ # 4. Load Models
235
+ hydra_models = getattr(self.proc.guardian_hydra, 'models', {}) if self.proc.guardian_hydra else {}
236
+ hydra_cols = getattr(self.proc.guardian_hydra, 'feature_cols', []) if self.proc.guardian_hydra else []
237
+ legacy_v2 = getattr(self.proc.guardian_legacy, 'model_v2', None)
238
+
239
+ # 5. 🔥 PRE-CALCULATE LEGACY V2 (GLOBAL) 🔥
240
+ # V2 depends only on structure, not entry price. We can predict for ALL rows at once.
241
+ global_v2_probs = np.zeros(len(fast_1m['close']))
242
+
243
+ if legacy_v2:
244
+ print(f" 🚀 Pre-calculating Legacy V2 for entire history...", flush=True)
245
+ try:
246
+ # 1m Feats
247
+ l_log = fast_1m['log_ret']
248
+ l_rsi = fast_1m['rsi'] / 100.0
249
+ l_fib = fast_1m['fib_pos']
250
+ l_vol = fast_1m['volatility']
251
+
252
+ # HTF Feats Mapped to 1m
253
+ l5_log = numpy_htf['5m']['log_ret'][map_1m_to_5m]
254
+ l5_rsi = numpy_htf['5m']['rsi'][map_1m_to_5m] / 100.0
255
+ l5_fib = numpy_htf['5m']['fib_pos'][map_1m_to_5m]
256
+ l5_trd = numpy_htf['5m']['trend_slope'][map_1m_to_5m]
257
+
258
+ l15_log = numpy_htf['15m']['log_ret'][map_1m_to_15m]
259
+ l15_rsi = numpy_htf['15m']['rsi'][map_1m_to_15m] / 100.0
260
+ l15_fib618 = numpy_htf['15m']['dist_fib618'][map_1m_to_15m]
261
+ l15_trd = numpy_htf['15m']['trend_slope'][map_1m_to_15m]
262
+
263
+ # Lags
264
+ lag_cols = []
265
+ for lag in [1, 2, 3, 5, 10, 20]:
266
+ lag_cols.append(fast_1m[f'log_ret_lag_{lag}'])
267
+ lag_cols.append(fast_1m[f'rsi_lag_{lag}'])
268
+ lag_cols.append(fast_1m[f'fib_pos_lag_{lag}'])
269
+ lag_cols.append(fast_1m[f'volatility_lag_{lag}'])
270
+
271
+ # Huge Matrix
272
+ X_GLOBAL_V2 = np.column_stack([
273
+ l_log, l_rsi, l_fib, l_vol,
274
+ l5_log, l5_rsi, l5_fib, l5_trd,
275
+ l15_log, l15_rsi, l15_fib618, l15_trd,
276
+ *lag_cols
277
+ ])
278
+
279
+ # Predict All in One Go
280
+ dm_glob = xgb.DMatrix(X_GLOBAL_V2)
281
+ preds_glob = legacy_v2.predict(dm_glob)
282
+ global_v2_probs = preds_glob[:, 2] if len(preds_glob.shape) > 1 else preds_glob
283
+
284
+ except Exception as e: print(f"V2 Error: {e}")
285
+
286
+ # 6. 🔥 PRE-ASSEMBLE HYDRA STATIC (GLOBAL) 🔥
287
+ # Hydra needs PnL (dynamic), but 90% features are static.
288
+ global_hydra_static = None
289
+ if hydra_models:
290
+ print(f" 🚀 Pre-assembling Hydra features...", flush=True)
291
+ try:
292
+ # Map columns that don't depend on PnL
293
+ h_rsi_1m = fast_1m['rsi']
294
+ h_rsi_5m = numpy_htf['5m']['rsi'][map_1m_to_5m]
295
+ h_rsi_15m = numpy_htf['15m']['rsi'][map_1m_to_15m]
296
+ h_bb = fast_1m['bb_width']
297
+ h_vol = fast_1m['rel_vol']
298
+ h_atr = fast_1m['atr']
299
+ h_close = fast_1m['close']
300
+
301
+ # We store these separate to combine inside loop efficiently
302
+ # [rsi1, rsi5, rsi15, bb, vol, atr, close]
303
+ global_hydra_static = np.column_stack([h_rsi_1m, h_rsi_5m, h_rsi_15m, h_bb, h_vol, h_atr, h_close])
304
+ except: pass
305
 
306
+ # 7. Candidate Filtering
307
  df_1h = frames['1h'].reindex(frames['5m'].index, method='ffill')
308
  df_5m = frames['5m'].copy()
309
  is_valid = (df_1h['rsi'] <= 70)
 
312
  final_valid_indices = [t for t in valid_indices if t >= start_dt]
313
 
314
  total_signals = len(final_valid_indices)
315
+ print(f" 🎯 Candidates: {total_signals}. Running Models...", flush=True)
316
 
 
 
 
 
 
 
 
 
317
  oracle_dir_model = getattr(self.proc.oracle, 'model_direction', None)
318
  oracle_cols = getattr(self.proc.oracle, 'feature_cols', [])
319
  sniper_models = getattr(self.proc.sniper, 'models', [])
320
  sniper_cols = getattr(self.proc.sniper, 'feature_names', [])
321
 
322
+ ai_results = []
323
+
324
+ # Pre-allocate Hydra time vector (0 to 240)
325
+ time_vec = np.arange(1, 241)
326
 
327
+ # --- MAIN LOOP (Optimized Lookups) ---
328
  for i, current_time in enumerate(final_valid_indices):
 
 
 
329
  ts_val = int(current_time.timestamp() * 1000)
 
 
330
  idx_1m = np.searchsorted(fast_1m['timestamp'], ts_val)
331
 
 
332
  if idx_1m < 500 or idx_1m >= len(fast_1m['close']) - 245: continue
333
 
 
334
  idx_1h = map_1m_to_1h[idx_1m]
 
335
  idx_15m = map_1m_to_15m[idx_1m]
336
+ idx_4h = np.searchsorted(numpy_htf['4h']['timestamp'], ts_val)
337
  if idx_4h >= len(numpy_htf['4h']['close']): idx_4h = len(numpy_htf['4h']['close']) - 1
338
 
339
  # === Oracle (Single Call) ===
 
368
  sniper_score = np.mean(s_preds)
369
  except: pass
370
 
371
+ # === RISK SIMULATION (ULTRA FAST) ===
 
 
 
 
372
  start_idx = idx_1m + 1
373
+ end_idx = start_idx + 240
 
 
 
 
 
 
374
 
375
+ # 1. LEGACY V2 (Instant Lookup)
 
 
 
 
376
  max_legacy_v2 = 0.0; legacy_panic_time = 0
377
+ if legacy_v2:
378
+ # Just slice the pre-calculated array!
379
+ probs_slice = global_v2_probs[start_idx:end_idx]
380
+ max_legacy_v2 = np.max(probs_slice)
381
+ panic_indices = np.where(probs_slice > 0.8)[0]
382
+ if len(panic_indices) > 0:
383
+ legacy_panic_time = int(fast_1m['timestamp'][start_idx + panic_indices[0]])
384
+
385
+ # 2. HYDRA (Semi-Vectorized)
386
+ max_hydra_crash = 0.0; hydra_crash_time = 0
387
+ if hydra_models and global_hydra_static is not None:
388
+ # Slice Static Feats
389
+ sl_static = global_hydra_static[start_idx:end_idx] # [rsi1, rsi5, rsi15, bb, vol, atr, close]
390
 
391
+ entry_price = fast_1m['close'][idx_1m]
392
+ sl_close = sl_static[:, 6]
393
+ sl_atr = sl_static[:, 5]
394
 
395
+ # Calc Dynamic Feats
396
  sl_dist = 1.5 * sl_atr
397
  sl_dist = np.where(sl_dist > 0, sl_dist, entry_price * 0.015)
398
 
 
399
  sl_pnl = sl_close - entry_price
400
  sl_norm_pnl = sl_pnl / sl_dist
401
 
 
402
  sl_cum_max = np.maximum.accumulate(sl_close)
 
403
  sl_cum_max = np.maximum(sl_cum_max, entry_price)
404
  sl_max_pnl_r = (sl_cum_max - entry_price) / sl_dist
405
 
406
  sl_atr_pct = sl_atr / sl_close
 
407
 
408
+ # Map to Hydra Cols Order (Hardcoded for max speed)
409
+ # Cols: rsi_1m, rsi_5m, rsi_15m, bb_width, rel_vol, dist_ema20_1h, atr_pct, norm_pnl_r, max_pnl_r, dists..., time, entry, oracle, l2, target
 
410
 
411
+ # Re-assemble only what is needed
412
+ # (Static 0-4) + (Zeros) + (Dynamic) + (Constants)
 
 
 
 
 
 
 
 
 
413
 
414
+ # Create arrays for constants
415
+ zeros = np.zeros(240)
416
+ oracle_arr = np.full(240, oracle_conf)
417
+ l2_arr = np.full(240, 0.7)
418
+ target_arr = np.full(240, 3.0)
419
+
420
+ X_hydra = np.column_stack([
421
+ sl_static[:, 0], sl_static[:, 1], sl_static[:, 2], # RSIs
422
+ sl_static[:, 3], sl_static[:, 4], # BB, Vol
423
+ zeros, # dist_ema
424
+ sl_atr_pct, sl_norm_pnl, sl_max_pnl_r,
425
+ zeros, zeros, # dists
426
+ time_vec, # time
427
+ zeros, oracle_arr, l2_arr, target_arr
428
+ ])
429
 
430
  try:
 
431
  probs_crash = hydra_models['crash'].predict_proba(X_hydra)[:, 1]
 
 
432
  max_hydra_crash = np.max(probs_crash)
 
 
433
  crash_indices = np.where(probs_crash > 0.6)[0]
434
  if len(crash_indices) > 0:
435
+ hydra_crash_time = int(fast_1m['timestamp'][start_idx + crash_indices[0]])
436
  except: pass
437
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
438
  ai_results.append({
439
  'timestamp': ts_val, 'symbol': sym, 'close': entry_price,
440
  'real_titan': 0.6,
 
451
  dt = time.time() - t0
452
  if ai_results:
453
  pd.DataFrame(ai_results).to_pickle(scores_file)
454
+ print(f" ✅ [{sym}] Completed {len(ai_results)} signals in {dt:.2f} seconds.", flush=True)
455
  else:
456
  print(f" ⚠️ [{sym}] No valid signals. Time: {dt:.2f}s", flush=True)
457
 
458
+ del frames, fast_1m, numpy_htf, global_v2_probs, global_hydra_static
459
  gc.collect()
460
 
461
  # ==============================================================
 
500
  for ts, group in grouped_by_time:
501
  active = list(wallet["positions"].keys())
502
  current_prices = {row['symbol']: row['close'] for _, row in group.iterrows()}
 
503
  for sym in active:
504
  if sym in current_prices:
505
  curr = current_prices[sym]
506
  pos = wallet["positions"][sym]
 
507
  h_risk = pos.get('risk_hydra_crash', 0)
508
  h_time = pos.get('time_hydra_crash', 0)
509
  is_crash = (h_risk > hydra_thresh) and (h_time > 0) and (ts >= h_time)
 
510
  pnl = (curr - pos['entry']) / pos['entry']
511
  if is_crash or pnl > 0.04 or pnl < -0.02:
512
  wallet['balance'] += pos['size'] * (1 + pnl - (fees_pct*2))
 
558
  'config': config, 'final_balance': final_bal, 'net_profit': net_profit,
559
  'total_trades': total_t, 'win_count': win_count, 'loss_count': loss_count,
560
  'win_rate': win_rate, 'max_single_win': max_win, 'max_single_loss': max_loss,
 
561
  'max_drawdown': max_drawdown * 100
562
  })
563
 
 
565
 
566
  async def run_optimization(self, target_regime="RANGE"):
567
  await self.generate_truth_data()
 
568
  oracle_range = [0.5, 0.6, 0.7]
569
  sniper_range = [0.4, 0.5, 0.6]
570
  hydra_range = [0.75, 0.85, 0.95]
 
571
  combinations = []
572
  for o, s, h in itertools.product(oracle_range, sniper_range, hydra_range):
573
  combinations.append({
 
581
 
582
  print(f"\n🧩 [Phase 2] Optimizing {len(combinations)} Configs (Full Stack) for {target_regime}...")
583
  best_res = self._worker_optimize(combinations, current_period_files, self.INITIAL_CAPITAL, self.TRADING_FEES, self.MAX_SLOTS)
 
584
  if not best_res: return None, None
585
  best = sorted(best_res, key=lambda x: x['final_balance'], reverse=True)[0]
586
 
 
603
  print("-" * 60)
604
  print(f" ⚙️ Oracle={best['config']['oracle_thresh']} | Sniper={best['config']['sniper_thresh']} | Hydra={best['config']['hydra_thresh']}")
605
  print("="*60)
606
+ return best['config'], best
607
 
608
  async def run_strategic_optimization_task():
609
  print("\n🧪 [STRATEGIC BACKTEST] Full Stack Mode...")