initial: import DOLPHIN baseline 2026-04-21 from dolphinng5_predict working tree
Includes core prod + GREEN/BLUE subsystems: - prod/ (BLUE harness, configs, scripts, docs) - nautilus_dolphin/ (GREEN Nautilus-native impl + dvae/ preserved) - adaptive_exit/ (AEM engine + models/bucket_assignments.pkl) - Observability/ (EsoF advisor, TUI, dashboards) - external_factors/ (EsoF producer) - mc_forewarning_qlabs_fork/ (MC regime/envelope) Excludes runtime caches, logs, backups, and reproducible artifacts per .gitignore.
This commit is contained in:
hjnormey
307
nautilus_dolphin/test_1m_calibration_sweep.py
Executable file
307
nautilus_dolphin/test_1m_calibration_sweep.py
Executable file
@@ -0,0 +1,307 @@
|
||||
"""1m Klines System Calibration — max_hold_bars × abs_max_lev sweep.
|
||||
|
||||
Problem: 1m system has DD=31.69% (elevated). Cause: 2-hour max hold (120 bars at 1min)
|
||||
× avg leverage 2.57x — long holds amplify adverse periods.
|
||||
|
||||
Sweep 1: max_hold_bars in [30, 45, 60, 90, 120] (with abs_max_lev=5.0)
|
||||
Sweep 2: abs_max_lev in [3.0, 4.0, 5.0, 6.0] (with best max_hold from sweep 1)
|
||||
Combined grid: 5 × 4 = 20 runs on 795-day klines window.
|
||||
|
||||
Full engine stack identical to klines_2y_experiment (ACBv6 + OB + MC-Forewarner + EsoF neutral).
|
||||
Thresholds: vel_div_threshold=-0.50, vel_div_extreme=-1.25 (klines-adapted).
|
||||
Saves: run_logs/1m_calib_{TS}.csv + .json
|
||||
"""
|
||||
import sys, time, json, csv
|
||||
sys.stdout.reconfigure(encoding='utf-8', errors='replace')
|
||||
from pathlib import Path
|
||||
from datetime import datetime
|
||||
import numpy as np
|
||||
import pandas as pd
|
||||
|
||||
sys.path.insert(0, str(Path(__file__).parent))
|
||||
|
||||
print("Compiling numba kernels...")
|
||||
t0c = time.time()
|
||||
from nautilus_dolphin.nautilus.alpha_asset_selector import compute_irp_nb, compute_ars_nb, rank_assets_irp_nb
|
||||
from nautilus_dolphin.nautilus.alpha_bet_sizer import compute_sizing_nb
|
||||
from nautilus_dolphin.nautilus.alpha_signal_generator import check_dc_nb
|
||||
from nautilus_dolphin.nautilus.ob_features import (
|
||||
OBFeatureEngine, compute_imbalance_nb, compute_depth_1pct_nb,
|
||||
compute_depth_quality_nb, compute_fill_probability_nb, compute_spread_proxy_nb,
|
||||
compute_depth_asymmetry_nb, compute_imbalance_persistence_nb,
|
||||
compute_withdrawal_velocity_nb, compute_market_agreement_nb, compute_cascade_signal_nb,
|
||||
)
|
||||
from nautilus_dolphin.nautilus.ob_provider import MockOBProvider
|
||||
_p = np.array([1.0, 2.0, 3.0], dtype=np.float64)
|
||||
compute_irp_nb(_p, -1); compute_ars_nb(1.0, 0.5, 0.01)
|
||||
rank_assets_irp_nb(np.ones((10, 2), dtype=np.float64), 8, -1, 5, 500.0, 20, 0.20)
|
||||
compute_sizing_nb(-0.55, -0.50, -1.25, 3.0, 0.5, 5.0, 0.20, True, True, 0.0,
|
||||
np.zeros(4, dtype=np.int64), np.zeros(4, dtype=np.int64),
|
||||
np.zeros(5, dtype=np.float64), 0, -1, 0.01, 0.04)
|
||||
check_dc_nb(_p, 3, 1, 0.75)
|
||||
_b = np.array([100.0, 200.0, 300.0, 400.0, 500.0], dtype=np.float64)
|
||||
_a = np.array([110.0, 190.0, 310.0, 390.0, 510.0], dtype=np.float64)
|
||||
compute_imbalance_nb(_b, _a); compute_depth_1pct_nb(_b, _a)
|
||||
compute_depth_quality_nb(210.0, 200.0); compute_fill_probability_nb(1.0)
|
||||
compute_spread_proxy_nb(_b, _a); compute_depth_asymmetry_nb(_b, _a)
|
||||
compute_imbalance_persistence_nb(np.array([0.1, -0.1], dtype=np.float64), 2)
|
||||
compute_withdrawal_velocity_nb(np.array([100.0, 110.0], dtype=np.float64), 1)
|
||||
compute_market_agreement_nb(np.array([0.1, -0.05], dtype=np.float64), 2)
|
||||
compute_cascade_signal_nb(np.array([-0.05, -0.15], dtype=np.float64), 2, -0.10)
|
||||
print(f" JIT: {time.time()-t0c:.1f}s")
|
||||
|
||||
from nautilus_dolphin.nautilus.esf_alpha_orchestrator import NDAlphaEngine
|
||||
from nautilus_dolphin.nautilus.adaptive_circuit_breaker import AdaptiveCircuitBreaker
|
||||
from mc.mc_ml import DolphinForewarner
|
||||
|
||||
VBT_DIR = Path(r"C:\Users\Lenovo\Documents\- DOLPHIN NG HD HCM TSF Predict\vbt_cache_klines")
|
||||
DATE_START = '2024-01-01'
|
||||
DATE_END = '2026-03-05'
|
||||
META_COLS = {'timestamp', 'scan_number', 'v50_lambda_max_velocity', 'v150_lambda_max_velocity',
|
||||
'v300_lambda_max_velocity', 'v750_lambda_max_velocity', 'vel_div',
|
||||
'instability_50', 'instability_150'}
|
||||
MC_MODELS_DIR = str(Path(r"C:\Users\Lenovo\Documents\- DOLPHIN NG HD HCM TSF Predict\nautilus_dolphin\mc_results\models"))
|
||||
MC_BASE_CFG = {
|
||||
'trial_id': 0, 'vel_div_threshold': -0.02, 'vel_div_extreme': -0.05,
|
||||
'use_direction_confirm': True, 'dc_lookback_bars': 7,
|
||||
'dc_min_magnitude_bps': 0.75, 'dc_skip_contradicts': True,
|
||||
'dc_leverage_boost': 1.00, 'dc_leverage_reduce': 0.50,
|
||||
'vd_trend_lookback': 10, 'min_leverage': 0.50, 'max_leverage': 5.00,
|
||||
'leverage_convexity': 3.00, 'fraction': 0.20, 'use_alpha_layers': True,
|
||||
'use_dynamic_leverage': True, 'fixed_tp_pct': 0.0099, 'stop_pct': 1.00,
|
||||
'max_hold_bars': 120, 'use_sp_fees': True, 'use_sp_slippage': True,
|
||||
'sp_maker_entry_rate': 0.62, 'sp_maker_exit_rate': 0.50,
|
||||
'use_ob_edge': True, 'ob_edge_bps': 5.00, 'ob_confirm_rate': 0.40,
|
||||
'ob_imbalance_bias': -0.09, 'ob_depth_scale': 1.00,
|
||||
'use_asset_selection': True, 'min_irp_alignment': 0.45, 'lookback': 100,
|
||||
'acb_beta_high': 0.80, 'acb_beta_low': 0.20, 'acb_w750_threshold_pct': 60,
|
||||
}
|
||||
BASE_ENGINE_KWARGS = dict(
|
||||
initial_capital=25000.0,
|
||||
vel_div_threshold=-0.50, vel_div_extreme=-1.25, # klines-adapted
|
||||
min_leverage=0.5, max_leverage=5.0, leverage_convexity=3.0,
|
||||
fraction=0.20, fixed_tp_pct=0.0099, stop_pct=1.0,
|
||||
use_direction_confirm=True, dc_lookback_bars=7, dc_min_magnitude_bps=0.75,
|
||||
dc_skip_contradicts=True, dc_leverage_boost=1.0, dc_leverage_reduce=0.5,
|
||||
use_asset_selection=True, min_irp_alignment=0.45,
|
||||
use_sp_fees=True, use_sp_slippage=True,
|
||||
sp_maker_entry_rate=0.62, sp_maker_exit_rate=0.50,
|
||||
use_ob_edge=True, ob_edge_bps=5.0, ob_confirm_rate=0.40,
|
||||
lookback=100, use_alpha_layers=True, use_dynamic_leverage=True, seed=42,
|
||||
)
|
||||
OB_ASSETS = ["BTCUSDT", "ETHUSDT", "BNBUSDT", "SOLUSDT"]
|
||||
|
||||
# ── Shared state ───────────────────────────────────────────────────────────────
|
||||
print("\nLoading MC-Forewarner...")
|
||||
forewarner = DolphinForewarner(models_dir=MC_MODELS_DIR)
|
||||
|
||||
parquet_files = sorted(
|
||||
p for p in VBT_DIR.glob("*.parquet")
|
||||
if 'catalog' not in str(p) and DATE_START <= p.stem <= DATE_END
|
||||
)
|
||||
date_strings = [pf.stem for pf in parquet_files]
|
||||
print(f"Dates: {len(parquet_files)} ({date_strings[0]} to {date_strings[-1]})")
|
||||
|
||||
# Vol calibration (first 5 dates for klines)
|
||||
all_vols = []
|
||||
for pf in parquet_files[:5]:
|
||||
df = pd.read_parquet(pf)
|
||||
if 'BTCUSDT' not in df.columns: continue
|
||||
pr = df['BTCUSDT'].values
|
||||
for i in range(60, len(pr)):
|
||||
seg = pr[max(0,i-50):i]
|
||||
if len(seg)<10: continue
|
||||
v = float(np.std(np.diff(seg)/seg[:-1]))
|
||||
if v > 0: all_vols.append(v)
|
||||
vol_p60 = float(np.percentile(all_vols, 60)) if all_vols else 1e-4
|
||||
print(f"Vol p60 (klines): {vol_p60:.6f}")
|
||||
|
||||
print(f"Pre-loading {len(parquet_files)} parquets...")
|
||||
t_load = time.time()
|
||||
pq_data = {}
|
||||
for i, pf in enumerate(parquet_files):
|
||||
df = pd.read_parquet(pf)
|
||||
ac = [c for c in df.columns if c not in META_COLS]
|
||||
bp = df['BTCUSDT'].values if 'BTCUSDT' in df.columns else None
|
||||
dv = np.full(len(df), np.nan)
|
||||
if bp is not None:
|
||||
for j in range(50, len(bp)):
|
||||
seg = bp[max(0,j-50):j]
|
||||
if len(seg)<10: continue
|
||||
dv[j] = float(np.std(np.diff(seg)/seg[:-1]))
|
||||
pq_data[pf.stem] = (df, ac, dv)
|
||||
if (i+1) % 200 == 0:
|
||||
print(f" {i+1}/{len(parquet_files)} loaded...")
|
||||
print(f" Done in {time.time()-t_load:.1f}s")
|
||||
|
||||
# ACB w750 from klines parquet
|
||||
acb_master = AdaptiveCircuitBreaker()
|
||||
acb_master.preload_w750(date_strings) # returns all-zero for klines (no NPZ files)
|
||||
for ds, (df, _, _) in pq_data.items():
|
||||
if 'v750_lambda_max_velocity' in df.columns:
|
||||
v750 = df['v750_lambda_max_velocity'].dropna()
|
||||
if len(v750) > 0:
|
||||
acb_master._w750_vel_cache[ds] = float(v750.median())
|
||||
_w750 = [v for v in acb_master._w750_vel_cache.values() if v != 0.0]
|
||||
if _w750:
|
||||
acb_master._w750_threshold = float(np.percentile(_w750, acb_master.config.W750_THRESHOLD_PCT))
|
||||
print(f"ACB w750 p60 (klines): {acb_master._w750_threshold:.6f}")
|
||||
|
||||
_mock_ob = MockOBProvider(
|
||||
imbalance_bias=-0.09, depth_scale=1.0, assets=OB_ASSETS,
|
||||
imbalance_biases={"BTCUSDT": -0.086, "ETHUSDT": -0.092,
|
||||
"BNBUSDT": +0.05, "SOLUSDT": +0.05},
|
||||
)
|
||||
ob_eng = OBFeatureEngine(_mock_ob)
|
||||
ob_eng.preload_date("mock", OB_ASSETS)
|
||||
|
||||
# ── Sweep grid ─────────────────────────────────────────────────────────────────
|
||||
MAX_HOLD_SWEEP = [30, 45, 60, 90, 120] # bars (=minutes at 1m timescale)
|
||||
ABS_MAX_LEV_SWEEP = [3.0, 4.0, 5.0, 6.0]
|
||||
|
||||
print(f"\n{'='*75}")
|
||||
print(f" 1m CALIBRATION SWEEP: max_hold × abs_max_lev")
|
||||
print(f" max_hold_bars: {MAX_HOLD_SWEEP}")
|
||||
print(f" abs_max_lev: {ABS_MAX_LEV_SWEEP}")
|
||||
print(f" Total runs: {len(MAX_HOLD_SWEEP) * len(ABS_MAX_LEV_SWEEP)}")
|
||||
print(f" Baseline: max_hold=120 abs_max_lev=5.0 (795-day: ROI=+172.34% DD=31.69%)")
|
||||
print(f"{'='*75}")
|
||||
|
||||
def run_klines(max_hold, abs_max_lev):
|
||||
kw = dict(BASE_ENGINE_KWARGS, max_hold_bars=max_hold, abs_max_leverage=abs_max_lev)
|
||||
engine = NDAlphaEngine(**kw)
|
||||
engine.set_ob_engine(ob_eng)
|
||||
engine.set_acb(acb_master)
|
||||
engine.set_mc_forewarner(forewarner, MC_BASE_CFG)
|
||||
engine.set_esoteric_hazard_multiplier(0.0)
|
||||
|
||||
all_daily = []
|
||||
for ds in date_strings:
|
||||
df, acols, dvol = pq_data[ds]
|
||||
vol_ok = np.where(np.isfinite(dvol), dvol > vol_p60, False)
|
||||
r = engine.process_day(ds, df, acols, vol_regime_ok=vol_ok)
|
||||
all_daily.append({'pnl': r.get('pnl', 0.0), 'capital': r.get('capital', 25000.0),
|
||||
'trades': r.get('trades', 0)})
|
||||
|
||||
tr = engine.trade_history
|
||||
wins = [t for t in tr if t.pnl_absolute > 0]
|
||||
losses = [t for t in tr if t.pnl_absolute <= 0]
|
||||
gw = sum(t.pnl_absolute for t in wins)
|
||||
gl = abs(sum(t.pnl_absolute for t in losses))
|
||||
roi = (engine.capital - 25000) / 25000 * 100
|
||||
pf = gw / gl if gl > 0 else 999.0
|
||||
wr = len(wins) / len(tr) * 100 if tr else 0.0
|
||||
pnls = np.array([s['pnl'] for s in all_daily])
|
||||
sharpe = float(pnls.mean() / pnls.std() * np.sqrt(252)) if pnls.std() > 0 else 0.0
|
||||
caps = [s['capital'] for s in all_daily]
|
||||
peak = 25000.0; max_dd = 0.0
|
||||
for c in caps:
|
||||
if c > peak: peak = c
|
||||
dd = (peak - c) / peak * 100
|
||||
if dd > max_dd: max_dd = dd
|
||||
tp_rate = engine.tp_exits / len(tr) * 100 if tr else 0.0
|
||||
avg_lev = float(np.mean([t.leverage for t in tr])) if tr else 0.0
|
||||
avg_bars = float(np.mean([t.bars_held for t in tr])) if tr else 0.0
|
||||
h1 = [r for r in all_daily if pq_data and date_strings[all_daily.index(r)] < '2025-01-01']
|
||||
h2 = [r for r in all_daily if pq_data and date_strings[all_daily.index(r)] >= '2025-01-01']
|
||||
# simpler h1/h2 split by index
|
||||
mid = len(all_daily) // 2
|
||||
h1_roi = sum(s['pnl'] for s in all_daily[:mid]) / 25000 * 100
|
||||
h2_roi = sum(s['pnl'] for s in all_daily[mid:]) / 25000 * 100
|
||||
h2h1 = h2_roi / h1_roi if h1_roi != 0 else float('nan')
|
||||
|
||||
return {
|
||||
'max_hold_bars': max_hold, 'abs_max_lev': abs_max_lev,
|
||||
'max_hold_min': max_hold, # 1min bars → minutes
|
||||
'roi': roi, 'pf': pf, 'dd': max_dd, 'sharpe': sharpe, 'wr': wr,
|
||||
'n_trades': len(tr), 'tp_rate_pct': tp_rate, 'avg_lev': avg_lev,
|
||||
'avg_bars_held': avg_bars, 'h1_roi': h1_roi, 'h2_roi': h2_roi, 'h2h1': h2h1,
|
||||
}
|
||||
|
||||
results = []
|
||||
t_sweep_start = time.time()
|
||||
run_n = 0
|
||||
|
||||
for max_hold in MAX_HOLD_SWEEP:
|
||||
for abs_max_lev in ABS_MAX_LEV_SWEEP:
|
||||
run_n += 1
|
||||
t0r = time.time()
|
||||
baseline_mark = " ← BASELINE" if (max_hold == 120 and abs_max_lev == 5.0) else ""
|
||||
print(f"\n[{run_n}/{len(MAX_HOLD_SWEEP)*len(ABS_MAX_LEV_SWEEP)}] "
|
||||
f"max_hold={max_hold}min abs_max_lev={abs_max_lev}x{baseline_mark}")
|
||||
|
||||
row = run_klines(max_hold, abs_max_lev)
|
||||
results.append(row)
|
||||
elapsed_r = time.time() - t0r
|
||||
|
||||
print(f" ROI={row['roi']:+.2f}% PF={row['pf']:.4f} DD={row['dd']:.2f}% "
|
||||
f"Sh={row['sharpe']:.3f} WR={row['wr']:.1f}% T={row['n_trades']} "
|
||||
f"TP%={row['tp_rate_pct']:.1f}% AvgLev={row['avg_lev']:.2f}x "
|
||||
f"AvgBars={row['avg_bars_held']:.1f} [{elapsed_r:.0f}s]")
|
||||
|
||||
total_elapsed = time.time() - t_sweep_start
|
||||
|
||||
# ── Analysis ────────────────────────────────────────────────────────────────────
|
||||
baseline = next(r for r in results if r['max_hold_bars'] == 120 and r['abs_max_lev'] == 5.0)
|
||||
best_roi = max(results, key=lambda r: r['roi'])
|
||||
best_sharpe = max(results, key=lambda r: r['sharpe'])
|
||||
best_dd = min(results, key=lambda r: r['dd'])
|
||||
# Best risk-adjusted: highest ROI with DD < 25%
|
||||
viable = [r for r in results if r['dd'] <= 25.0]
|
||||
best_viable = max(viable, key=lambda r: r['roi']) if viable else best_roi
|
||||
|
||||
print(f"\n{'='*75}")
|
||||
print(f" 1m CALIBRATION SWEEP COMPLETE ({total_elapsed/60:.1f} min)")
|
||||
print(f"{'='*75}")
|
||||
print(f" Baseline (hold=120m lev=5x): ROI={baseline['roi']:+.2f}% PF={baseline['pf']:.4f} "
|
||||
f"DD={baseline['dd']:.2f}% Sh={baseline['sharpe']:.3f}")
|
||||
print(f" Best ROI: hold={best_roi['max_hold_bars']}m lev={best_roi['abs_max_lev']}x "
|
||||
f"→ ROI={best_roi['roi']:+.2f}% DD={best_roi['dd']:.2f}%")
|
||||
print(f" Best Sharpe: hold={best_sharpe['max_hold_bars']}m lev={best_sharpe['abs_max_lev']}x "
|
||||
f"→ Sh={best_sharpe['sharpe']:.3f} DD={best_sharpe['dd']:.2f}%")
|
||||
print(f" Min DD: hold={best_dd['max_hold_bars']}m lev={best_dd['abs_max_lev']}x "
|
||||
f"→ DD={best_dd['dd']:.2f}% ROI={best_dd['roi']:+.2f}%")
|
||||
print(f" Best viable (DD≤25%): hold={best_viable['max_hold_bars']}m "
|
||||
f"lev={best_viable['abs_max_lev']}x → ROI={best_viable['roi']:+.2f}% "
|
||||
f"DD={best_viable['dd']:.2f}% Sh={best_viable['sharpe']:.3f}")
|
||||
|
||||
print(f"\n Grid summary (ROI | DD):")
|
||||
print(f" {'':>12}", end='')
|
||||
for lev in ABS_MAX_LEV_SWEEP:
|
||||
print(f" lev={lev:.0f}x ", end='')
|
||||
print()
|
||||
for mh in MAX_HOLD_SWEEP:
|
||||
print(f" hold={mh:3d}min ", end='')
|
||||
for lev in ABS_MAX_LEV_SWEEP:
|
||||
row = next(r for r in results if r['max_hold_bars'] == mh and r['abs_max_lev'] == lev)
|
||||
mk = '*' if (row == best_viable) else (' ' if (mh != 120 or lev != 5.0) else 'B')
|
||||
print(f" {row['roi']:+6.1f}%/{row['dd']:4.1f}%{mk} ", end='')
|
||||
print()
|
||||
|
||||
# ── Save ────────────────────────────────────────────────────────────────────────
|
||||
ts = datetime.now().strftime('%Y%m%d_%H%M%S')
|
||||
run_dir = Path(__file__).parent / 'run_logs'
|
||||
run_dir.mkdir(exist_ok=True)
|
||||
|
||||
with open(run_dir / f'1m_calib_{ts}.csv', 'w', newline='') as f:
|
||||
w = csv.DictWriter(f, fieldnames=list(results[0].keys()))
|
||||
w.writeheader(); w.writerows(results)
|
||||
|
||||
summary = {
|
||||
'experiment': '1m_klines_calibration_sweep',
|
||||
'date_range': f'{DATE_START}_to_{DATE_END}',
|
||||
'max_hold_sweep': MAX_HOLD_SWEEP,
|
||||
'abs_max_lev_sweep': ABS_MAX_LEV_SWEEP,
|
||||
'baseline': baseline,
|
||||
'best_roi': best_roi,
|
||||
'best_sharpe': best_sharpe,
|
||||
'best_dd_reduction': best_dd,
|
||||
'best_viable_dd25': best_viable,
|
||||
'elapsed_s': total_elapsed,
|
||||
'run_ts': ts,
|
||||
'all_results': results,
|
||||
}
|
||||
with open(run_dir / f'1m_calib_{ts}.json', 'w') as f:
|
||||
json.dump(summary, f, indent=2)
|
||||
|
||||
print(f"\nSaved: run_logs/1m_calib_{ts}.csv + .json")
|
||||
Reference in New Issue
Block a user