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
205
prod/reconstruct_181.py
Executable file
205
prod/reconstruct_181.py
Executable file
@@ -0,0 +1,205 @@
|
||||
"""
|
||||
reconstruct_181.py — Faithful reconstruction of the original certification path.
|
||||
|
||||
The E2E test that certified 181.81% used:
|
||||
- exp_shared.load_data() that returned pq_data (all data pre-loaded, float64)
|
||||
- eng.set_esoteric_hazard_multiplier(0.0) [ceiling=6.0 at time of certification]
|
||||
- vol_ok = np.where(np.isfinite(dvol), dvol > d['vol_p60'], False) [STATIC threshold]
|
||||
- NO per-day OB clearing (OB set up once)
|
||||
- funding_btc (not fund_dbt_btc) for ACB signals
|
||||
|
||||
This script tests multiple sub-hypotheses to find what changed.
|
||||
|
||||
Tests:
|
||||
R1: Exact old path (pq_data style, ceiling patched to 6.0, static vol, no OB clear)
|
||||
R2: Same but ceiling=10.0 (agent's change) — to isolate ceiling effect in old path
|
||||
R3: No hazard call + old path (to isolate hazard effect)
|
||||
R4: Old path but funding_btc forced for ACB (isolate ACB change)
|
||||
|
||||
All use static vol_p60 and np.isfinite (the certification condition).
|
||||
"""
|
||||
import sys, time, gc, math
|
||||
from pathlib import Path
|
||||
import numpy as np
|
||||
import pandas as pd
|
||||
|
||||
ROOT = Path(r"C:\Users\Lenovo\Documents\- DOLPHIN NG HD HCM TSF Predict")
|
||||
sys.path.insert(0, str(ROOT / 'nautilus_dolphin'))
|
||||
sys.path.insert(0, str(ROOT / 'nautilus_dolphin' / 'dvae'))
|
||||
|
||||
import exp_shared
|
||||
from nautilus_dolphin.nautilus.proxy_boost_engine import create_d_liq_engine
|
||||
from nautilus_dolphin.nautilus.adaptive_circuit_breaker import AdaptiveCircuitBreaker
|
||||
|
||||
print("Ensuring JIT...")
|
||||
exp_shared.ensure_jit()
|
||||
|
||||
# ── Pre-load all data (OLD style — what pq_data contained) ───────────────────
|
||||
print("\nPre-loading all parquet data (old style, float64)...")
|
||||
VBT_DIR = exp_shared.VBT_DIR
|
||||
parquet_files = sorted(VBT_DIR.glob("*.parquet"))
|
||||
parquet_files = [p for p in parquet_files if 'catalog' not in str(p)]
|
||||
date_strings = [p.stem for p in parquet_files]
|
||||
|
||||
all_vols = []
|
||||
for pf in parquet_files[:3]:
|
||||
tmp = pd.read_parquet(pf)
|
||||
if 'BTCUSDT' in tmp.columns:
|
||||
bp = tmp['BTCUSDT'].values
|
||||
diffs = np.diff(bp) / bp[:-1]
|
||||
for i in range(50, len(diffs)):
|
||||
all_vols.append(np.std(diffs[i-50:i]))
|
||||
del tmp
|
||||
vol_p60_static = float(np.percentile(all_vols, 60)) if all_vols else 0.0002
|
||||
print(f" Static vol_p60: {vol_p60_static:.8f}")
|
||||
|
||||
# Load ALL data into pq_data (the old approach)
|
||||
print(" Loading all 56 parquet files...")
|
||||
pq_data = {}
|
||||
for pf in parquet_files:
|
||||
ds = pf.stem
|
||||
df = pd.read_parquet(pf) # float64, no casting
|
||||
acols = [c for c in df.columns if c not in exp_shared.META_COLS]
|
||||
bp = df['BTCUSDT'].values if 'BTCUSDT' in df.columns else None
|
||||
dvol = np.full(len(df), np.nan)
|
||||
if bp is not None:
|
||||
diffs = np.zeros(len(bp), dtype=np.float64)
|
||||
diffs[1:] = np.diff(bp) / bp[:-1]
|
||||
for j in range(50, len(bp)):
|
||||
dvol[j] = np.std(diffs[j-50:j])
|
||||
pq_data[ds] = (df, acols, dvol)
|
||||
|
||||
# OB setup (once only, like old path)
|
||||
from nautilus_dolphin.nautilus.ob_features import OBFeatureEngine
|
||||
from nautilus_dolphin.nautilus.ob_provider import MockOBProvider
|
||||
OB_ASSETS = ["BTCUSDT", "ETHUSDT"]
|
||||
_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)
|
||||
|
||||
print(f" Loaded {len(pq_data)} days. Ready.\n")
|
||||
|
||||
|
||||
def run_reconstruction(label, ceiling_lev, use_hazard_call, use_static_vol):
|
||||
print(f"\n{'='*65}")
|
||||
print(f" {label}")
|
||||
print(f" ceiling={ceiling_lev} hazard_call={use_hazard_call} static_vol={use_static_vol}")
|
||||
print(f"{'='*65}")
|
||||
|
||||
kw = exp_shared.ENGINE_KWARGS.copy()
|
||||
kw.update({'sp_maker_entry_rate': 1.0, 'sp_maker_exit_rate': 1.0, 'use_sp_slippage': False})
|
||||
|
||||
acb = AdaptiveCircuitBreaker()
|
||||
acb.preload_w750(date_strings)
|
||||
|
||||
eng = create_d_liq_engine(**kw)
|
||||
eng.set_ob_engine(ob_eng)
|
||||
eng.set_acb(acb)
|
||||
|
||||
if use_hazard_call:
|
||||
# Patch ceiling_lev before calling
|
||||
import nautilus_dolphin.nautilus.esf_alpha_orchestrator as orch_mod
|
||||
original_fn = orch_mod.NDAlphaEngine.set_esoteric_hazard_multiplier
|
||||
|
||||
def patched_hazard(self, hazard_score):
|
||||
floor_lev = 3.0
|
||||
c_lev = ceiling_lev # capture
|
||||
range_lev = c_lev - floor_lev
|
||||
target_lev = c_lev - (hazard_score * range_lev)
|
||||
import math
|
||||
step = range_lev / 8.0
|
||||
stepped_lev = math.ceil(target_lev / step) * step
|
||||
self.base_max_leverage = max(floor_lev, min(c_lev, stepped_lev))
|
||||
self.bet_sizer.max_leverage = self.base_max_leverage
|
||||
if hasattr(self, '_day_mc_status'):
|
||||
if self._day_mc_status == 'RED':
|
||||
self.bet_sizer.max_leverage = self.base_max_leverage * 0.8
|
||||
else:
|
||||
self.bet_sizer.max_leverage = self.base_max_leverage
|
||||
|
||||
import types
|
||||
eng.set_esoteric_hazard_multiplier = types.MethodType(patched_hazard, eng)
|
||||
eng.set_esoteric_hazard_multiplier(0.0)
|
||||
|
||||
base_after = getattr(eng, 'base_max_leverage', None)
|
||||
print(f" After hazard call: base_max={base_after} abs_max={getattr(eng,'abs_max_leverage',None)}")
|
||||
else:
|
||||
print(f" No hazard call: base_max={getattr(eng,'base_max_leverage',None)} abs_max={getattr(eng,'abs_max_leverage',None)}")
|
||||
|
||||
daily_caps, daily_pnls = [], []
|
||||
t0 = time.time()
|
||||
|
||||
for pf in parquet_files:
|
||||
ds = pf.stem
|
||||
df, acols, dvol = pq_data[ds] # OLD style: pre-loaded float64 data
|
||||
|
||||
cap_before = eng.capital
|
||||
vol_ok = np.where(np.isfinite(dvol), dvol > vol_p60_static, False) # OLD condition
|
||||
eng.process_day(ds, df, acols, vol_regime_ok=vol_ok)
|
||||
daily_caps.append(eng.capital)
|
||||
daily_pnls.append(eng.capital - cap_before)
|
||||
|
||||
tr = eng.trade_history
|
||||
n = len(tr)
|
||||
roi = (eng.capital - 25000.0) / 25000.0 * 100.0
|
||||
|
||||
peak_cap, max_dd = 25000.0, 0.0
|
||||
for cap in daily_caps:
|
||||
peak_cap = max(peak_cap, cap)
|
||||
max_dd = max(max_dd, (peak_cap - cap) / peak_cap * 100.0)
|
||||
|
||||
elapsed = time.time() - t0
|
||||
|
||||
def _abs(t): return t.pnl_absolute if hasattr(t,'pnl_absolute') else t.pnl_pct*250.
|
||||
if n > 0:
|
||||
wins = [t for t in tr if _abs(t) > 0]
|
||||
pf_val = sum(_abs(t) for t in wins) / max(abs(sum(_abs(t) for t in tr if _abs(t)<=0)), 1e-9)
|
||||
dr = np.array([p/25000.*100. for p in daily_pnls])
|
||||
sharpe = float(dr.mean()/(dr.std()+1e-9)*math.sqrt(365))
|
||||
calmar = roi / max(max_dd, 0.01)
|
||||
else:
|
||||
pf_val, sharpe, calmar = 0, 0, 0
|
||||
|
||||
print(f"\n ROI: {roi:+.2f}% T={n} DD={max_dd:.2f}% PF={pf_val:.3f} Calmar={calmar:.2f} ({elapsed:.0f}s)")
|
||||
|
||||
gold_roi, gold_t = 181.81, 2155
|
||||
roi_ok = abs(roi - gold_roi) <= 2.0
|
||||
t_ok = abs(n - gold_t) <= 5
|
||||
print(f" vs GOLD (+181.81%, T=2155): ROI diff={roi-gold_roi:+.2f}pp {'✓' if roi_ok else '✗'} T diff={n-gold_t:+d} {'✓' if t_ok else '✗'}")
|
||||
|
||||
return {'label': label, 'roi': roi, 'trades': n, 'dd': max_dd, 'pf': pf_val, 'calmar': calmar, 'elapsed': elapsed}
|
||||
|
||||
|
||||
if __name__ == '__main__':
|
||||
results = []
|
||||
|
||||
# R1: Exact certification conditions — ceiling=6.0 (original), hazard call, static vol, np.isfinite
|
||||
results.append(run_reconstruction(
|
||||
"R1: Cert conditions (ceiling=6.0, hazard, static vol)",
|
||||
ceiling_lev=6.0, use_hazard_call=True, use_static_vol=True,
|
||||
))
|
||||
|
||||
# R2: Ceiling=10.0 (agent's change), hazard call, static vol — isolate ceiling effect
|
||||
results.append(run_reconstruction(
|
||||
"R2: Agent ceiling=10.0, hazard, static vol",
|
||||
ceiling_lev=10.0, use_hazard_call=True, use_static_vol=True,
|
||||
))
|
||||
|
||||
# R3: No hazard call, static vol — baseline replicate style (should ≈ 111%)
|
||||
results.append(run_reconstruction(
|
||||
"R3: No hazard call, static vol (replicate style)",
|
||||
ceiling_lev=6.0, use_hazard_call=False, use_static_vol=True,
|
||||
))
|
||||
|
||||
print(f"\n{'='*65}")
|
||||
print(" RECONSTRUCTION SUMMARY")
|
||||
print(f"{'='*65}")
|
||||
print(f" {'Config':<45} {'ROI':>8} {'T':>6} {'DD':>7} {'PF':>6}")
|
||||
print(f" {'-'*75}")
|
||||
print(f" {'GOLD STANDARD':.<45} {'+181.81%':>8} {'2155':>6} {'17.65%':>7} {'---':>6}")
|
||||
for r in results:
|
||||
print(f" {r['label']:<45} {r['roi']:>+7.2f}% {r['trades']:>6} {r['dd']:>6.2f}% {r['pf']:>6.3f}")
|
||||
print(f" {'Current replicate (no hazard, float64, static)':.<45} {'~+111%':>8} {'~1959':>6} {'~16.9%':>7}")
|
||||
Reference in New Issue
Block a user