421 lines
20 KiB
Python
421 lines
20 KiB
Python
|
"""vel_div Adverse-Turn Exit Calibration
|
|||
|
|
========================================
|
|||
|
|
Runs the champion engine (vd_enabled=False), then post-hoc reconstructs
|
|||
|
|
the vel_div trajectory for every trade and sweeps thresholds.
|
|||
|
|
|
|||
|
|
For each (threshold, debounce_bars) pair:
|
|||
|
|
- % of trades that would trigger early exit
|
|||
|
|
- Sensitivity: % of true losers captured (want HIGH)
|
|||
|
|
- Specificity: % of true winners avoided (want HIGH)
|
|||
|
|
- Separation score: P(loser|triggered) - P(loser|not triggered) (want HIGH)
|
|||
|
|
- avg bars saved for triggered losers (want HIGH)
|
|||
|
|
- avg pnl of triggered vs not-triggered
|
|||
|
|
|
|||
|
|
Output:
|
|||
|
|
run_logs/vd_calib_<ts>.csv — per-trade data
|
|||
|
|
run_logs/vd_sweep_<ts>.csv — full threshold sweep
|
|||
|
|
Printed: top-10 configs by separation score
|
|||
|
|
"""
|
|||
|
|
import sys, time, json, csv
|
|||
|
|
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(-0.03, -0.02, -0.05, 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
|
|||
|
|
|
|||
|
|
# ── Config (identical to main test) ────────────────────────────────────────────
|
|||
|
|
VBT_DIR = Path(r"C:\Users\Lenovo\Documents\- DOLPHIN NG HD HCM TSF Predict\vbt_cache")
|
|||
|
|
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'}
|
|||
|
|
ENGINE_KWARGS = dict(
|
|||
|
|
initial_capital=25000.0, vel_div_threshold=-0.02, vel_div_extreme=-0.05,
|
|||
|
|
min_leverage=0.5, max_leverage=5.0, leverage_convexity=3.0,
|
|||
|
|
fraction=0.20, fixed_tp_pct=0.0099, stop_pct=1.0, max_hold_bars=120,
|
|||
|
|
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,
|
|||
|
|
)
|
|||
|
|
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': -0.020, 'vel_div_extreme': -0.050,
|
|||
|
|
'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,
|
|||
|
|
}
|
|||
|
|
|
|||
|
|
# ── Setup ───────────────────────────────────────────────────────────────────────
|
|||
|
|
print("\nLoading MC-Forewarner...")
|
|||
|
|
forewarner = DolphinForewarner(models_dir=MC_MODELS_DIR)
|
|||
|
|
print(" OK")
|
|||
|
|
|
|||
|
|
parquet_files = sorted(VBT_DIR.glob("*.parquet"))
|
|||
|
|
parquet_files = [p for p in parquet_files if 'catalog' not in str(p)]
|
|||
|
|
|
|||
|
|
acb = AdaptiveCircuitBreaker()
|
|||
|
|
date_strings = [pf.stem for pf in parquet_files]
|
|||
|
|
acb.preload_w750(date_strings)
|
|||
|
|
|
|||
|
|
all_vols = []
|
|||
|
|
for pf in parquet_files[:2]:
|
|||
|
|
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))
|
|||
|
|
|
|||
|
|
pq_data = {}
|
|||
|
|
for pf in 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 i in range(50, len(bp)):
|
|||
|
|
seg = bp[max(0, i-50):i]
|
|||
|
|
if len(seg) < 10: continue
|
|||
|
|
dv[i] = float(np.std(np.diff(seg)/seg[:-1]))
|
|||
|
|
pq_data[pf.stem] = (df, ac, dv)
|
|||
|
|
|
|||
|
|
OB_ASSETS = ["BTCUSDT", "ETHUSDT", "BNBUSDT", "SOLUSDT"]
|
|||
|
|
_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)
|
|||
|
|
|
|||
|
|
# ── Build global_bar_idx → (date_str, row_idx) mapping ─────────────────────────
|
|||
|
|
# Mirrors the _global_bar_idx increment logic in process_day():
|
|||
|
|
# every row increments the counter, including NaN-vd rows.
|
|||
|
|
print("\nBuilding global bar index map...")
|
|||
|
|
gbar_map = {} # global_bar_idx → {'date': str, 'row': int, 'vel_div': float}
|
|||
|
|
g = 0
|
|||
|
|
for pf in parquet_files:
|
|||
|
|
ds = pf.stem
|
|||
|
|
df, _, _ = pq_data[ds]
|
|||
|
|
for ri in range(len(df)):
|
|||
|
|
row = df.iloc[ri]
|
|||
|
|
vd_val = row.get('vel_div')
|
|||
|
|
vd_f = float(vd_val) if (vd_val is not None and np.isfinite(float(vd_val))) else np.nan
|
|||
|
|
gbar_map[g] = {'date': ds, 'row': ri, 'vel_div': vd_f}
|
|||
|
|
g += 1
|
|||
|
|
print(f" Total global bars: {g}")
|
|||
|
|
|
|||
|
|
# ── Engine run (champion exits, vd_enabled=False) ──────────────────────────────
|
|||
|
|
print(f"\n=== Champion run (vd_enabled=False) ===")
|
|||
|
|
t0 = time.time()
|
|||
|
|
engine = NDAlphaEngine(**ENGINE_KWARGS)
|
|||
|
|
engine.set_ob_engine(ob_eng)
|
|||
|
|
engine.set_acb(acb)
|
|||
|
|
engine.set_mc_forewarner(forewarner, MC_BASE_CFG)
|
|||
|
|
engine.set_esoteric_hazard_multiplier(0.0)
|
|||
|
|
|
|||
|
|
for pf in parquet_files:
|
|||
|
|
ds = pf.stem
|
|||
|
|
df, acols, dvol = pq_data[ds]
|
|||
|
|
vol_ok = np.where(np.isfinite(dvol), dvol > vol_p60, False)
|
|||
|
|
engine.process_day(ds, df, acols, vol_regime_ok=vol_ok)
|
|||
|
|
|
|||
|
|
tr = engine.trade_history
|
|||
|
|
elapsed = time.time() - t0
|
|||
|
|
print(f" Done: {elapsed:.1f}s | {len(tr)} trades | ROI={(engine.capital-25000)/25000*100:+.2f}%")
|
|||
|
|
|
|||
|
|
# ── Post-hoc: reconstruct vel_div sequence per trade ───────────────────────────
|
|||
|
|
print("\nReconstructing vel_div trajectories per trade...")
|
|||
|
|
|
|||
|
|
trade_records = []
|
|||
|
|
for t in tr:
|
|||
|
|
# Collect vel_div values from entry_bar to exit_bar (inclusive)
|
|||
|
|
vd_seq = []
|
|||
|
|
for b in range(t.entry_bar, t.exit_bar + 1):
|
|||
|
|
info = gbar_map.get(b)
|
|||
|
|
if info and not np.isnan(info['vel_div']):
|
|||
|
|
vd_seq.append(info['vel_div'])
|
|||
|
|
|
|||
|
|
if not vd_seq:
|
|||
|
|
continue
|
|||
|
|
|
|||
|
|
vd_arr = np.array(vd_seq)
|
|||
|
|
is_winner = t.pnl_absolute > 0
|
|||
|
|
|
|||
|
|
# entry_vel_div not stored in NDTradeRecord — look it up from gbar_map
|
|||
|
|
entry_vd_info = gbar_map.get(t.entry_bar)
|
|||
|
|
entry_vd = entry_vd_info['vel_div'] if entry_vd_info else np.nan
|
|||
|
|
|
|||
|
|
# Basic vel_div stats
|
|||
|
|
rec = {
|
|||
|
|
'trade_id': t.trade_id,
|
|||
|
|
'asset': t.asset,
|
|||
|
|
'exit_reason': t.exit_reason,
|
|||
|
|
'bars_held': t.bars_held,
|
|||
|
|
'pnl_pct': t.pnl_pct * 100,
|
|||
|
|
'pnl_abs': t.pnl_absolute,
|
|||
|
|
'leverage': t.leverage,
|
|||
|
|
'is_winner': is_winner,
|
|||
|
|
'entry_vd': entry_vd,
|
|||
|
|
'exit_vd': vd_arr[-1],
|
|||
|
|
'max_vd': float(np.max(vd_arr)),
|
|||
|
|
'min_vd': float(np.min(vd_arr)),
|
|||
|
|
'mean_vd': float(np.mean(vd_arr)),
|
|||
|
|
'pct_bars_positive_vd': float(np.mean(vd_arr > 0) * 100),
|
|||
|
|
'n_vd_bars': len(vd_arr),
|
|||
|
|
'vd_seq': vd_seq,
|
|||
|
|
}
|
|||
|
|
trade_records.append(rec)
|
|||
|
|
|
|||
|
|
print(f" {len(trade_records)} trades with vel_div data")
|
|||
|
|
|
|||
|
|
# ── Threshold sweep ──────────────────────────────────────────────────────────────
|
|||
|
|
# For each (threshold, debounce): compute separation metrics.
|
|||
|
|
THRESHOLDS = [0.0, 0.005, 0.01, 0.015, 0.02, 0.025, 0.03, 0.04, 0.05, 0.07, 0.10, 0.15, 0.20]
|
|||
|
|
DEBOUNCE = [1, 2, 3, 5, 10, 20]
|
|||
|
|
|
|||
|
|
print(f"\nRunning threshold sweep ({len(THRESHOLDS)} thresholds × {len(DEBOUNCE)} debounces)...")
|
|||
|
|
|
|||
|
|
def first_sustain(vd_seq, threshold, n_consec):
|
|||
|
|
"""Return index of first bar where vd_seq has n_consec consecutive values > threshold.
|
|||
|
|
Returns -1 if never sustained."""
|
|||
|
|
count = 0
|
|||
|
|
for i, v in enumerate(vd_seq):
|
|||
|
|
if v > threshold:
|
|||
|
|
count += 1
|
|||
|
|
if count >= n_consec:
|
|||
|
|
return i - n_consec + 1 # bar where sustained crossing started
|
|||
|
|
else:
|
|||
|
|
count = 0
|
|||
|
|
return -1
|
|||
|
|
|
|||
|
|
sweep_rows = []
|
|||
|
|
for thr in THRESHOLDS:
|
|||
|
|
for deb in DEBOUNCE:
|
|||
|
|
triggered = [] # (is_winner, bars_held, pnl_pct, trigger_bar)
|
|||
|
|
not_triggered = []
|
|||
|
|
|
|||
|
|
for rec in trade_records:
|
|||
|
|
tb = first_sustain(rec['vd_seq'], thr, deb)
|
|||
|
|
if tb >= 0:
|
|||
|
|
triggered.append({
|
|||
|
|
'is_winner': rec['is_winner'],
|
|||
|
|
'bars_held': rec['bars_held'],
|
|||
|
|
'pnl_pct': rec['pnl_pct'],
|
|||
|
|
'trigger_bar': tb,
|
|||
|
|
'bars_saved': rec['bars_held'] - tb,
|
|||
|
|
})
|
|||
|
|
else:
|
|||
|
|
not_triggered.append({
|
|||
|
|
'is_winner': rec['is_winner'],
|
|||
|
|
'pnl_pct': rec['pnl_pct'],
|
|||
|
|
})
|
|||
|
|
|
|||
|
|
n_tot = len(trade_records)
|
|||
|
|
n_trig = len(triggered)
|
|||
|
|
n_ntrig = len(not_triggered)
|
|||
|
|
if n_tot == 0: continue
|
|||
|
|
|
|||
|
|
# True losers / winners
|
|||
|
|
n_losers = sum(1 for r in trade_records if not r['is_winner'])
|
|||
|
|
n_winners = sum(1 for r in trade_records if r['is_winner'])
|
|||
|
|
|
|||
|
|
# Triggered stats
|
|||
|
|
trig_losers = sum(1 for r in triggered if not r['is_winner'])
|
|||
|
|
trig_winners = sum(1 for r in triggered if r['is_winner'])
|
|||
|
|
wr_trig = trig_winners / n_trig * 100 if n_trig > 0 else 0.0
|
|||
|
|
avg_pnl_trig = float(np.mean([r['pnl_pct'] for r in triggered])) if triggered else 0.0
|
|||
|
|
avg_bars_saved = float(np.mean([r['bars_saved'] for r in triggered if r['bars_saved'] > 0])) if triggered else 0.0
|
|||
|
|
|
|||
|
|
# Not-triggered stats
|
|||
|
|
ntrig_losers = sum(1 for r in not_triggered if not r['is_winner'])
|
|||
|
|
wr_ntrig = sum(1 for r in not_triggered if r['is_winner']) / n_ntrig * 100 if n_ntrig > 0 else 0.0
|
|||
|
|
avg_pnl_ntrig = float(np.mean([r['pnl_pct'] for r in not_triggered])) if not_triggered else 0.0
|
|||
|
|
|
|||
|
|
# Sensitivity: % of true losers that triggered (captured)
|
|||
|
|
sensitivity = trig_losers / n_losers * 100 if n_losers > 0 else 0.0
|
|||
|
|
# Specificity: % of true winners NOT triggered (avoided)
|
|||
|
|
specificity = (n_winners - trig_winners) / n_winners * 100 if n_winners > 0 else 0.0
|
|||
|
|
# Precision: % of triggered that are true losers
|
|||
|
|
precision = trig_losers / n_trig * 100 if n_trig > 0 else 0.0
|
|||
|
|
# Separation: how much better wr_ntrig is vs wr_trig (want high)
|
|||
|
|
separation = wr_ntrig - wr_trig
|
|||
|
|
|
|||
|
|
sweep_rows.append({
|
|||
|
|
'threshold': thr, 'debounce': deb,
|
|||
|
|
'pct_triggered': round(n_trig / n_tot * 100, 2),
|
|||
|
|
'sensitivity': round(sensitivity, 2),
|
|||
|
|
'specificity': round(specificity, 2),
|
|||
|
|
'precision': round(precision, 2),
|
|||
|
|
'separation': round(separation, 2),
|
|||
|
|
'wr_triggered': round(wr_trig, 2),
|
|||
|
|
'wr_not_triggered': round(wr_ntrig, 2),
|
|||
|
|
'avg_pnl_triggered': round(avg_pnl_trig, 4),
|
|||
|
|
'avg_pnl_not_triggered': round(avg_pnl_ntrig, 4),
|
|||
|
|
'avg_bars_saved': round(avg_bars_saved, 1),
|
|||
|
|
'n_triggered': n_trig,
|
|||
|
|
'n_not_triggered': n_ntrig,
|
|||
|
|
})
|
|||
|
|
|
|||
|
|
print(f" Sweep complete ({len(sweep_rows)} combos)")
|
|||
|
|
|
|||
|
|
# ── Vel_div distribution summary ────────────────────────────────────────────────
|
|||
|
|
winners = [r for r in trade_records if r['is_winner']]
|
|||
|
|
losers = [r for r in trade_records if not r['is_winner']]
|
|||
|
|
|
|||
|
|
def pct_positive_summary(records, label):
|
|||
|
|
if not records: return
|
|||
|
|
pcts = [r['pct_bars_positive_vd'] for r in records]
|
|||
|
|
max_vds = [r['max_vd'] for r in records]
|
|||
|
|
print(f" {label:8s} n={len(records):5d} "
|
|||
|
|
f"pct_bars_vd>0: mean={np.mean(pcts):5.1f}% p25={np.percentile(pcts,25):5.1f}% p75={np.percentile(pcts,75):5.1f}% "
|
|||
|
|
f"max_vd: mean={np.mean(max_vds):+.4f} p50={np.percentile(max_vds,50):+.4f} p90={np.percentile(max_vds,90):+.4f}")
|
|||
|
|
|
|||
|
|
print(f"\n{'─'*70}")
|
|||
|
|
print(f" VEL_DIV DISTRIBUTION DURING OPEN TRADES")
|
|||
|
|
print(f"{'─'*70}")
|
|||
|
|
pct_positive_summary(winners, "WINNERS")
|
|||
|
|
pct_positive_summary(losers, "LOSERS")
|
|||
|
|
print(f" Overall: % trades where vel_div EVER crosses 0: "
|
|||
|
|
f"{sum(1 for r in trade_records if r['max_vd'] > 0)/len(trade_records)*100:.1f}%")
|
|||
|
|
print(f" Overall: % trades where vel_div > +0.02 ever: "
|
|||
|
|
f"{sum(1 for r in trade_records if r['max_vd'] > 0.02)/len(trade_records)*100:.1f}%")
|
|||
|
|
print(f" Overall: % trades where vel_div > +0.05 ever: "
|
|||
|
|
f"{sum(1 for r in trade_records if r['max_vd'] > 0.05)/len(trade_records)*100:.1f}%")
|
|||
|
|
|
|||
|
|
# ASCII histogram of max_vd distribution (winners vs losers)
|
|||
|
|
def ascii_hist(values, lo, hi, bins=16, label=""):
|
|||
|
|
edges = np.linspace(lo, hi, bins+1)
|
|||
|
|
counts, _ = np.histogram(values, bins=edges)
|
|||
|
|
total = max(1, len(values))
|
|||
|
|
bar_max = max(1, max(counts))
|
|||
|
|
width = 20
|
|||
|
|
print(f"\n {label} (n={total})")
|
|||
|
|
for i in range(bins):
|
|||
|
|
bar = '█' * int(counts[i] / bar_max * width)
|
|||
|
|
pct = counts[i] / total * 100
|
|||
|
|
print(f" [{edges[i]:+.3f},{edges[i+1]:+.3f}) {bar:<{width}} {pct:4.1f}%")
|
|||
|
|
|
|||
|
|
ascii_hist([r['max_vd'] for r in winners], -0.15, 0.20, bins=14, label="WINNERS max vel_div during trade")
|
|||
|
|
ascii_hist([r['max_vd'] for r in losers], -0.15, 0.20, bins=14, label="LOSERS max vel_div during trade")
|
|||
|
|
|
|||
|
|
# ── Print sweep results: top by separation ──────────────────────────────────────
|
|||
|
|
print(f"\n{'═'*80}")
|
|||
|
|
print(f" THRESHOLD SWEEP — TOP 20 BY SEPARATION SCORE")
|
|||
|
|
print(f" (separation = WR_not_triggered - WR_triggered; higher = better loser/winner split)")
|
|||
|
|
print(f"{'─'*80}")
|
|||
|
|
print(f" {'thr':>7} {'deb':>4} {'%trig':>6} {'sens%':>6} {'spec%':>6} {'prec%':>6} {'sep':>6} "
|
|||
|
|
f"{'wr_t%':>6} {'wr_nt%':>7} {'bars_sv':>8} {'avg_pnl_t':>10}")
|
|||
|
|
print(f"{'─'*80}")
|
|||
|
|
|
|||
|
|
top20 = sorted(sweep_rows, key=lambda x: x['separation'], reverse=True)[:20]
|
|||
|
|
for r in top20:
|
|||
|
|
print(f" {r['threshold']:7.3f} {r['debounce']:4d} {r['pct_triggered']:6.1f} "
|
|||
|
|
f"{r['sensitivity']:6.1f} {r['specificity']:6.1f} {r['precision']:6.1f} "
|
|||
|
|
f"{r['separation']:6.2f} {r['wr_triggered']:6.1f} {r['wr_not_triggered']:7.1f} "
|
|||
|
|
f"{r['avg_bars_saved']:8.1f} {r['avg_pnl_triggered']:10.4f}")
|
|||
|
|
|
|||
|
|
# Also print the Pareto frontier: best separation at each pct_triggered level
|
|||
|
|
print(f"\n{'─'*80}")
|
|||
|
|
print(f" PARETO FRONTIER (best separation per coverage bracket)")
|
|||
|
|
print(f"{'─'*80}")
|
|||
|
|
brackets = [(0,5), (5,10), (10,20), (20,30), (30,50), (50,100)]
|
|||
|
|
for lo, hi in brackets:
|
|||
|
|
subset = [r for r in sweep_rows if lo <= r['pct_triggered'] < hi]
|
|||
|
|
if not subset: continue
|
|||
|
|
best = max(subset, key=lambda x: x['separation'])
|
|||
|
|
print(f" Coverage {lo:3d}-{hi:3d}%: thr={best['threshold']:.3f} deb={best['debounce']} "
|
|||
|
|
f"sep={best['separation']:.2f} sens={best['sensitivity']:.1f}% spec={best['specificity']:.1f}% "
|
|||
|
|
f"prec={best['precision']:.1f}% bars_saved={best['avg_bars_saved']:.1f}")
|
|||
|
|
|
|||
|
|
# ── Save logs ───────────────────────────────────────────────────────────────────
|
|||
|
|
LOG_DIR = Path(__file__).parent / "run_logs"
|
|||
|
|
LOG_DIR.mkdir(exist_ok=True)
|
|||
|
|
ts = datetime.now().strftime("%Y%m%d_%H%M%S")
|
|||
|
|
|
|||
|
|
# Per-trade CSV (without raw vd_seq — too large)
|
|||
|
|
calib_path = LOG_DIR / f"vd_calib_{ts}.csv"
|
|||
|
|
with open(calib_path, 'w', newline='') as f:
|
|||
|
|
cw = csv.writer(f)
|
|||
|
|
cw.writerow(['trade_id','asset','exit_reason','bars_held','pnl_pct','pnl_abs',
|
|||
|
|
'leverage','is_winner','entry_vd','exit_vd','max_vd','min_vd','mean_vd',
|
|||
|
|
'pct_bars_positive_vd','n_vd_bars'])
|
|||
|
|
for r in trade_records:
|
|||
|
|
cw.writerow([r['trade_id'], r['asset'], r['exit_reason'], r['bars_held'],
|
|||
|
|
f"{r['pnl_pct']:.6f}", f"{r['pnl_abs']:.4f}", f"{r['leverage']:.4f}",
|
|||
|
|
int(r['is_winner']), f"{r['entry_vd']:.6f}", f"{r['exit_vd']:.6f}",
|
|||
|
|
f"{r['max_vd']:.6f}", f"{r['min_vd']:.6f}", f"{r['mean_vd']:.6f}",
|
|||
|
|
f"{r['pct_bars_positive_vd']:.2f}", r['n_vd_bars']])
|
|||
|
|
|
|||
|
|
# Sweep CSV
|
|||
|
|
sweep_path = LOG_DIR / f"vd_sweep_{ts}.csv"
|
|||
|
|
with open(sweep_path, 'w', newline='') as f:
|
|||
|
|
cw = csv.writer(f)
|
|||
|
|
cw.writerow(['threshold','debounce','pct_triggered','sensitivity','specificity',
|
|||
|
|
'precision','separation','wr_triggered','wr_not_triggered',
|
|||
|
|
'avg_pnl_triggered','avg_pnl_not_triggered','avg_bars_saved',
|
|||
|
|
'n_triggered','n_not_triggered'])
|
|||
|
|
for r in sweep_rows:
|
|||
|
|
cw.writerow([r['threshold'], r['debounce'], r['pct_triggered'],
|
|||
|
|
r['sensitivity'], r['specificity'], r['precision'],
|
|||
|
|
r['separation'], r['wr_triggered'], r['wr_not_triggered'],
|
|||
|
|
r['avg_pnl_triggered'], r['avg_pnl_not_triggered'],
|
|||
|
|
r['avg_bars_saved'], r['n_triggered'], r['n_not_triggered']])
|
|||
|
|
|
|||
|
|
print(f"\n{'═'*80}")
|
|||
|
|
print(f" per-trade → {calib_path} ({len(trade_records)} rows)")
|
|||
|
|
print(f" sweep → {sweep_path} ({len(sweep_rows)} rows)")
|
|||
|
|
print(f"{'═'*80}")
|