ORBIT Campaign (SYNTHETIC — machinery validation)

generated 2026-08-10 19:45 UTC · 15m session-ORB breakout→retest→expansion · ATR stops/targets · 1.0% risk · costs embedded in every fill
OOS trades
263
Expectancy
+0.206 R
net of costs
Win rate
34.6%
Profit factor
1.37
Total
+54.1 R
$38,762 @ 1% risk
Max DD
9.1 R
t-stat 2.15

Per-symbol ATR parameters (walk-forward, plateau-selected)

symbol ATR period SL xATR TP xATR setups OOS n OOS expectancy R OOS total R
XAUUSD 14 2.000 4.000 93 65 0.070 4.600
US30 21 1.000 4.000 95 59 0.786 46.400
US100 21 1.000 2.500 103 74 -0.148 -11.000
US500 21 1.500 4.000 85 65 0.217 14.100

SL/TP grid — out-of-sample total R

Distribution & segments

By session

session n win% expectancy_R total_R PF
ASIA 87 29.900 0.090 7.900 1.150
LONDON 85 35.300 0.249 21.200 1.490
NEWYORK 91 38.500 0.275 25.000 1.490

By symbol

symbol n win% expectancy_R total_R PF
US100 74 25.700 -0.148 -11.000 0.780
US30 59 45.800 0.786 46.400 2.910
US500 65 38.500 0.217 14.100 1.400
XAUUSD 65 30.800 0.070 4.600 1.120

By exit

exit_reason n win% expectancy_R total_R PF
be 27 0.000 -0.014 -0.400 0.000
day_flat 1 100.000 1.354 1.400 inf
sl 142 0.000 -1.011 -143.600 0.000
time 15 80.000 1.054 15.800 14.600
tp 78 100.000 2.318 180.800 inf

ML gate (ensemble: logistic + gradient boosting, purged walk-forward)

gate NOT validated — keep pass-through, keep logging — threshold τ=0.0, keep rate 100%, uplift +0.000 R/trade, bootstrap p=0.507.

feature |logit coef|
direction 0.714
trend1h 0.519
rsi 0.454
dist_ema200_atr 0.431
displacement 0.413
ob_dist_atr 0.380
session_id 0.332
bbz 0.326
above_vwap 0.321
trend15 0.270
breakout_body 0.254
breakout_delay 0.253

Drawdown monitor (Landolfi, vol-unit accounting)

status: ok
years: 0.33
ann_vol_R: 40.23
max_dd_vol_units: 0.18
dd90_threshold: 1.4
flag: False
lo_sharpe_se: 2.14
note: flag=True => drawdown beyond 90th-pct Gaussian line: de-risk & review (Landolfi SSRN 7122358; short-history Sharpe uncertainty per Lo SE).
ORBIT engine v0.1 — research software; virtual trades only; nothing here is financial advice and no performance is promised. Out-of-sample = walk-forward test segments; all R figures are net of spread/slippage/commission. Evidence policy per docs/KNOWLEDGE_BASE.md.