BFG · Variance Lab
The paddock · every egg gets its chance to hatch
P2C is the DNA. Every variant is an egg: one changed gene, a complete build, and a chance to hatch. Most eggs do not make it. The ones that do have to prove they can hunt before they ever see live capital. Nothing is lost, nothing is judged on a handful of trades.
First raptor · 2026-10-01
Extended-history backtest (2023-10-01 to 2026-09-22): 142 trades, profit factor 1.10, net +$922.19. Net > 0, PF > 1.00, expectancy > 0R, total R > 0, 100+ trades. Every gate holds. This is a backtest milestone, not a live signal. Next: frozen-parameter out-of-sample, then forward demo. See the full Raptor 2 record → · Read the full report → · Full backstory and analysis brief →
Audio briefing, just over four minutes. What Raptor 3 and Raptor 4 change, why we are taking the chance, and why it might work.
The reference implementation. Backtested on 10,605 H1 bars of EURUSD (2025-01-01 to 2026-09-22):
P2B was the original attempt. It had an error code and was retired. P2C is the sole current base and control.
Expectancy (R per trade) for every version, P2B through the target bar. Below zero pays the market. Above zero is the hunt.
P2B died after 4 trades. P2C survived 100 and still pays the market -0.139R per trade. The raptors change one gene each. The bar is not a prediction, it is the standard. 2026-10-01: Raptors 1 and 3 extinct, Raptor 2 promoted on 142 trades with all five gates passing, Raptor 4 tabled.
P2 was the original pattern scanner build. P2B was the first demo EA: 4 trades, 4 full stop-outs, -$3,547.42. Retired 2026-09-30. Dead, never a baseline, never a fallback.
P2C is the control. Smart-entry order fix, magic 20260922. Backtested on 10,605 H1 bars of EURUSD, 2025-01-01 to 2026-09-22:
As specified, the strategy loses. The rule stands: fix the logic, not the test. Raptor 1 (P2D) inverts the bias before detection. Raptor 2 (P2E) inverts the detector output. Raptor 3 (P2F) widens the ATR stop from 0.5x to 1.0x. Raptor 4 (P2G) moves TP1 from 1.5R to 1.0R. One gene each, everything else frozen. All four backtested 2026-10-01: Raptors 1 and 3 extinct, Raptor 2 promoted after clearing all five gates on 142 trades, Raptor 4 tabled for TP1 repair.
Locked 2026-09-30. EURUSD · H1 · 2025-01-01 to 2026-09-22 · every tick based on real ticks · optimization OFF. Tester gates: InpLiveTrading=true, InpAllowOrders=true (tester only, source defaults stay false). Run order: P2C first to reproduce the baseline, then Raptor 1, then Raptor 2, then Raptor 3, then Raptor 4. Nothing changes between runs except the EA.
A run clears only if all hold: net profit > $0, profit factor > 1.00, expectancy > 0R, total R > 0, 100+ trades. Then out-of-sample on a fresh period with frozen parameters. Win rate is not a criterion.
Everything moves forward here, on the site, not in chat. Updated 2026-10-01.
2026-10-01 session: all four raptors compiled clean and backtested on the locked settings. Raptor 2 (P2E) first passed 4 of 5 gates on 78 trades (net +7.35%), then cleared all five gates on extended history back to 2023-10-01: 142 trades, PF 1.10, net +$922.19. Promoted to Raptor. Raptors 1 and 3 are extinct. Raptor 4 is tabled: its TP1 partial never triggered, so the 1.0R level was never actually tested.
The change: tradedBias = -BiasDirection() before the pattern check. Buys dips in daily downtrends, sells rallies in daily uptrends.
The bet: the daily trend is exhausted; trade the snap-back, not the continuation.
Result: failed all gates. Bias-flip hypothesis eliminated.
The change: raw bias unchanged, baseSig inverted to sig. P2C BUY becomes SELL, P2C SELL becomes BUY.
The bet: the signal is right about timing but wrong about direction.
Result: 142 trades · PF 1.10 · net +$922.19 · all five gates pass on extended history (2023-10-01 to 2026-09-22). First variant promoted to Raptor. Next: out-of-sample, then forward demo.
The change: new input InpStopAtrMult = 1.0; ATR stop widens from 0.5x to 1.0x ATR on both sides. 1% risk and R targets adjust automatically.
The bet: 69% of P2C trades stopped out fully; the 0.5 ATR stop sat inside H1 noise.
Result: failed all gates. Stop-distance hypothesis eliminated.
The change: InpTp1RMultiple default 1.5 → 1.0; 75% of the position banks at 1.0R.
The bet: only 16% of P2C trades ever reached TP2; take the first achievable leg.
Result: invalid test, the TP1 partial never triggered. Tabled for mechanism repair.
Restricted
Type the password to enter.