Raptor 1: Flip the Signal

The exact change

In OnTick, the raw bias is computed exactly as P2C computes it, then inverted before it reaches the detector. Raw 0 (unknown) stays 0, so no pattern can pass the bias test on an unknown read.

int rawBias = BiasDirection();
int tradedBias = -rawBias;
int sig = DetectPattern(rates, atr, rsiBuf[0], zoneBuf[0], tradedBias);

The on-chart label reports both values so the traded side can never be misread, e.g. Bias(PERIOD_D1) raw: UP | traded: DOWN.

Verification · 2026-09-30

Single-variable confirmed. Token-by-token diff against P2C: the bias inversion is the only behavior change. Everything else is identical.

Source review: no issues found. In-house token diff plus an independent external MQL5 review (2026-09-30) found zero syntax problems across all 1,017 lines. Not a MetaEditor compile, that still needs F7 on a real terminal. Full independent report (PDF).

Status: compiled and backtested 2026-10-01. Verdict: extinct.

What it does

• D1 UP is traded as DOWN: waits for a bullish H1 bar in the pullback zone and sells it, betting the bounce fails and the down-move resumes.
• D1 DOWN is traded as UP: waits for a bearish H1 bar in the pullback zone and buys it, betting the dip snaps back.

Hypothesis

The daily trend is exhausted by the time the pullback prints. P2C bought dips that kept falling and sold rallies that kept rising. P2D trades the snap-back instead of the continuation.

Unchanged from P2C

Pattern geometry, bar-shape and RSI filters, pullback zone, ATR stops, 1.5R / 3.0R targets, 75% partial, runner, breakeven +0.1R, scale-in, 1% risk, 12-bar pending expiry, 21:00 cutoff, 3% daily loss halt, max setups. If P2D wins where P2C lost, it is purely the direction call.

Identity and safety

File

Blackfish_EA_P2D.mq5

Base

P2C (Blackfish_EA_P2C.mq5). P2B abandoned.

Magic number

20260930

Trading gates

InpLiveTrading = false, InpAllowOrders = false (defaults)

Demo only

No live trading. No real money.

Experiment status

Backtest results · 2026-10-01

EURUSD H1, 2025-01-01 to 2026-09-22, every tick based on real ticks. Tester gates InpLiveTrading=true, InpAllowOrders=true.

Total trades

130

Win rate

35.38%

Profit factor

0.53

Net

-29.88%

Max drawdown

34.07%

Gates passed

0 of 5

Verdict: extinct. The bias-flip hypothesis is eliminated. Flipping the D1 bias made every gate worse than the control.

Download

Complete .mq5 source. Compile in MetaEditor before attaching to any chart.