Raptor 2: Backtest Report and Investment Assessment

Reading notice

Backtest results only. This is a historical simulation, not a live trading signal. No real capital has touched this strategy. Nothing in this report is a promise of future returns.

1. Which Raptor Is Which

Five versions were built and tested. Only one passed. This table is the official lineup so there is no confusion about names.

Raptor 2 is P2E. Not Raptor C, not Raptor D. P2E is the file, Raptor 2 is the name. P2C is the control and was never a Raptor.

2. Executive Summary: What Raptor 2 Is

Raptor 2 is an automated trading program (an Expert Advisor) for MetaTrader 5. It trades EURUSD on the 1-hour chart. It watches for one specific price pattern: a 1-2-3 reversal pullback in the direction of the daily trend. When the pattern appears, it places the trade.

The one thing that makes Raptor 2 different from the original strategy: it does the exact opposite. Where the original bought, Raptor 2 sells. Where the original sold, Raptor 2 buys. The entry timing, the stop loss, and the profit targets are identical. Only the direction is flipped.

Over nearly three years of historical data (October 2023 to September 2026), it placed 142 trades and finished up $922 on a $10,000 account. That is a 9.22% total gain, roughly 3% per year. It is the only variant out of five that passed all five advancement gates.

It is a real, verified edge on historical data. It is not ready for anyone's money yet.

3. The Numbers: Full Backtest Results

Tested in the MetaTrader 5 Strategy Tester. Every tick based on real ticks. History quality 99%.

4. The Five Gates: 5 of 5 PASS

A variant only advances if every single gate holds. Raptor 2 is the first variant to clear all five.

For comparison: the control (P2C) lost 16.82% over a shorter window. Raptor 1 (P2D) lost 29.88%. Raptor 3 (P2F) lost 18.32%. Raptor 4 (P2G) produced results identical to the control, which means its test was invalid and it is tabled for repair.

5. What Makes It Different: The Action-Flip Insight

The original strategy lost money in every test. Raptor 2 changed exactly one variable: the direction of the trade. Same pattern, same timing, same stop, same targets. The only difference is that buy became sell and sell became buy.

The result flipped from a consistent loss to a consistent gain. The honest interpretation: the signal's TIMING may contain real information, but the original DIRECTION was wrong. The market was telling the truth about WHEN to trade and lying about WHICH WAY.

That is a hypothesis supported by one backtest, not proof. It held up over 142 trades across nearly three years, which is enough to take seriously and not enough to bet money on. The next tests will try to break it.

6. The Honest Limitations

Anyone who puts money near this needs to hear these first. No exceptions.

• The return is modest. 9.22% over nearly three years is about 3% per year. A plain savings account does similar with zero risk and zero drawdown. The edge is real but small at 1% risk per trade.
• Seven out of ten trades lose. A 29.58% win rate means long losing streaks are guaranteed, not just possible. The math works because the winners are bigger than the losers, but most people cannot sit through that many red trades without overriding the system. If you cannot stomach it, the edge does you no good.
• The drawdown is real. 18.87% max drawdown means the account was down almost $1,900 at its worst point. On a $10,000 account, that is the kind of hole that makes people quit or do something stupid.
• This is a backtest, nothing more. Historical simulation. No out-of-sample test on unseen data. No forward demo test in live market conditions. No live trading. Past simulated performance does not predict future results, and anyone who says otherwise is selling something.
• One market, one timeframe. EURUSD on the 1-hour chart only. There is zero evidence it works on any other pair, any other timeframe, or in different market regimes.
• Costs are modeled, not real. Spreads and commissions in the tester are estimates. Live conditions include slippage, requotes, and spread widening around news that the tester does not fully capture.

7. What Would Make It Investable

Four steps, in order. No skipping.

1. Step 1: Frozen-parameter out-of-sample test. Run Raptor 2 with zero changes on a different, unseen time period. It must pass all five gates again. If it fails, the edge was a mirage of the test window.
2. Step 2: Forward demo test, 3 to 6 months. Run it on a demo account in real time. Live spreads, live slippage, news events, everything the backtest cannot simulate. The numbers must hold up.
3. Step 3: Improve the return or cut the risk. Through disciplined one-variable variations: either lift the ~3%/year return or reduce the ~19% drawdown. The current ratio is not good enough for outside capital.
4. Step 4: Small live capital only. After steps 1 through 3 pass, live trading starts with money that can be fully lost. Position sizes stay small until a full year of live results exists.

8. Cleanup Guide: What to Keep, What to Delete

The MetaEditor file list has five versions. Two are dead weight. Confirm the exact filenames in MetaEditor before deleting anything. Deletion is permanent.

Delete (extinct, failed every gate)

• Blackfish_EA_P2D.mq5 and its compiled .ex5 (Raptor 1, flipped the signal, lost 29.88%)
• Blackfish_EA_P2F.mq5 and its compiled .ex5 (Raptor 3, wider stop, lost 18.32%)

Keep

• Blackfish_EA_P2C.mq5 (the control. Every future variant is measured against this. Never delete.)
• Blackfish_EA_P2E.mq5 (Raptor 2, the winner. This is the one.)
• Blackfish_EA_P2G.mq5 (Raptor 4, tabled. Its take-profit test was invalid and it needs repair, not deletion.)

Suggested folder hygiene: move the two extinct versions into an archive subfolder rather than deleting outright, until the next variant is built and confirmed working. Cheap insurance.

Closing

On the night of October 1, 2026, after four failed variants and one invalid test, a single flipped variable turned a losing strategy into a winning one. Raptor 2 passed all five gates over 142 trades and nearly three years of data. It is the first real lead the BFG project has produced.

It is a lead, not a victory. The work now is to try to break it: out-of-sample, forward demo, disciplined variations. If it survives that, then it earns capital. Not before.