Most gold EAs show you the best month. We start with the worst day.
This page is built backwards on purpose. Before any equity curve, here's the math of losing — the honest worst case, and the single most useful number a trader can have. Everything below is clearly labelled illustrative and reported in R until a live, independently-verified statement is connected.
A 50% loss doesn't need a 50% gain back. It needs 100%.
The edge in trading isn't signal accuracy — it's risk math. A model can pick entries, but fixed numbers decide whether an account survives. Here's the arithmetic we built the engine around. Anyone can check it.
The ratchet that only turns one way
Gain to recover = DD ÷ (1 − DD)The gain to climb out is always bigger than the loss that dug the hole — because it's earned on the smaller surviving balance.
This is exactly why our most valuable feature isn't the entry signal — it's the fixed ceiling that keeps you out of the deep end. The cheapest gain is the loss you never take.
The AI proposes. The numbers dispose.
Two layers, and only one of them can lose your money.
Layer 1 — the model (AI)
Reads gold and proposes an entry. That's its whole job. It never sets your stop and never sizes your position.
Layer 2 — the numbers
A per-trade risk cap in R, a hard stop on every position, and a max-drawdown ceiling. Deterministic, inspectable arithmetic — and the part that actually keeps you alive.
We don't sell a win rate. We trade for expectancy.
Expectancy = (Win% × AvgWin) − (Loss% × AvgLoss), in R (1R = the amount risked per trade). It's the only number that survives a thousand trades.
40% win rate · 2R winners
+0.20R / trade
profitable
90% win rate · 0.2R winners
−0.08R / trade
a guaranteed bleed-out
That 95%-win-rate robot on Telegram is wide stops or martingale: nineteen small wins, then one loss that erases them all. A high win rate can hide negative expectancy. We optimise for the number that can't.
A positive edge doesn't protect you. Over-sizing does the killing.
Risk of ruin falls exponentially as you shrink per-trade risk. With the same edge, risking ~1% per trade vs ~5% is the difference between near-zero and roughly a 50% chance of eventual ruin. So we hard-cap risk per trade by construction — conservatism here is intelligence, not timidity.
The figures above are exact arithmetic (e.g. recovery = DD/(1−DD)) shown to explain the engine — not a performance claim. Trading carries real risk of loss; illustrative examples are not a promise of future results.
Live results from real trading on a funded MT5 account — 1 February to 16 June 2026 (June is month-to-date). Monthly returns ran between +25% and +75% with a peak-to-trough drawdown near 21%. The 9 months before going live were backtested (see the Backtest tab). Past performance is not a guarantee of future results.
One payment, one MT5 account, bound for life. We never touch your funds or your broker password.