GOLD VIPER Got a Second Engine: The Full Backtest at Four Risk Levels, Including the One We Rejected

GOLD VIPER is our XAUUSD breakout EA, and its original engine is deliberately selective: 294 trades in about 7.7 years of real-tick testing. From version 2.35 it ships with a second engine built in, a fractal-breakout engine with eight sub-strategies that we developed separately and ported over. It is always on, and it has its own risk input, FB_RiskPercent, scaled by the preset multiplier. Below is the full test at four settings of that input, including the one we decided not to ship.

XAUUSD M30, real ticks, 2019-01-01 to 2026-09-01, $10,000 starting balance, on a real-tick XAUUSD feed with a 1:500 account. The original engine ran at its default 3% risk in every run. The only thing that changed between runs was FB_RiskPercent. All figures are net.

Setting | Net profit | PF | Balance max DD | Equity max DD | Engine max lot | Engine-only DD | Engine share

Second engine OFF | +229,585 | 1.60 | 20.4% | 27.5% | – | – | –

FB_Risk 0.25% | +384,804 | 1.48 | 23.6% | 30.3% | 0.77 | 13.6% | +145k

FB_Risk 0.5% (shipped default) | +910,766 | 1.51 | 24.9% | 30.8% | 2.95 | 17.3% | +550k

FB_Risk 1.0% (rejected) | +2,809,799 | 1.57 | 26.1% | 32.2% | 5.0 (cap) | 29.0% | +2,008k

Trade counts: 294 with the second engine off; 2,806 at the shipped 0.5% (275 original engine, 2,531 second engine). So the original engine took 19 fewer trades with the second engine on. We have not traced why. The plausible reason is that the two engines share one account, so the balance, margin and open positions of one can affect the other, but that is untested, and we will not claim the engines are fully independent.

The second engine fired the same 2,531 trades at every risk level. Only the lot size differs. So the jump in the engine's share from +145k to +550k to +2,008k is not better signals; it is percent-risk compounding on the same signals.

At 1.0% the engine reaches a 5.0 lot, which is the lot cap. From there the upside is capped and cannot compound further, while the engine-only drawdown nearly doubles, from 17.3% to 29.0%. At 0.5% the engine tops out at 2.95 lots, below the cap, with an engine-only drawdown of 17.3%. That is the trade-off we chose, and the input is there if you want to choose differently.

Percent risk over 7.7 years compounds, so the late-year dollar figures are inflated by the balance the early years built. +910,766 on $10,000 is not a number to plan around. The drawdown percentages and the profit factor are stable across balance size; the net profit is not.

With the second engine on, the original engine's profit also rose. That is not evidence that it traded better: it took 275 trades instead of 294, and the rise is explained by the larger balance, since 3% risk on a larger balance is a larger lot.

Same test, shipped 0.5%, buy-only: 1,668 trades, +246,486, PF 1.85, DD 24.17%. Sell-only: 1,164 trades, +42,400, PF 1.40, DD 22.15%. The long side carried the period, which is unsurprising for gold from 2019 to 2026.

The engine's own equity curve, taken on its own, was positive in every year of the test; we mention it as a qualitative fact, not a promise. Our public GOLD VIPER forward account is a demo started 2026-07-25 and has taken exactly one trade so far, so the forward record says nothing yet, in either direction. And a backtest on one broker's feed is one sample. We have not tested this engine on a second feed.

Our EAs in this article: GOLD VIPER https://www.mql5.com/en/market/product/182297 · GOLD NEURON (AI) https://www.mql5.com/en/market/product/187329