We Ran One of GOLD NEURON's Three Engines Alone for Five Years: Two Years Were Red

GOLD NEURON (AI) trades XAUUSD with three engines sharing one risk budget: a Trend engine, a breakout engine, and a Pivot engine gated by a machine-learning model that scores each pivot setup and only fires above a fixed threshold. Before shipping, we ran each engine on its own. This is what the Pivot engine did alone. Two of the five years were red, and that is why the product has three engines instead of one.

Pivot engine only, the other two switched off. XAUUSD M30, Ultra preset (×3.5 on a 1% base risk), threshold fixed at 0.502. Each year started fresh at $10,000 so no year can hide another. A real-tick XAUUSD feed. Net figures in dollars.

Year | Trades | Win rate | Gross win / loss | Net | PF | End balance

2022 | 37 | 86.5% | +3,633 / −4,182 | −550 | 0.87 | 9,055

2023 | 39 | 89.7% | +3,068 / −4,475 | −1,407 | 0.69 | 8,598

2024 | 41 | 100.0% | +6,619 / 0 | +6,619 | ∞ (no losing trade) | 16,671

2025 | 44 | 93.2% | +9,749 / −4,620 | +5,129 | 2.11 | 15,045

2026 (Jan–Sep) | 26 | 73.1% | +6,472 / −5,727 | +746 | 1.13 | 10,746

As one 4.5-year test the engine looks strong. Split by year, nearly all of the return came from 2024, with 41 trades and zero losers, plus 2025, with only three losses. Without those two years, what remains is two red years and one that barely broke even. Win rates of 86.5% and 89.7% still produced losses, because the few losing trades were larger than the many winners.

Between 27 and 45 setups passed the threshold every year. The model was not more selective in the good years or looser in the bad ones. What changed was the market, not the model. A seller who shows you the 2024 line is showing you the market of 2024, not a filter that knew something.

We expected higher-scored setups to win more often. In 2022 the losing trades averaged a higher score, 0.630, than the winning ones at 0.591. A score that is higher on the losers than on the winners is not ranking anything useful. It gates the entry count; it does not tell you which entries will pay.

Lowering the threshold to let more setups through made the results worse in testing, not better. We never shipped a pivot-only configuration: an engine that is red in two years out of five cannot carry a product by itself. The three engines are there for diversification across mechanisms, not because any one of them is a better single signal.

Our public forward account for GOLD NEURON is a demo account, started on 17 July 2026. As of this writing: 36 trades, PF 1.03, gain +1.72%, drawdown 39.38%. We will not dress that up. It is a short sample in a drawdown, consistent with what the yearly table already says about this kind of engine.

A high win rate engine with a few large losses is judged by its worst year, not by its average. If 2023 alone, −1,407 on $10,000 at an 89.7% win rate, is more than you would sit through, the five-year average is not the number to look at. Ask any seller for the yearly split. If they only show the total, you now know why.

Our three best-selling EAs: GOLD NEURON (AI) https://www.mql5.com/en/market/product/187329 · ATLAS PORTFOLIO https://www.mql5.com/en/market/product/182751