Metals

A Year of Real-Tick Testing on Gold: Why Most XAUUSD Backtests Flatter You

18 August 2026, 07:00
Kenichiro Sakamoto
Kenichiro Sakamoto
0
24
Gold is where most EA money is spent and where most EA disappointment happens. After a year of testing gold strategies — our own and reconstructed versions of popular approaches — on real broker ticks across full market cycles, here is what the data kept repeating.

1. The recent window is doing most of the work.
Gold has spent the last few years in a strong uptrend. Almost any buy-side logic — averaging down, dip buying, breakout chasing — looks brilliant on 2023-2026 data. The same configurations run through windows containing 2013, 2020 or 2022 tell a different story: what looked like an edge was mostly the era. If a gold backtest starts after 2022, that choice is informative.

2. Whipsaw is gold's defining feature, not an exception.
On real ticks, gold's sharp intraday reversals systematically punish two popular archetypes: tight-stop breakout entries (stopped out by the wick, then price goes without you) and thin take-profit scalps (the spread and slippage eat the edge). Strategies that survive tend to work on larger scales — daily ranges, wide ATR-based stops — where a single wick cannot decide the trade.

3. Session behavior is real but fragile.
Gold does behave differently in Asian, London and New York hours, and time-filtered systems exploit this. But the finer the time window, the faster it decays and the more broker-timezone details matter. An edge that needs one specific hour is a rounding error away from no edge.

4. Costs are a first-order factor.
Gold spreads vary several-fold between brokers and across the day. A strategy that is profitable at one broker's average spread can be structurally unprofitable at another's. Always re-test on your own broker's data — the free demo of any Market EA runs in your own tester precisely for this.

What we concluded for ourselves
We hold gold strategies to the same gate as everything else: profitable on at least two real feeds, across windows that include crashes, with equity drawdown reported. Very few gold ideas pass. The ones we sell publish their losing periods openly, because on this symbol especially, anyone promising smooth monthly profit is describing the recent past, not the instrument.

The gold system we do sell, with its losing periods published in full: https://www.mql5.com/en/market/product/187329

Our measured backtest data for every EA (profit factor, equity drawdown, trade count, year-by-year results) is published at fxea365.com/ea/ranking

#gold, xauusd, expert advisor, backtest, real ticks