To backtest an MT5 gold EA the right way, you need real tick data, multi-year coverage, and a clear-eyed reading of the strategy tester report. Anything less, and you're optimizing for a fantasy that won't survive its first live week on XAUUSD.

This guide explains how to set up the MT5 Strategy Tester properly and how to interpret the metrics that actually matter — drawdown, profit factor, recovery factor, and Sharpe ratio.

Setting up the MT5 Strategy Tester

Open View → Strategy Tester (Ctrl+R) and configure the run:

  1. Expert Advisor: select the EA you're testing.
  2. Symbol: XAUUSD on the same broker server you'll trade live. Different brokers have different historical data.
  3. Period: H1 or M15 — match the EA's design.
  4. Modeling: always "Every tick based on real ticks." The other modes are faster but lie.
  5. Date range: minimum 3 years. 5+ is better. Anything less than 1 year is a coin flip.
  6. Initial deposit: match what you'll trade live.

MetaQuotes' official testing documentation covers each setting in depth.

Reading the report: the four metrics that matter

1. Maximum drawdown (%)

The biggest peak-to-trough equity decline. If the report shows 28% max drawdown, ask yourself honestly: would you keep trading after losing more than a quarter of your account?

  • Under 15%: excellent — psychologically tradable.
  • 15–25%: realistic for a working strategy with edge.
  • Over 30%: red flag, even if returns look great.

2. Profit factor

Gross profit divided by gross loss. PF of 1.5 means the EA earns $1.50 for every $1 lost.

  • 1.2–1.6: realistic and sustainable.
  • 1.6–2.5: very good — if drawdown is also reasonable.
  • Over 4: almost always over-optimized. Be skeptical.

3. Recovery factor

Net profit divided by max drawdown. Tells you how efficient the strategy is at producing returns relative to the pain. A recovery factor over 3 is solid; over 5 is excellent. Below 1 means the EA spends more time clawing back losses than making money.

4. Sharpe ratio

Risk-adjusted return. The CME Group's Sharpe ratio explainer covers the math, but the rule of thumb: above 1.0 is good, above 1.5 is excellent, above 2.0 is rare and worth investigating for over-fitting.

The biggest backtest trap: curve fitting

Curve fitting is when an EA's parameters are tuned so tightly to historical data that they no longer work on anything else. Symptoms:

  • Win rate above 85% with shallow drawdown.
  • Massive profit on the optimized period, garbage on out-of-sample.
  • Tiny changes to inputs collapse performance.

The defense is walk-forward testing — optimize on 2018–2022, validate on 2023–2025 untouched. If the validation period stays profitable with similar drawdown, you have a real edge. If it falls apart, you have a beautifully fitted curve and nothing more.

Personal insight: the trade list tells more than the headline

After reading hundreds of strategy tester reports, the single most useful section is the trade-by-trade list. Look for:

  • Cluster of losses in any single month — usually a regime the EA can't handle.
  • Outsized winners doing all the heavy lifting — fragile dependency on a few trades.
  • Long stretches of small wins followed by one giant loss — classic martingale signature.

A healthy equity curve climbs steadily with limited variance, not in dramatic bursts. For an example, see how NextTrade documents its rule logic openly.

Related reading