Backtest in gold trading

A backtest runs a strategy’s rules over historical data to see how it would have performed. It can be manual, stepping through old charts bar by bar and recording each trade, or automated with code. The output is a sample of trades from which win rate, average R, expectancy, maximum drawdown and longest losing streak can be calculated.

How it applies to gold trading

Gold backtests need care with three things. Data: spot and futures prices differ, and the daily break and weekend gap must be in the data. Costs: a 30-cent spread plus swap over a multi-day hold can turn a marginal method negative. Sessions: a rule that works in the London session may fail in the Asian range, so results should be split by time of day. A manual backtest of 100 trades on a gold breakout rule, recorded in R, is enough to show whether the idea deserves a forward test.

Backtests overstate performance when rules are tuned to the same data they are tested on, a problem known as curve fitting. Testing on one period and checking on another, and keeping rules simple, reduces it.

A manual backtest also trains pattern recognition: after a hundred screenshots of a setup forming and failing, the trader sees it faster in real time, which is a benefit no automated test provides.

A common mistake

Testing on a few months of a strong gold trend and concluding the method works. Gold’s behavior in trends, ranges and crisis spikes differs, and a method should be tested across all three.

See also: Forward test, Expectancy, Drawdown, Trading journal.

Frequently asked questions