Glossary
Backtest
A test of fixed trading rules on historical data, as if they had been traded at the time. It is only as honest as its data and costs.
A backtest applies fixed trading rules to historical prices and records the trades they would have produced. A useful backtest uses only information available at each moment, fills orders at realistic prices and deducts spread, commission, swap and slippage. Every change made after looking at the results uses up some of its credibility.
Example
A rule on gold produces 1,000 trades from 2012 to 2021, averaging +0.3 bp per trade before costs. With costs of 1.1 bp per trade, the average becomes 0.3 − 1.1 = −0.8 bp. Over 1,000 trades that adds up to about −800 bp: a losing rule that looked profitable before costs.
Why it matters
Most rules that look good in a backtest fail later, usually because of costs, look-ahead errors or too many variants tried on the same data. That is why our lab publishes the rules that failed, with their costs, next to the few that held up.