Backtest
What does replaying a trading rule on past prices actually prove, and what does it leave unproven?
In plain words
A backtest applies a trading rule to historical data and records what it would have done, showing hypothetical entries, exits and results. It is a replay of the rule against one particular sequence of past prices, not a forecast.
Why it matters
Backtests are how rules get sized, stress-tested and compared before capital is committed. But the same rule can look brilliant in one window and flat in another, and past data cannot supply the fills, costs or liquidity a live market will.
In a real week: ETH, week of 18 August 2025
ETH's week of 18 August 2025 shows the limits: a rule judged on the weekly close of 4780.15, up 6.88%, would have hidden a 22.09% range, a 5.5% worst day on 19 August and a 14.36% best day on 22 August, with realised volatility at 85.9%. Volume of 24.04bn was 1.8 times its prior 12-week average, and the week ranked in the 73rd percentile of 392 weeks of history.
A common mistake
Reading a strong backtest result as proof the rule will keep working. It describes one historical path only; a different sequence, or the same path with real costs, can change the outcome sharply.