Backtesting without lying to yourself
4 min read · educational content, not investment advice
A backtest is an experiment. Experiments go wrong in predictable ways, and knowing the failure modes matters more than any indicator.
The four ways it lies
- Overfitting. Tuning parameters until the past looks perfect. The curve fits history and nothing else.
- Look-ahead. Using information you could not have had at the time — a close before it happened, news before it printed.
- Survivorship. Testing on the names that exist today, ignoring the ones that went to zero.
- Cost blindness. Assuming clean fills at the printed price with no spread and no slippage.
Three checks that catch most of it
- Out-of-sample: build the rules on one period, then test untouched on a later one. If it collapses, you overfit.
- Add costs: rerun with realistic slippage. A strategy that only works at zero cost does not work.
- Count the trades: thirty trades prove nothing. Statistical significance needs a sample your sample size does not have.
A backtest tells you a strategy is possible. Only live, logged trading tells you whether you can execute it.
Once you trade it, run it through the review page — that is the only test that includes you.