A backtest applies a rule to historical data to see what would have happened. It can be a powerful educational tool because it turns an idea into something measurable. But backtests are also easy to abuse. A beautiful equity curve can be the result of hidden assumptions, not a durable idea.
Overfitting
Overfitting happens when a rule is tuned too closely to the past. If a student tries hundreds of parameter combinations and only reports the best one, the result may simply be luck. A rule can memorize historical noise and fail when conditions change.
Look-ahead bias
Look-ahead bias occurs when the backtest uses information that would not have been available at the time. For example, using a final revised dataset inside an earlier decision point can make performance look unrealistic. The rule should only use information available up to that moment.
Survivorship bias
Survivorship bias appears when failed or delisted assets are removed from the sample. If only surviving winners remain, historical performance can look cleaner than reality. This is especially important when studying stocks, funds, or crypto tokens.
Costs and friction
Backtests often ignore fees, spreads, taxes, slippage, and delays. These frictions can turn an exciting chart into an ordinary or negative result. Educational backtests should show assumptions clearly and treat costs as part of the model.
Good backtest habits
Use train/test splits, report failed experiments, show sensitivity to parameter changes, and write limitations in plain language. A credible backtest does not try to impress the reader. It helps the reader understand where the idea is strong, weak, and uncertain.