Backtesting
Backtesting is evaluating a trading strategy or forecasting model by applying it to historical data to see how it would have performed in the past.
A backtest takes clear rules, such as 'buy when the 50-day average crosses above the 200-day average,' and simulates trades on past prices. The results show hypothetical returns, volatility, maximum drawdown, win rate and other statistics, which can be compared with a benchmark like buy-and-hold.
Good backtests avoid several traps. Look-ahead bias means using data that would not have been available at the time. Survivorship bias means testing only on companies that still exist, ignoring those that went bankrupt or were delisted. Overfitting means tweaking rules until they fit past noise perfectly, which usually fails on new data.
To reduce these risks, analysts test on out-of-sample data, use walk-forward methods that re-estimate a model only with information up to each point in time, and include trading costs and taxes.
Even a sound backtest is not a guarantee. Markets change, and past performance does not ensure future results. Example: a strategy that returned 15% a year from 2010 to 2020 might perform very differently in a period of higher interest rates.