Backtesting applies a trading strategy to historical data to estimate how it would have performed. You specify the rules — buy when the 50-day SMA crosses above the 200-day SMA, sell when it crosses below — and the backtest engine calculates what would have happened if you had followed those rules in the past. The result is a historical performance record: return, drawdown, trade frequency, and win rate.
The uses and abuses of backtesting are equally significant. The honest uses: understanding the historical behaviour of a strategy in different market regimes, stress-testing a rule against its worst-case historical periods, and comparing strategy variants on an apples-to-apples basis. The common abuses: over-fitting rules to look-back data (curve fitting), ignoring transaction costs and slippage, and treating a good backtest as a reliable forecast.
The core limitation is that financial markets are not stationary: the conditions that made a strategy profitable in one era may not persist. A strategy that was never backtested is guesswork; a strategy whose live results are far worse than its backtest suggests the backtest was misleading. Both failure modes are common and worth anticipating before real capital is committed.