Is backtesting accurate?

A well-built backtest is accurate about the past and silent about the future. If the engine models fees, slippage and only uses information available at each bar, its numbers faithfully describe how the rules would have traded that period. What it cannot promise is that the next period behaves like the last one.

Accuracy problems come in two layers: engine bugs (lookahead, unrealistic fills, repainting indicators) that make even the past wrong, and interpretation errors (curve-fitting, cherry-picked windows) that make a correct past misleading. The first layer is the platform's job; the second is what robustness scoring, walk-forward analysis and out-of-sample tests exist to catch.