How much historical data do you need for a reliable backtest?

Enough to produce at least 30–100 closed trades across more than one market regime — for daily-bar crypto strategies that usually means 3+ years of data. Trade count matters more than calendar length: 15 trades tell you almost nothing regardless of how many years they span.

The regime point is easy to underestimate. A trend-following strategy tested only on 2023–2024 has never met a real bear market; a mean-reversion strategy tested only on a ranging year has never met a trend that runs away from it. If your data window contains a single regime, your backtest measures that regime, not your strategy.