Which timeframe should I backtest — 1h, 4h or daily?

Start with daily bars. They have the most history per unit of noise, the smallest fee-and-slippage impact per unit of profit, and results that survive imperfect execution. Intraday timeframes multiply trade counts, which multiplies costs — a strategy that looks great on 15-minute bars before fees is often unprofitable after them.

Move down in timeframe only when the strategy's edge genuinely lives there (e.g. a volatility breakout that resolves within hours), and re-test fee sensitivity every step down. A rule of thumb: if halving your assumed fees flips the strategy from losing to winning, the edge is too thin for that timeframe.