How can you tell if a backtest is trustworthy?

Interrogate five things. (1) Window: does it span at least one ugly regime (2022), or only a bull run? (2) Benchmark: is buy-and-hold of the same asset shown on the same window? (3) Costs: are fills, fees, and slippage modeled or at least disclosed? (4) Trade list: can you inspect individual decisions, or only the headline curve? (5) Provenance: who ran it — the strategy's author (cherry-picking risk) or a neutral engine on standardized terms? A trustworthy backtest also states its assumptions plainly: starting capital, data resolution, execution timing. Distrust any result you can't reproduce, and treat every backtest — including honest ones — as an upper bound on live performance, not a forecast.