I once ran a backtest that came back with an 86.5% win rate over 680 trades. My first reaction was excitement. My second, about ten seconds later, was suspicion — and it turned out to be justified: the strategy was allowing unlimited re-entries within the same session, and it was quietly compounding the same winning move over and over. The headline number looked great. The strategy was broken.
That’s the trap with backtest results: a single return number can’t tell you whether you found an edge or a bug. You need to look at the shape of the results, not just the total.
The checks that actually catch this
Here are the patterns worth checking on every trade log before you trust it, in roughly the order I’d check them:
Sample size. Below roughly 100-300 trades, the statistical margin of error is too wide to distinguish real edge from luck. A great-looking 40-trade backtest tells you almost nothing.
Win rate plausibility. Above ~75% is unusual enough that it’s almost always a logic bug — an unclosed position, a look-ahead leak, a re-entry loop — rather than genuine skill.
Directional asymmetry. If longs are profitable and shorts aren’t (or vice versa), check whether that’s a real property of the strategy or just the market’s bias during the tested window.
Trade frequency. A strategy that’s supposed to trade once per session but shows 12 trades/day on average is very likely re-entering without a real exit signal in between.
Time coverage. A backtest confined to a few months is a backtest confined to one market regime. It tells you how the strategy did in that regime, not how it will do generally.
Transaction cost realism. Suspiciously clean P&L numbers (round figures, no fractional cents) often mean slippage and commission were never modeled.
Two more — look-ahead bias and walk-forward validation — need extra columns most trade logs don’t include (entry price plus the trade’s high/low, or an explicit in-sample/out-of-sample split). When those columns are missing, the honest answer is that the check can’t run, not a fabricated pass.
Why this matters more than the return number
A backtest bug that inflates your win rate doesn’t just give you a wrong number — it gives you false confidence right before you risk real capital. The checks above won’t tell you your strategy is good. They’ll tell you whether the result you’re looking at is even measuring what you think it’s measuring.
Try the free interactive demo first if you want to see it in action before anything else: it runs entirely in your browser, no install, no upload.