quantcheck — Test Your Backtest for Overfitting (Free)

Why most backtests lie

It's rarely deliberate. It's the math. Optimise a strategy long enough, try enough indicators and settings, and something will look spectacular on historical data — purely by chance. You didn't find an edge; you found the luckiest of the variants you tried.

Three quiet killers do most of the damage:

• Overfitting — the strategy memorised the noise of the past instead of a real pattern.
• Data-mining — test 10,000 combinations and one looks like genius. It isn't.
• No out-of-sample — a curve that only shines on the data it was tuned on tells you nothing about tomorrow.

What the free check gives you

Upload your backtest's returns and get a plain-language verdict in seconds — no code, no signup for the result:

• Deflated Sharpe Ratio — is your Sharpe real, or just the best of the many variants you tried?
• Probability of Backtest Overfitting (PBO) — does your in-sample winner hold up out-of-sample?
• Stationary bootstrap — a confidence interval on your Sharpe, so you know how stable it really is.

It reads your export from TradingView, MetaTrader 4/5, freqtrade, or a generic returns / equity / pnl CSV (European ; / decimal-comma is auto-detected). Your data is processed in memory and never stored.

Deep — for when it has to be right

The paid Deep tier adds the heavier, institutional-grade tests you'd want before trusting — or selling — a strategy:

• White's Reality Check & Hansen's SPA — the data-snooping test: does the best of all your variants genuinely beat the rest, or is it noise?
• Combinatorially-Purged Cross-Validation (CPCV) — a far more honest picture of out-of-sample performance than a single split.
• A shareable, dated report — the evidence, if you want to prove an edge to someone else.

Trading without knowing your real edge is like driving with your eyes closed. You might get lucky — but you'll crash.