Guide
What a "reliability" score actually measures
Most signal marketplaces show a reliability percentage next to every strategy. It's easy to read "99%" as "99% likely to make me money." That's not what it measures, and the gap between the two readings is where a lot of buyers get surprised.
What it's actually tracking
A reliability score is almost always an execution-consistency metric, not a profitability one. It's built from things like: does the account trade often enough to evaluate, are there unexplained gaps in activity, does leverage or deposit load change unexpectedly, is there enough history to say anything at all. Those are real signals — a strategy that goes quiet for six days at a time is harder to copy reliably — but none of them say whether the underlying edge is real.
The result: a strategy can carry a 90%+ reliability score while its Sharpe ratio sits near zero, because reliability and edge are measuring completely different things. And a strategy can score low on reliability for a boring reason — it trades infrequently by design — while the edge itself is fine. Reliability tells you how easy the account is to observe, not how good the strategy is.
What to check instead
None of these are visible in a single headline number, which is exactly why they're worth pulling up individually:
- Sharpe ratio
- Return per unit of volatility. A strategy with huge growth and a Sharpe near zero got there on a few lucky trades, not a repeatable edge — the growth number alone can't tell you which one you're looking at.
- Maximum drawdown
- How far the account fell from its peak before recovering. A strategy that has already lost 40%+ of its equity once will very plausibly do it again — past drawdown is one of the few backtest numbers that tends to repeat.
- Profit concentration
- What share of total profit came from the single best trade, or the best handful of days. A track record where 80% of the growth happened in 2% of the trading days isn't a strategy with a proven edge — it's one good trade wearing a growth chart.
- Sample size
- How many closed trades the numbers above are actually based on. Twenty trades can produce an extraordinary-looking Sharpe ratio by chance; a hundred trades starts to mean something.
- In-sample vs out-of-sample performance
- The single strongest check there is. A strategy tuned against its own history will look better on that history than on anything that comes after — the size of that drop is a direct read on how much of the edge was real.
A five-minute checklist
Before subscribing to any signal, worth pulling up the strategy's own stats page and checking, in order: how many closed trades (sample size), what the Sharpe ratio and max drawdown actually are (not just the growth %), whether any single trade or short window accounts for most of the gains (concentration), and whether there's any out-of-sample period to compare against the strategy's original testing at all. If a seller can't or won't show that last one, that's itself informative.
This is exactly what StrategyAudit checks, automatically.
Upload a trade history and get all four numbers above scored against fixed, published thresholds — not a reliability percentage, a verdict.