Edge or noise?
A 75% win rate on 12 trades tells you almost nothing. Drag the wins and losses and watch how much — or how little — your sample can actually prove.
The blue→gold band is the Wilson 95% confidence interval — the range of true win rates your sample is compatible with. If it still straddles 50%, your history can't yet tell a real edge from a coin flip, however pretty the headline number looks. This is descriptive — it measures what your past sample can prove, not what your next trade will do.
Why a win rate on its own means almost nothing
A win rate is a single number squeezed out of a sample of trades — and a small sample is easy to fool. Flip a fair coin twelve times and you will land on nine heads roughly one time in twenty. Nothing about that coin changed; you just caught a lucky stretch. A trading strategy is no different: a 75% win rate over 12 trades is exactly what a coin flip looks like on a good run.
The honest question is not "what was my win rate?" but "given this sample, what range of true win rates is compatible with what I saw?" That range is a confidence interval, and it does the one thing a headline percentage can't: it shrinks as your evidence grows.
What the Wilson interval tells you
The blue-to-gold band in the lab is the Wilson 95% confidence interval — the standard, well-behaved way to put error bars on a proportion, even when the sample is small or the rate is near 0% or 100%. Read it like this:
- The whole band sits above 50% — even the pessimistic end beats a coin flip. That is real evidence of an edge.
- The band still straddles 50% — your sample cannot yet tell a genuine edge from luck, however high the headline number looks.
- The whole band sits below 50% — a real pattern, just the losing kind.
Try it: 75% on 12 trades is inconclusive — the band still crosses 50%. The same 75% on 24 trades clears it. Identical win rate, opposite conclusion. Sample size is the whole story.
Why this matters for reading an indicator
Every signal statistic you will ever see — a backtest win rate, a "hit rate", a strategy's advertised accuracy — is a sample estimate with an interval hiding behind it. Probalist reports those intervals instead of a lone flattering number, because a claim you can bound is a claim you can trust. This lab is descriptive: it measures what a past sample can prove, never what your next trade will do.