Updated 2026-07-25
The Wilson score interval is a confidence interval for a proportion, such as a hit rate or a precision figure. It exists because the textbook normal approximation misbehaves exactly where it matters most: small samples, and proportions close to zero or one. In those regions the simple interval can extend below zero or above one, or be far too narrow. The Wilson interval stays inside the valid range and keeps its coverage.
Why this belongs in a trading glossary: any measured rate computed from a limited number of observations is an estimate, and the interval is what tells you how much the estimate could move with more data. A rate with a wide interval and a small sample is not evidence of skill. This is why we publish the sample size and the lower bound with every measurement rather than a bare figure, and why a candidate has to clear its lower bound before it counts. Method and current numbers: proof page.