infolol Editorial guide
How to read win-rate confidence intervals
A practical explanation of why the same displayed win rate carries different uncertainty at different sample sizes.
A single win-rate point does not show the range
A displayed 52% win rate means 52% of the observed games were wins. It does not promise exactly 52% in every future set of games. A different sample can produce a different result, especially when the sample is small.
infolol therefore shows a 95% confidence interval beside role win rate. The interval is a supporting measure of uncertainty under repeated sampling, not a prediction for one player's next match.
The difference between 100 and 1,000 games
One result moves a 100-game win rate by one percentage point. It moves a 1,000-game result by only 0.1 point. That is why the same win rate normally has a narrower interval with more games.
If two builds differ by one point while their intervals overlap widely, pick rate and item purpose are more useful than a strict ranking. A repeated difference across larger samples provides stronger comparative evidence.
An interval is not a verdict
The calculation assumes a sampling process that real match data cannot perfectly satisfy. Mastery, premades, time of day, and patch adaptation can all create bias outside the formula.
A narrow interval also does not make one build optimal for a specific player. Separate the stability of an aggregate observation from its fit for the match in front of you.
A reading order for detail pages
Check patch and collection date first, then role sample and confidence interval. Compare build-level samples afterward. When differences are small, the more common build is usually the safer starting point to investigate.
- Avoid strong rankings when intervals are wide.
- When intervals overlap, prioritize item function and composition.
- Current context may matter more than a narrow interval from an old patch.