How Lunar ranks heroes
Win rate is the most quoted stat in hero shooters and one of the most misleading. A hero you've played twice and won twice has a 100% win rate, which says almost nothing. This guide explains how Lunar avoids that trap.
The problem with raw win rate
Imagine three heroes in your history:
| Hero | Record | Raw win rate |
|---|---|---|
| A | 2 wins, 0 losses | 100% |
| B | 9 wins, 6 losses | 60% |
| C | 40 wins, 25 losses | 62% |
Sorted by raw win rate, hero A comes first. But two games is far too few to know anything. You might be a 30% player on that hero who got lucky twice. Hero C, with 65 games, is the one you can actually rely on.
The Wilson lower bound
Lunar ranks small samples by the Wilson lower bound of the win rate. In plain terms, it asks: given this record, what's the lowest win rate we can be reasonably confident you have? With few games the answer is cautious. With many games it moves close to the raw percentage.
For the example above, hero A's lower bound is low (two games prove little), hero B's sits somewhere in the middle, and hero C's is the highest of the three. So the ranking follows the evidence, not the luck.
Top heroes vs most-played heroes
/career shows two separate lists:
- Most played simply sorts by playtime.
- Top heroes combines the Wilson-adjusted win rate with playtime. Lunar takes the geometric mean of the two, and scales playtime logarithmically so one hero with a huge number of hours doesn't drown out everything else. A hero with no wins scores zero.
The result is a list of heroes you both play enough and win on, which is usually what "best heroes" should mean.
Markers you'll see
- Every percentage shows its sample size, for example "58% (24 games)".
- A † marks entries with few games. Treat them as early hints, not conclusions.
- Preliminary labels appear on pairings and matchups below the minimum game count. They are ranked below fully supported entries.
Detailed playtime, not just the hero you started on
Players swap heroes mid-match. Wherever the match data has per-hero playtime, Lunar uses it, so a match where you played 8 minutes of one hero and 2 of another counts mostly toward the first. Preferred roles are weighted the same way.
Why this matters for you
Stats that ignore sample size push you toward the wrong conclusions: "I should main hero A, I never lose on it." Lunar would rather tell you "not enough games yet" than hand you a confident-looking number that's mostly noise. The more you play and the more history Lunar has, the sharper every ranking gets.