How the ranking works

Every score on Game Battle is calculated from real votes. Here is exactly how.

Why not just count votes?

Because raw vote totals measure how popular a *battle* was, not how good a *game* is. One matchup going viral would pin both of its games to the top of the leaderboard forever, regardless of whether people actually preferred them.

The four ideas behind the score

1. Vote share, not vote count

A game that wins 900 to 100 has done something more impressive than one that wins 5,000 to 4,900. We score the proportion, not the pile.

2. Statistical confidence

A game with a perfect 3–0 record has not proven much. We use the lower bound of the Wilson score interval, which discounts small samples. A game needs a meaningful number of votes before it can outrank one with a long record.

3. A cap on any single battle

No individual battle may contribute more than a set share of a game's total evidence. This is the specific mechanism that stops one popular matchup from deciding the board.

4. Opponent variety

Beating eight different games says more than beating the same game eight times. Games only ever compared against one rival are discounted.

Recency

Votes lose influence over time on a configurable half-life, so the board reflects what the community thinks now rather than what it thought two years ago.

The threshold

Games below a minimum vote count are shown as "needs more votes" rather than being given a position they have not earned.

What never affects the score

• Sponsorship or any form of payment.
• Featuring a battle on the homepage.
• Staff opinion.

Community ranking equals community votes. That is the whole product.