ELO Matchlog

How each game is rated

• Winner-advance formula: change = K(playercount) x (actual - expected) x (setting_weight / 21).
• K-factors by lobby size: 3p=8, 4p=12, 5p=15, 6p=18.
• Expected values come from each player's pre-game ELO relative to the whole lobby.
• Tournament games need 4+ players and a recorded winner, and enter ELO after the vote gates: event average above 4, and at least 5 non-skip setting votes averaging above 2. Games that just crossed the gates are picked up by the next ELO run (every two hours, at :15).
• ELO Lounge games (weight 1-21) need 3+ players and are rated at the next ELO run, no votes required.

How the setting weight works

The setting weight is the setting_weight / 21 factor in the formula above. It starts from a community score and is then clamped by the strength of the lobby, so weak lobbies can never produce big ELO swings, no matter how highly the settings are rated.

1. Starting weight (0-21)

• Tournament games: the average of the setting votes cast for that exact game on the profile page's Votes tab. Votes use the scale 1 (should not count towards ELO), 7 (acceptable), 14 (just average) and 21 (good settings for ELO). Skip votes are ignored. A game needs at least 5 non-skip votes with an average above 2 before it is rated at all, and the average is recomputed on every ELO run, so late votes keep adjusting the weight of already rated games.
• ELO Lounge games: the fixed weight (1-21) the host picked when the lounge game was created, shown in the Event column below. No voting.

2. Clamp by lobby strength

The engine measures the lobby by the trimmed average of the players' pre-game ELO: the highest and lowest ELO in the lobby are dropped and the rest is averaged (3-player lobbies use the median). This is close to, but not identical to, the AVG Elo column in the table. The starting weight is then multiplied by a factor and capped, both depending on that trimmed average:

Example: settings voted 21 in a lobby with a trimmed average of 1012 give 21 x 0.74 = 15.5, capped to 7, so the game counts for 7/21 of a full-weight game. Only lobbies averaging 1055+ get the full weight.

3. Lounge reduction

After the clamp, ELO Lounge games are multiplied by a further 0.75, so a lounge game is never worth more than 75% of a comparable tournament game.

The final weight actually used for a game is shown as Setting weight used in the ELO details card, and is stored per player in the matchlog. The clamp is the reason two games with identical votes can move ELO by very different amounts.