Enter two ratings, pick a K-factor and a result, and see each player's win probability, rating change and new rating.
The formula
Two steps: work out what was expected, then pay the difference between that and what happened.
EA = 1 / (1 + 10(RB − RA) / 400)
R is a rating. E is the expected score, a number between 0 and 1: the share of the points a player should take, with a win counting 1 and a draw 0.5. With no draws it is the win probability. Only the gap between the two ratings matters, and both players' expected scores add up to 1.
R′A = RA + K × (SA − EA)
S is the actual score: 1 for a win, 0.5 for a draw, 0 for a loss. Do better than expected and the rating goes up; do worse and it goes down. This calculator rounds the change to a whole number and gives player B the exact opposite of player A's change, so no points appear or vanish.
Worked example
Player A is rated 1500 and player B 1700, a gap of 200 points. A's expected score is 0.24 and B's is 0.76: B should win about 76 times in 100.
A wins anyway. With K = 32, A gains 32 × (1 − 0.24), which rounds to 24 points, and B loses the same 24. The new ratings are 1524 and 1676.
Had B won as expected, B would gain only 8 points, for 1708 against 1492. The less likely a result, the more it is worth.
A draw is a good result for the lower-rated player: A gains 8 points and B loses 8.
K sets the size of every change. It is the most a single match can be worth: an even match pays the winner half of K, and beating someone you were almost certain to lose to pays nearly all of it.
A low K (10 to 16) keeps a ladder steady and suits players with many games behind them. A high K (40 and up) lets ratings find their level quickly, which suits new players and short events. FIDE uses 40 for new players, 20 for most and 10 for players who have reached 2400. Scoreboards starts every leaderboard at 32.
| K-factor | Lower-rated player wins | Higher-rated player wins | Draw |
|---|---|---|---|
| 16 | ±12 | ±4 | ±4 |
| 24 | ±18 | ±6 | ±6 |
| 32 | ±24 | ±8 | ±8 |
| 40 | ±30 | ±10 | ±10 |
More than two players
The same formula stretches to both. This is how Scoreboards does it.
Each team's ratings are averaged and the formula runs on the two averages. Everyone on a team gets the same change, added to their own rating. Example: 1400 and 1600 average 1500; if they beat a team averaging 1700 at K = 32, both gain 24 and both opponents lose 24. To work out a team match here, enter the two team averages as player A and player B.
In a free-for-all of 3 to 8 players, every pair of players is rated against each other by finishing position, and K is divided by the number of opponents, so one game moves a rating about as much as one match. Six equally rated players at K = 32 finish +16, +10, +3, −3, −10, −16, which adds up to exactly zero.
From the rating gap alone: 1 / (1 + 10^(gap / 400)) for the lower-rated player. A 100-point lead means an expected score of about 64%, 200 points about 76%, and 400 points about 91%, or ten to one. Strictly this is the expected score, which counts a draw as half a win.
32 is a sound default for a community ladder. Go down to 16 or 20 if ratings should move slowly over a long season, and up to 40 for new players or a short event. In chess, FIDE uses 40, 20 and 10 depending on the player's experience and rating.
Yes, unless both players have the same rating. A draw scores 0.5 for each side, so the lower-rated player did better than expected and gains points, and the higher-rated player loses the same number.
In this calculator and on Scoreboards, yes: both players use the same K, and the loser's change is the exact opposite of the winner's. Systems that give each player their own K, as FIDE does, can move the two ratings by different amounts.
Whatever the ladder decides. Only the difference between two ratings enters the formula, so the starting number is a convention. Scoreboards starts players at 1200 by default and lets you set any other number per leaderboard.
It gives the classic ELO result. Federations add rules of their own, such as different K-factors per player and rating periods, and Chess.com and Lichess use Glicko systems instead of plain ELO, so an official change can differ by a few points.
Yes. Scoreboards keeps an ELO leaderboard in your Discord server: players report a match on the leaderboard, the opponent confirms it, and both ratings update with this calculation. Team matches, free-for-alls and a matchmaking queue are built in.
Scoreboards is a Discord bot that keeps the ELO leaderboard for you. Players report a match, the opponent confirms, and every rating updates with the maths on this page.
Add to Discord