The Glicko system, created by Mark Glickman, improves on Elo by adding a rating deviation (RD): a measure of how uncertain a rating is. A player who has not played for months, or who has played few games, has a high RD, so the rating moves more after the next games. This calculator applies the Glicko-1 update for one rating period with up to three opponents and gives the new rating, the new RD and a 95 % interval.
How the calculation works
- q = ln 10 ÷ 400
- g(RD) = 1 ÷ √(1 + 3q²RD² ÷ π²), which reduces the weight of an opponent whose rating is uncertain
- Expected score E = 1 ÷ (1 + 10^(−g(RD_j)(r − r_j) ÷ 400))
- d² = 1 ÷ (q² × Σ g² × E × (1 − E))
- New rating r′ = r + q ÷ (1 ÷ RD² + 1 ÷ d²) × Σ g × (score − E)
- New RD = √(1 ÷ (1 ÷ RD² + 1 ÷ d²))
Example
This is the worked example from Glickman’s paper. A player rated 1500 with an RD of 200 plays three games: a win against a 1400 player (RD 30), a loss against a 1550 player (RD 100) and a loss against a 1700 player (RD 300). The expected scores are 63.9 %, 43.2 % and 30.3 %. The player scored 1.0 points and was expected to score 1.37. The new rating is 1,464.1, a change of -35.9, and the new RD is 151.4, so the 95 % interval runs from 1,167 to 1,761.
Why it matters
A single number cannot tell a veteran from a newcomer with the same rating. The RD does. After a good run, a newcomer’s rating can jump, while that of an established player moves slowly. Games against opponents whose own ratings are uncertain count less. Online chess and many video games use Glicko or its successor Glicko-2 for this reason.
Good to know
Glicko-1 assumes the games of a rating period happen at the same time, and the rating period should hold about 5 to 10 games per player. Between periods, the RD grows with inactivity: RD = √(RD² + c²t), where c is chosen so that an unrated RD of 350 is reached after a given time. This calculator does not apply that growth, so enter the RD after any inactivity adjustment. Ratings from different sites are not comparable. Glicko-2 adds a volatility term.
In practice
Use the tool to check a result from a rating site, to plan which opponents help you most, or to see how much an upset win counts. A win against a much stronger opponent moves the rating more than a win against a weaker one.