Glicko

Glicko is a probabilistic rating system used to estimate players’ relative skill in two-player or small-team competitive games. Developed by Mark E. Glickman, it extends the traditional Elo framework by explicitly modeling uncertainty in a player's rating and how that uncertainty evolves over time. Each player has a rating and a rating deviation (RD) that expresses confidence in that rating; a larger RD indicates greater uncertainty.

The core idea is that a player’s true strength is a latent value and game results are noisy observations of that strength. After a set of games, the system updates the player’s rating to better reflect observed outcomes, while adjusting the RD to reflect the certainty of the new rating. Inactivity increases RD, making the rating less certain until the player competes again. This mechanism allows ratings to adapt to recent performance and to degrade gracefully when a player is inactive.

Glicko-2 is an extended version that adds a volatility parameter. This parameter captures how much a player’s rating is expected to change over time, even with no new results. Glicko-2 also uses a different scale and transformation to improve numerical stability and forecasting accuracy. Many online platforms implement Glicko-2 rather than the original Glicko, often converting ratings to familiar Elo-like figures for display.

Compared with Elo, Glicko and its variants provide a more explicit account of uncertainty and time-based drift, potentially offering faster and more accurate reflection of a player's current level. They are widely used in online chess, esports, and other competitive games to rank players and pair opponents.