What the Glicko Rating System Is
The Glicko system is a skill-rating method developed by statistician Mark Glickman in the 1990s as a direct response to Elo’s biggest blind spot: Elo treats every rating as equally reliable. Glicko attaches an explicit uncertainty measure — the ratings deviation (RD) — to every rating.
Glicko-2 (2001) adds a third parameter, volatility, tracking how erratically a player performs over time. Major online chess platforms, including Chess.com and Lichess, use Glicko-family systems.
What the Glicko System Ranks
Glicko ranks competitors in head-to-head games — chess, board games, and online competitive games — on an Elo-like scale where each player’s state is a rating plus RD rather than a bare number.
Core Inputs Used by the Glicko System
- Game results within a rating period — wins, draws, losses against rated opponents
- Each player’s rating and RD — and every opponent’s rating and RD
- Time elapsed — RD grows automatically during inactivity
How Glicko Ratings Are Calculated (High-Level)
The update follows Elo’s expectation logic, filtered through uncertainty:
- High RD = fast movement — a new or returning player’s rating swings widely per game, because the system knows it knows little.
- Low RD = slow movement — an established player’s rating inches, because the evidence base is deep.
- Opponent RD matters too — beating a poorly measured opponent tells the system less than beating a precisely measured one, so gains shrink accordingly.
- Inactivity inflates RD — stop playing and your rating stays put but becomes less trusted, moving faster again when you return.
Glicko-2’s volatility parameter additionally tracks how consistently a player performs, refining how quickly their rating may change.
Conceptual model: Elo answers “what’s your strength?”; Glicko answers “what’s your strength, and how sure are we?” — then lets the second answer drive the first.
Key Parameters or Factors
- Rating — the strength estimate, Elo-comparable in scale
- Ratings deviation (RD) — the uncertainty; 350 is the standard initial value for new players, and implementations bound how low it can fall for active ones
- Volatility (Glicko-2) — expected fluctuation in a player’s true performance over time
- Rating period — updates are batch-computed over periods, not game-by-game
Update Frequency
Ratings are recalculated per rating period (implementations typically treat each game or day as a period in practice); RD decays upward with calendar time between games.
Known Limitations and Criticisms
- Complexity — three interacting parameters make outcomes harder for players to intuit than Elo
- Tuning burden — RD growth rates and volatility dynamics need careful per-game calibration
- Batch design — the original formulation assumes rating periods; real-time platforms adapt it
- Still outcome-only — like Elo, Glicko reads results, not in-game performance
Where the Glicko System Is Used
Glicko-family systems are used in:
- Online chess platforms — Chess.com uses a modified Glicko-1, Lichess uses Glicko-2
- Board-game and competitive gaming sites
- The Pokémon Showdown ladder and similar competitive ladders
- Academic benchmarking of rating-system performance
Summary
Glicko’s insight is that a rating without a confidence interval is only half an answer. By making uncertainty a first-class parameter — rising with absence, falling with evidence — Glicko turned Elo’s flat number into a living estimate, and set the template TrueSkill later carried into team sports.
References and Sources
- Glickman, M. E. The Glicko System and Glicko-2 System (method documentation).
- Wikipedia. Glicko rating system.
- Elo, A. E. The Rating of Chessplayers, Past and Present (background).