Glicko rating system

Glicko

Glicko is a family of rating systems devised by statistician Mark Glickman to measure player strength while also tracking the uncertainty of that estimate. Unlike the classic Elo system, Glicko attaches a second number—Rating Deviation (RD)—to each rating, and (in the widely used Glicko-2 variant) also tracks a player’s volatility. These additions let ratings move quickly when confidence is low and stabilize when confidence is high, producing responsive and robust ratings across time controls and activity levels.

Definition

At its core, Glicko models a player with two (or three) parameters:

• Rating (R): Your estimated playing strength, in rating points.
• Rating Deviation (RD): A measure of uncertainty in R. Smaller RD means the system is more confident about your strength; larger RD means less confidence.
• Volatility (σ): Glicko-2 adds this parameter to capture how erratic a player’s performance tends to be. Higher volatility allows bigger rating swings after each game.

Glicko (mid-1990s) introduced rating plus RD; Glicko-2 (1999) refined the math by adding volatility and a scaling that improves stability in practice.

How it is used in chess

Many online chess servers use Glicko or Glicko-2 to update ratings after each game in Blitz, Rapid, Bullet, and other pools. Broadly:

• After each result (win/draw/loss), your rating changes based on opponent strength, your RD, and (in Glicko-2) your volatility.
• Inactivity increases RD: If you don’t play for a while, RD rises, signaling less certainty. When you return, results move your rating more until RD settles again.
• Frequent play reduces RD: Playing regularly shrinks RD, so rating changes become smaller and more stable.
• Provisional phase: New accounts commonly start with a high RD; early results can cause large rating swings while the system “finds” your level.
• Separate pools: Each time control (e.g., Blitz vs Rapid) typically has its own Glicko parameters, reflecting different playing strength across formats.

Strategic implications for players

• Expect bigger swings after breaks: If you’ve been inactive, your RD is higher, so a streak will shift your rating more than usual.
• Stabilize by playing regularly: Consistent play drives RD down, making your rating a steadier indicator.
• Facing uncertain opponents: Playing someone with high RD gives the system more license to move both players’ ratings, especially after decisive results.
• Improving quickly? High RD helps the system “catch up” to your true strength faster than Elo typically would.
• Time-control specialization: Your Glicko Blitz rating may stabilize differently than your Rapid rating depending on how often you play each.

Historical background

Mark Glickman developed Glicko to address known limitations of Elo, chiefly the fixed K-factor and lack of an uncertainty measure. Glicko appeared in the mid-1990s, with Glicko-2 published in 1999. Over-the-board federations such as FIDE primarily use Elo-based systems, while many online platforms adopted Glicko or Glicko-2 because they handle fast-moving, high-volume game data and intermittent activity more gracefully.

Examples

The numbers below are illustrative; actual changes depend on server settings and your current parameters.

• High-RD newcomer climbs quickly: Player A starts at R=1500, RD=200. They beat a stable opponent B at R=1700, RD=50. Because A’s RD is large and the upset is informative, A might gain on the order of +60 to +100 points in one game, while B loses noticeably less (their RD is small, so the system was already confident).
• Established player, small adjustment: Player C at R=2000, RD=40 draws D at R=2100, RD=35. Both RDs are low; the draw is mildly favorable to C. Ratings adjust only slightly (C +3, D −3 is a typical scale).
• Inactivity and volatile return: Player E at R=1800, RD=45 doesn’t play for months; RD drifts to 90. Upon returning, a 3-game winning streak versus ~1800 opposition could yield a large jump (e.g., +50 to +80 total) before RD contracts again.

Interesting facts and anecdotes

• Confidence band: A handy interpretation is that your true strength is likely within roughly R ± 2×RD. So if you’re 1900 with RD=50, think “about 1800–2000.”
• No fixed K-factor: Glicko’s RD and volatility play the role of a dynamic K, letting the system adapt automatically to your activity and consistency.
• Cross-discipline impact: Beyond chess, Glicko variants are used in esports, board games, and other competitive rankings where activity and uncertainty vary widely.
• Streamer swings: It’s common to see large rating swings when popular players return from breaks—classic Glicko behavior as high RD meets a burst of results.

Common misconceptions

• “I was punished for being inactive.” RD growth doesn’t change your rating by itself; it just increases how much your rating can move once you play again.
• “Draws don’t matter.” Draws can move ratings noticeably when the draw is unexpected (e.g., a much lower-rated player drawing a higher-rated one), especially with high RD.
• “All platforms use the same math.” Implementations differ: some use Glicko, others Glicko-2, with different scaling constants and update frequencies.

Related terms and further exploration

• Compare with Elo and Glicko-2.
• See also: Rating Deviation (RD), Volatility, Provisional rating, Rating floor.
• Your own trends: • Peak: