Glicko-2 Rating System for Chess
How the Glicko-2 rating system works for chess: rating deviation, rating periods, tournament results, and how it differs from Elo for organizers.
Contents
- What the Glicko-2 system measures
- How rating deviation shapes updates
- The rating period in tournament play
- Glicko-2 compared with Elo
- Setting up Glicko-2 for your group
- Handling tournament results and byes
- Software and calculation steps
- Choices that keep standings fair
- Questions about Glicko-2 chess ratings
- What is the Glicko-2 rating system in chess?
- How is Glicko-2 different from Elo for chess?
- What does rating deviation mean in chess ratings?
- How long is a rating period in Glicko-2?
- Do unrated players need a starting rating for Glicko-2?
- Can tournament results be entered all at once into Glicko-2?
The Glicko-2 rating system is a statistical method for calculating chess ratings that pairs each player's rating with a measure of how uncertain that rating is. Instead of treating every result as equally informative, Glicko-2 adjusts a rating based on how much is already known about the player. Tournament organizers can use it to keep standings fair when players compete unevenly.
- Glicko-2 stores a rating plus a rating deviation, which shows how confident the system is in that rating.
- Ratings update once per rating period, a block of games processed together rather than one game at a time.
- A player with a high rating deviation moves further after a result than a player with a low one.
- Glicko-2 rewards upsets more than expected wins, which keeps standings responsive to real results.
- Organizers must decide the rating period length and the starting deviation before the first round.
What the Glicko-2 system measures
Glicko-2 measures chess skill as a rating number plus a rating deviation, often shortened to RD. The rating estimates a player's strength on a continuous scale. The rating deviation expresses how unsure the system is about that estimate.
A new player enters with a wide rating deviation because the system has seen no results. A player with many recorded games carries a narrow deviation. Two players can share the same rating while holding very different deviations.
Glicko-2 also tracks a volatility value that describes how erratic a player's results have been. Volatility affects how far a rating can swing after a rating period. Mark Glickman designed the system as an extension of the earlier Glicko method.
How rating deviation shapes updates
Rating deviation in chess decides how much a single result can move a player's rating. A high deviation signals uncertainty, so the system allows a larger change. A low deviation signals confidence, so the system moves the rating less.
The effect matters for newcomers and returning players. Someone who has played few rated games can climb or fall quickly as results arrive. A long-established player needs repeated evidence before the rating shifts far.
Deviation also grows during inactivity in the full model, because the system loses track of a player's current form. That growth means a returning player's first results count for more. Organizers should record the gap between events so the calculation reflects it.
The rating period in tournament play
A rating period is the block of time over which games are collected and processed together. Glicko-2 updates ratings once per period rather than after each game. All results inside the period are treated as simultaneous evidence.
This design suits a tournament that runs over a weekend or a league that meets on a fixed schedule. The organizer sets the period length before play begins. Players and spectators then know when the new list takes effect.
Shorter rating periods give faster feedback but rest on fewer games per update. Longer periods smooth out noise but delay the visible effect of a strong result. The right length depends on how often your group meets.
Glicko-2 compared with Elo
Elo and Glicko-2 both convert results into rating changes, but they treat uncertainty differently. Elo assigns a fixed K-factor that sets how much a rating can move. Glicko-2 derives the size of the change from each player's rating deviation and volatility.
Elo treats a veteran and a newcomer with the same rating as equally settled. Glicko-2 separates them through the deviation. That difference matters in a small group where new players join between events.
Glicko-2 also accounts for the reliability of an opponent's rating. A win against a player with a wide deviation counts differently than a win against a settled one. Elo handles this less directly through the expected score.
| Feature | Elo | Glicko-2 |
|---|---|---|
| Uncertainty measure | Fixed K-factor | Rating deviation per player |
| Update timing | Game by game | Once per rating period |
| New player handling | Same K-factor as veterans | Wide deviation, larger moves |
| Result volatility | Not modeled | Tracked as a separate value |
Setting up Glicko-2 for your group
Start by choosing a starting rating and a starting rating deviation for every unrated player. The organizer assigns these values before the first event. The deviation should be wide enough to let early results shape the rating.
Next, fix the rating period and state it in the rules. Publish whether results are processed after each round, each event, or each month. Consistency keeps standings comparable across events.
Record every game with both players, the result, and the date. Tournament results feed the calculation as a set, so a missing game distorts the update. Keep a backup of the raw results before running the ratings.
Handling tournament results and byes
Tournament results enter the calculation as wins, losses, and draws for each player. A bye or an unplayed round should be left out rather than scored as a win. Treating a bye as a game inflates a rating with no real evidence.
Players who face the same opponents repeatedly generate correlated results. The system still processes them, but the standing reflects that schedule. Organizers can note scheduling patterns when they publish results.
A withdrawn player keeps the games already completed. Any unplayed rounds are excluded from that player's period. This rule keeps the rating tied to games actually played.
Software and calculation steps
Glicko-2 runs as a sequence of formulas that most organizers handle with software. The steps compute each player's expected score, then a new rating, deviation, and volatility. Manual computation for a whole field is error-prone and slow.
Free implementations exist in several programming languages, and some tournament managers include the system. The organizer should verify the version and test it on a small set of known results. A dry run on past events builds confidence before a live event.
The calculation needs a tuning constant that controls how quickly volatility adapts. Changing that constant changes how jumpy ratings become. Keep the chosen value fixed for the season and document it alongside the rules.
Choices that keep standings fair
Fair standings depend on consistent inputs more than on the formula itself. Every organizer should publish the starting values, the period length, and the handling of byes. Players then understand why their rating moved.
A wide starting deviation lets new players find their level quickly. A narrow one protects established ratings but slows adjustment. The choice depends on how much turnover your group sees.
Transparency about volatility settings and period length prevents disputes later. A short written policy covering these settings is enough for a recreational league. Review the policy between seasons rather than mid-event.
Questions about Glicko-2 chess ratings
What is the Glicko-2 rating system in chess?
Glicko-2 is a rating method that estimates chess skill while also tracking how certain that estimate is. It updates each player's rating once per rating period using all games in that block. The method was designed by Mark Glickman as an extension of Glicko.
How is Glicko-2 different from Elo for chess?
Glicko-2 differs from Elo by attaching a rating deviation to each player instead of using a fixed K-factor. Elo updates game by game, while Glicko-2 processes a whole rating period at once. Glicko-2 also models each player's result volatility separately.
What does rating deviation mean in chess ratings?
Rating deviation measures how unsure the system is about a player's rating. A wide deviation means the rating can move further after new results. A narrow deviation means the rating is well established and moves less.
How long is a rating period in Glicko-2?
A rating period has no fixed length and depends on your schedule. A weekly club might process ratings monthly, while a tournament might use the whole event as one period. The organizer sets the length and publishes it in the rules.
Do unrated players need a starting rating for Glicko-2?
Yes, every unrated player needs a starting rating and a starting rating deviation before play. The organizer assigns both values and states them in the rules. A wide starting deviation lets early results shape the rating.
Can tournament results be entered all at once into Glicko-2?
Yes, Glicko-2 is built to process tournament results as a batch for each rating period. All games in the period are treated as evidence gathered at the same time. Byes and unplayed rounds should be excluded from the calculation.