Elo Rating System
Also known as: Elo Rating System, Elo Chess Rating, Elo Skill Rating, Elo Derecelendirme Sistemi
The Elo Rating System is a pairwise comparison-based ranking method developed by Hungarian-American physicist and chess master Arpad Elo and formally published in 1978. Originally designed to assess the relative skill levels of chess players, it assigns each competitor a numerical rating that rises or falls after each encounter based on the expected versus actual outcome. The system assumes that player performance follows a logistic distribution, enabling probabilistic predictions of match results and continuous rating refinement over time.
Read the full method
Sign in with a free account to read this section.
Method map
The neighbourhood of related methods — select a node to explore.
When to use it
Use the Elo Rating System when you have a stream of pairwise comparison data — contests, head-to-head evaluations, or preference judgments — and need a simple, interpretable ranking that updates incrementally. Key assumptions include that outcomes are independent, performance is stationary over time, and the logistic model adequately captures winning probability. Elo is less suited when competitors play very few games (high variance), when non-transitivity exists (A beats B, B beats C, C beats A cyclically), or when margin of victory is informative. Alternatives such as Glicko or TrueSkill address rating uncertainty and team-based settings.
Strengths & limitations
- Computationally simple: requires only basic arithmetic and updates in constant time per match.
- Self-correcting: consistent over- or under-performance automatically adjusts ratings toward true skill.
- Interpretable: rating differences directly translate into win probabilities via the logistic formula.
- Domain-agnostic: applicable to chess, sports, games, crowdsourcing preference tasks, and any pairwise comparison setting.
- Ignores rating uncertainty: all players at the same rating are treated identically regardless of how many games support that estimate.
- Sensitive to K-factor choice: too high causes volatility; too low makes the system slow to reflect genuine skill changes.
- Assumes stationarity: players whose skill is rapidly improving or declining are poorly tracked until many games have elapsed.
- Vulnerable to sparse data: players with few games have unreliable ratings, and new entrants require provisional handling.
Frequently asked
Why is the divisor 400 in the expected score formula?
The 400-point constant is a scaling choice made by Elo so that a rating difference of 400 points corresponds to an approximately 91% win probability for the higher-rated player. It has no deeper mathematical necessity; alternative implementations use 173.7 (for a natural-log scale) or other constants. The choice of 400 was retained for historical continuity and ease of interpretation.
How does Elo differ from the Glicko system?
Glicko extends Elo by introducing a rating deviation (RD) that quantifies uncertainty around each player's true rating. When a player competes infrequently, RD grows, reflecting lower confidence. Updates then scale with RD: high-uncertainty players change ratings more per game, while well-established players change less. Elo has no such uncertainty term and treats all ratings at the same level of confidence.
Can Elo handle draws, and does it work for non-chess contests?
Yes. Draws are accommodated by assigning S_A = 0.5, placing the actual outcome exactly between a win and a loss. The formula is entirely general: any pairwise comparison with binary or ternary outcomes — sports matches, preference tests, or A/B evaluations — can use Elo without modification, provided the logistic performance model is a reasonable approximation.
Sources
- Elo, A. E. (1978). The Rating of Chessplayers, Past and Present. Arco Publishing. ISBN: 978-0-668-04721-0
How to cite this page
ScholarGate. (2026, June 2). Elo Rating System. ScholarGate. https://scholargate.app/en/decision-making/elo-rating
Which method?
Set this method beside its closest kin and read them side by side — the library lays the books on the table; the choice is yours.
- Bradley-Terry ModelDecision-making↔ compare
- Plackett-Luce ModelDecision-making↔ compare