Elo rating

What is an Elo rating?

An Elo rating is a skill-rating system that estimates competitors’ relative strength from match results. It predicts the likely result from the gap between two ratings, then adjusts both ratings based on what happened.

Named after chess player and professor Árpád Élő, the system was developed for chess and is now widely adapted. “Elo” often means a player’s rating or skill bracket. It is not an absolute score or permanent verdict, but a moving estimate of how likely one competitor is to outperform another in a particular player pool.

A rating only has meaning in context. A 1000 rating may represent a beginner, average player, or strong player depending on the scale, starting point, rules, and population.

How the Elo rating system works

Elo uses a repeated cycle of prediction and correction. It learns most from surprising results, such as an underdog beating a stronger opponent.

1. Give each competitor a starting rating, often a common baseline.
2. Compare ratings to calculate each competitor’s expected score.
3. Record the actual result: usually 1 for a win, 0.5 for a draw, and 0 for a loss.
4. Find the difference between actual and expected score.
5. Raise or lower ratings by an amount based on that difference.
6. Repeat over many matches so ratings reflect results across the pool.

If a favorite wins as predicted, ratings change little. If an underdog wins, the change is larger. In a simple two-player system with matching settings, one player’s gain often equals the other’s loss. That is common, but not universal in Elo-inspired systems.

The Elo formula and its key components

The standard update rule is Rnew = Rold + K × (S − E). In plain language, start with the old rating and add or subtract points according to how surprising the result was.

R is the rating, E is the expected score, and S is the actual score. A draw normally gives each side 0.5. Expected score is calculated from the rating difference, often with a logistic formula. K, the K-factor, controls how quickly ratings move: a higher K reacts faster and a lower K is steadier.

The rating-difference scale is a convention. In many classic Elo systems, a 400-point difference gives the higher-rated side an expected score of about 0.91. Platforms may use different scales, formulas, or adjustment rules. Anyone building a ladder through a software development kit or custom service should document those choices.

A practical Elo rating example

Imagine Player A is rated 1400 and Player B is rated 1600. B is favored because of the 200-point gap. If B wins, the result is close to expectation, so both ratings move modestly.

If A wins, it is an upset. A gains more points and B loses more because the result was unexpected. If they draw, A may gain points and B may lose points because a draw is relatively better for A. The exact change depends on the K-factor and expected-score formula.

What Elo numbers mean, and what they do not mean

A rating is most useful when compared with others playing under similar conditions. Numbers alone cannot reliably compare skill across unrelated systems.

Where Elo ratings are used

Elo is useful wherever repeated results reveal comparative strength. Most real systems modify the original model.

• Chess uses ratings for pairings, titles, seeding, and competitive progress.
• Online games and esports use Elo-style matchmaking and ranked ladders. Teams using application development tools may keep ratings behind the scenes.
• Team sports rate teams, sometimes adding home-field advantage and margin-of-victory rules.
• Combat sports and head-to-head leagues can update competitors despite incomplete schedules.
• Forecasting models use rating gaps with factors such as injuries, lineups, venue, or recent performance.
• Multimode games often keep separate ratings for solo, team, fast, and slow modes.

Benefits of an Elo rating system

Elo combines a clear concept with practical benefits. It can work even when not everyone plays everyone else.

• Opponent-sensitive updates reward wins over stronger opponents more.
• Continuous updates reflect changing performance without waiting for a season to end.
• Simple ratings support understandable standings and matchmaking.
• Expected scores help predict balanced contests.
• It supports open ladders and irregular schedules better than a complete round-robin table.
• It can be adapted for draws, teams, stakes, and provisional players.

Limitations and common pitfalls

An Elo rating estimates skill from available results. Poor design or data can make it misleading.

• Small samples can create volatile ratings.
• Inflation or deflation can arise when players enter or leave, or point flows are poorly controlled.
• Smurfing, alternate accounts, deliberate losses, and collusion can distort rankings.
• Inactive ratings may not reflect current skill without careful decay, re-entry, or uncertainty rules.
• Rapid improvement, patches, rule changes, and roster changes can outpace a slow rating.
• Team results do not show each teammate’s contribution.
• Groups that rarely play each other are difficult to compare.
• Not every Elo-inspired system is zero-sum. Bonuses and other adjustments can add or remove points.

Elo vs. Glicko, TrueSkill, and ranking points

Classic Elo is not the only option. The best choice depends on whether uncertainty, multiplayer teams, or event results matter most.

Designing or evaluating an Elo-based ladder

A credible ladder needs rules that fit the activity, not just a familiar formula.

• Define the pool by region, mode, time control, and eligibility.
• Choose an initial rating and label new ratings as provisional where appropriate.
• Tune K-factor by experience and match frequency.
• Handle draws, forfeits, disconnects, remakes, and abuse consistently.
• Keep incompatible modes separate.
• Monitor calibration by checking whether predicted win rates match actual results over time.
• Publish rules, update schedules, and appeal processes.

Why ratings change over time

Ratings move because performance, opponents, and the system change. New competitors often receive larger updates because there is less evidence about their skill. A high K-factor creates bigger swings, while a low K-factor favors stability but can lag behind real improvement.

Some systems apply inactivity decay, while others keep the rating unchanged but treat it as less certain on return. Population changes can also shift the scale. A player may improve while their visible Elo stays flat if opponents improve at the same pace. Elo is therefore most useful as a comparative signal over a defined period, not as a permanent label or complete account of performance.