MMR, short for matchmaking rating, is a numeric estimate of a player’s or team’s skill that a competitive online game uses to create balanced matches. After every match, the rating rises or falls depending on the result and on how strong the opponents were expected to be. In esports and ranked play, MMR is the statistical engine behind matchmaking and ladders, even when players see only a medal, a division or a league name.
Most MMR systems descend from the Elo rating system developed for chess, extended with ideas from later models such as Glicko and TrueSkill. Their shared goal is to predict match outcomes accurately, so that players face opponents they have a roughly even chance of beating.
How MMR is calculated
An MMR system first computes an expected result for each side based on the rating difference. In the classic Elo formula, a player rated 400 points above an opponent is expected to win about 91% of the time, while equal ratings give 50%. After the match, the rating changes in proportion to the gap between the actual and the expected result.
The size of each change is controlled by a factor often called the K-factor. Beating a stronger opponent yields a large gain, while beating a much weaker one yields very little. Losing to a weaker opponent costs more than losing to a stronger one. Over many matches, a player’s rating converges toward their true skill level.
| Rating difference | Expected win chance of stronger side | Gain for an upset win |
|---|---|---|
| 0 points | 50% | Moderate |
| 100 points | About 64% | Somewhat larger than average |
| 200 points | About 76% | Large |
| 400 points | About 91% | Very large |
These figures follow from the standard Elo curve with a 400-point scale. Games may use different scales and constants, so absolute MMR numbers are not comparable between titles.
Rating models behind matchmaking rating
Modern games usually extend Elo by tracking how certain the system is about each rating. A new player has high uncertainty, so their rating moves quickly; a veteran with hundreds of matches has low uncertainty, so single results change little. This is why the first matches, often called placement matches, have such a strong effect.
| Model | Key idea | Typical use |
|---|---|---|
| Elo | Single number updated by expected versus actual result | Chess, simple one-versus-one ladders |
| Glicko | Adds a rating deviation for uncertainty | Online chess, many competitive games |
| Glicko-2 | Adds volatility for erratic performance | Games with varied activity patterns |
| TrueSkill | Skill mean and uncertainty for team games | Team-based and free-for-all matches |
| Proprietary hybrids | Custom blends with performance data | Large commercial esports titles |
Team games add complexity, because one result must be shared among several players. Some systems also include individual performance statistics, although this can reward playing for numbers rather than for the team’s victory, so many designers weigh it carefully or avoid it.
MMR compared with visible rank
Many games separate the hidden MMR used for matchmaking from the visible rank shown to players, such as bronze, silver, gold or a league with points. The visible rank is designed to be motivating and understandable, while MMR remains a precise statistical value.
| Aspect | Hidden MMR | Visible rank |
|---|---|---|
| Purpose | Predict outcomes and build fair matches | Show progress and status |
| Form | Precise number with uncertainty | Tiers, divisions or points |
| Change per match | Depends on expected result | Often smoothed or fixed steps |
| Visibility | Often hidden or shown only partly | Always shown in the ranked mode |
| Seasonal reset | Usually kept, sometimes softened | Often reset or lowered |
When MMR is higher than the visible rank, a player typically gains more points per win than they lose per defeat until the rank catches up. This mechanism explains many apparently uneven point changes in a ranked mode.
MMR in esports and competitive play
In organized esports, MMR and similar ratings serve as a pipeline for talent. Top ladder positions attract attention from teams, and some qualifiers or open tournaments use ranked standing as an entry requirement. Third-party ranking systems for professional teams often use comparable Elo-based methods to seed tournaments and compare teams across regions.
- Matchmaking queues use MMR to group players and keep waiting times acceptable.
- Leaderboards display the highest-rated players in each region.
- Scrim partners are often chosen by comparable skill ratings.
- Tournament seeding places stronger teams apart in early rounds.
Matchmaking always balances match quality against queue time. At the very top of a ladder, few players have similar ratings, so the system must accept larger rating gaps or longer waits.
Common problems and misconceptions
Because MMR shapes every ranked match, players frequently attribute losing streaks to the system itself. In reality, a correctly working system makes long-term win rates move toward 50%, which is a sign of balanced matchmaking rather than unfairness.
| Problem | What happens | Typical countermeasure |
|---|---|---|
| Smurfing | Skilled players use new accounts to face weaker opponents | Faster rating adjustment, phone verification |
| Boosting | Another person plays to raise an account’s rating | Detection systems and bans |
| Rating inflation | Average ratings drift upward over time | Recalibration and seasonal adjustments |
| Belief in rating traps | Players think they cannot climb | Statistics show skilled players rise over time |
In simple terms
MMR is a hidden skill score that a game gives you. When you win, it goes up; when you lose, it goes down. Beating a strong opponent raises it considerably, while beating a weak one raises it only a little.
The game uses this score to find opponents who are about as good as you are. The aim is that every match feels fair and could go either way, which makes competitive play more exciting.
A practical example
A new player completes ten placement matches in a five-versus-five competitive game, winning seven. Because the system is still uncertain about the player’s skill, each result moves the rating strongly, and the player is placed in a mid-level tier.
Over the next fifty matches, the rating keeps rising while the visible rank lags behind, so wins grant about 25 points and losses cost about 15. Eventually the visible rank catches up and the win rate settles near 50%, signaling that the player now faces opponents of matching skill.
Once the player reaches the upper part of the regional ladder, that position helps a local esports team notice them and invite them to weekly scrims.