Skill-based matchmaking, commonly called SBMM, is a system that uses an estimate of each player’s ability to create online multiplayer matches. Its goal is usually to reduce extreme skill differences and give players or teams reasonably similar predicted chances of winning.
However, matchmaking does not consider skill alone. It may also account for server location, connection quality, game mode, platform, party size, preferred role and waiting time.
What Is MMR?
Most SBMM systems use a hidden value known as matchmaking rating, or MMR. This rating estimates how likely a player is to perform well against other rated players.
MMR is different from visible rank. A player’s displayed badge or division may include seasonal resets, promotion rules or progress points, while hidden MMR may remain more stable.
The matchmaker uses MMR and other requirements to decide which players can enter the same lobby. This is why players with different visible ranks may sometimes appear in the same match.
How Matchmaking Works
When a player enters a queue, the game may record information such as:
- Skill estimate.
- Server region.
- Connection quality.
- Platform.
- Party size.
- Preferred role.
- Time already spent waiting.
The system first removes players who are not compatible because of different modes, platforms, regions or team requirements.
It then compares possible groups based on skill differences, predicted win probabilities, connection quality and team composition.
The search normally begins with strict criteria. If no suitable match is found, the system may gradually allow a wider MMR range, broader region or greater variation between parties.
This creates a trade-off. Strict matchmaking can produce closer skill levels but longer waiting times. Faster matchmaking may reduce waiting but create wider skill differences.
How Skill Ratings Change
After a match, the rating system compares the result with what it expected.
Defeating a stronger opponent may produce a larger rating increase than defeating a weaker opponent. Losing to a lower-rated opponent may cause a larger decrease.
Some systems also consider rating uncertainty. A new player’s rating may change quickly because the system has little information. An established player’s rating usually changes more slowly because it is based on more matches.
Games may use different rating models, including Elo-like systems, Glicko or TrueSkill. Some rely mainly on wins and losses, while others may consider placement, objectives, assists, roles or individual performance.
Why Balanced Matches Can Feel Unfair
A match with equal predicted chances can still end in a one-sided result. Ratings are estimates rather than guarantees.
Unexpected outcomes can happen because:
- A player performs unusually well or poorly.
- One team communicates more effectively.
- A player disconnects.
- Team roles do not work well together.
- A map favors one team’s experience.
- The rating system has incomplete information.
Two teams with a predicted 50% chance of winning are not guaranteed to finish with a close score. The prediction only means that neither team is strongly favored before the match begins.
New Players, Parties and Smurf Accounts
New accounts are difficult to place because the system has limited performance history. Genuine beginners, experienced players using new accounts and players transferring from another platform may initially receive inaccurate ratings.
Premade parties also complicate matchmaking. A coordinated group may perform better than individual ratings suggest because its members communicate and practise together.
Some systems therefore try to match parties against groups of similar size or adjust the party’s estimated strength.
Does SBMM Force a 50% Win Rate?
SBMM may try to create matches with similar predicted win chances, but this does not mean it deliberately alternates wins and losses.
A player whose rating accurately reflects their ability may move toward an average win rate over time because they regularly face similarly skilled opponents. Winning and losing streaks can still occur naturally.
Conclusion
Skill-based matchmaking combines a player’s estimated ability with practical factors such as latency, server region, party size, roles and queue time.
It cannot create perfectly balanced matches because ratings are uncertain, player performance changes and the available population is limited. SBMM is best understood as a system that searches for a reasonable compromise between competitive balance, connection quality and waiting time.
Frequently Asked Questions
No. Skill-based matchmaking selects players using an estimate of ability, while ranked matchmaking normally includes a visible progression structure. Casual and unranked modes can also use hidden skill estimates without displaying a formal rank.
MMR is calculated from match evidence according to a game-specific model. Common inputs include wins, losses, draws, opponent strength and rating uncertainty. Some systems may also use placement or performance information. Glicko and TrueSkill demonstrate how uncertainty can be included alongside the central skill estimate.
No. It may try to construct matches with approximately even predicted chances once a player’s rating stabilizes, but it does not need to predetermine or alternate outcomes. Improvement, team variation, streaks and incomplete ratings can move the recorded rate above or below 50%.
The displayed rank may not be the value used directly for matchmaking. Seasonal resets, progression points, rank protection and placement rules can separate visible progress from hidden MMR. Party constraints and widened search ranges can create additional visible differences.
Fewer suitable opponents are usually available at the extremes of the rating distribution. The matchmaker must find those players in the same mode, region and time window while preserving acceptable connections. Longer searches can be used to avoid widening the skill range too quickly.
The treatment varies by game. A system may use the party’s average, highest rating, weighted rating or complete skill distribution. It may also seek another premade group because coordinated parties can perform differently from solo players.
The rating system begins with limited information. New-player ratings carry greater uncertainty and may change rapidly. Experienced players on new accounts and beginners starting at the same provisional value make the cold-start period especially difficult.
Both can matter, but their weighting varies. A matchmaker must keep latency acceptable while finding similarly skilled players. A strict skill match on a distant server may provide worse gameplay than a slightly broader match on an appropriate server, so systems balance the two rather than applying a universal priority.
