Skill-Based Matchmaking (SBMM)
Skill-based matchmaking (SBMM) is a system that groups players into matches based on their measured ability — trying to create games where everyone has a reasonable chance rather than placing new players against veterans.
SBMM is simultaneously one of the most effective tools for player retention and one of the most debated design decisions in modern multiplayer games. Done well, it keeps matches competitive. Done poorly — or communicated poorly — it creates the perception that the system is punishing good players.
How SBMM works
Every player is assigned a skill rating — usually derived from win/loss record, performance statistics, or both. When a player enters the matchmaking queue, the system searches for other players whose ratings are close enough to form a balanced match.
The matchmaker balances two competing goals: match quality (how closely rated the players are) and queue time (how long players wait). A tight skill window produces fairer matches but longer queues. A loose window fills matches faster but may produce mismatches. Most implementations widen the skill window gradually as a player’s queue time increases — starting strict, then relaxing until a match can be formed.
SBMM and lobby fill time
SBMM is more demanding on server infrastructure than naive matchmaking. Matching players by skill requires holding them in a queue until a suitable group is assembled — sometimes seconds, sometimes minutes. During that time, the matchmaker is evaluating many candidate groups across different server regions, which means maintaining queues across multiple active server locations.
At low concurrent player counts, SBMM queues grow long because there are fewer players at any given skill level available to match against. This is why SBMM works well in popular games (large player pools make finding close matches fast) but creates problems for smaller titles (thin player pools force a choice between long queues or bad matches).
The “SBMM debate”
SBMM is controversial among dedicated players, primarily in casual modes. The criticism is that it eliminates easy matches — every game becomes competitive because the system will not let a skilled player stomp beginners. Players who want to occasionally play below their level, carry friends, or simply have a relaxing session find SBMM prevents it.
Defenders argue this is exactly the point. New players who get repeatedly crushed by skilled veterans churn quickly. SBMM protects the lower end of the skill spectrum at the cost of occasionally frustrating the top end.
Many studios now run separate pools: strict SBMM for ranked modes (where fairness is expected) and relaxed or no SBMM for casual modes (where variety and fast queue times are more important).
Skill rating systems
The most common rating systems used in SBMM:
Elo — the original chess rating system, adapted for games. Each player has a numeric rating; winning against a higher-rated opponent gains more points than beating a lower-rated one. Simple and transparent, but slow to converge on accurate ratings for new players.
Glicko / Glicko-2 — an extension of Elo that tracks rating uncertainty. A new player with few games has high uncertainty — their rating can swing quickly. An established player with hundreds of games has low uncertainty — ratings are more stable. Glicko-2 is widely used in competitive games.
TrueSkill — Microsoft’s system, designed for team games. It models skill as a probability distribution and updates all players’ ratings based on team outcomes. Handles team games and free-for-alls where Elo’s head-to-head assumptions break down.
Opaque performance metrics — many games use proprietary systems that combine K/D ratio, damage, objective play, and other signals rather than pure win/loss. These are harder for players to understand or game deliberately.
See also: Matchmaker · MMR · Lobby · CCU