Two goals that conflict
A matchmaker has two jobs. It should produce games that are even, so both sides have a fair chance. It should also start those games quickly, so players are not left staring at a timer. These goals pull in opposite directions, and nearly every design decision in matchmaking is a way of choosing a point between them.
The underlying reason is simple. A queue contains only the players who are searching right now. The tighter the requirements for a match, the fewer players qualify, and the longer the system must wait for enough of them to appear. Demanding a perfect match from a small pool means waiting; accepting any match means games that may be lopsided.
How a search window works
A common approach is a search window around a player’s rating. When you join the queue, the system looks for opponents within a narrow band. If it cannot fill a match, the band widens as your wait grows. Early on, the matchmaker is strict; after a minute or so, it relaxes; eventually it accepts nearly anyone available.
A few consequences follow.
- Most players get reasonably close matches quickly, because the bulk of players sit near the middle of the rating range where the pool is dense.
- Players at the extremes wait longer or get looser games. A very high-rated or very low-rated player has fewer peers online, so the window must grow wider before it finds anyone. Top-ranked players often report long waits or matches against much lower-rated opposition for exactly this reason.
- The same rating can mean different queue experiences at different times of day or in different regions, because the number of players online changes.
What counts as match quality
Quality is not one number. Systems may weigh several things.
- Rating gap. The difference between sides, which determines the predicted win probability, as described in the Elo ratings formula.
- Uncertainty. Whether the ratings are well established; a match involving a newcomer is harder to judge than one between veterans, and systems using rating uncertainty may account for this.
- Team balance. In team modes, sides need to be balanced, not just have equal averages, a limitation discussed under team ratings.
- Roles and composition. Some games need specific roles filled, which narrows the pool further.
- Connection quality. Distance to a server and latency also affect fairness, and a good match on skill can still feel bad if the connection is poor. Even with rollback netcode in fighting games, high latency changes how fair a match feels.
- Recent opponents and party size. Systems may avoid repeats and keep groups of similar size on both sides.
Each added requirement shrinks the pool of valid matches, so every extra quality constraint costs queue time.
Choosing a balance
There is no objectively correct setting, because it depends on what players value and how large the population is.
Competitive and ranked queues usually lean toward quality. Players there accept longer waits in exchange for results that mean something, since lopsided games add noise to ratings and make progress feel arbitrary.
Casual queues usually lean toward speed. The cost of a mismatch is lower, and long waits drive players away.
Small populations force compromise. When few players are online, such as in a new game, a niche mode or an off-peak hour, a strict matchmaker would leave people waiting indefinitely, so it has to loosen.
Large populations can afford strictness. More players means tight windows still fill fast, so a game with a big active base can deliver both speed and fairness for most of its players.
The hidden costs of both extremes
Sacrificing quality for speed produces stomps. Games that are decided early are frustrating on both sides, and they also give ratings little information, because the result was close to predetermined. Sacrificing speed for quality increases the chance that players give up and quit the queue, which in turn shrinks the pool and lengthens the wait for everyone else. Left alone, that can become a spiral in which a mode empties out.
Related design levers
Beyond the window, designers have other tools.
- Separate pools by region, mode or rating range, which can improve match quality but split the population.
- Bracketed queues that restrict very different ranks from playing together, sometimes with exceptions for parties.
- Displaying estimated wait times so players can choose to wait or accept a looser game.
- Flexible roles so that players willing to fill any position get shorter queues.
- Treating new players differently, for example using placement matches with deliberately wider matching so the system learns their level quickly.
Interpreting your own experience
If your games feel uneven, the explanation is often structural rather than a flaw in your rating. You may be at a thin part of the rating range, playing at an off-peak time, or in a small region. Occasional lopsided games are the price of keeping the queue moving. Judging whether a ladder is working well requires looking at many matches, and the usual caution applies: a few unbalanced games say little. An honest picture needs a decent sample size, not just the memory of the most recent stomp.