What a fair game means
In most competitive games, a fair match is one where each side has a similar chance of winning. If the outcome is nearly certain before it starts, the weaker side has little to play for and the stronger side learns little. Matchmaking is the process of finding opponents and teammates so that games land close to an even chance.
The simplest way to express this goal uses a rating system. If two players have the same Elo rating, the expected score for each is 0.5. A 100 point gap shifts that to about 0.64 for the higher-rated player, and a 200 point gap to about 0.76. A matchmaker can treat those expected scores as a quality measure and try to keep them near 0.5.
The basic process
A typical system works roughly like this:
- A player joins a queue with a rating, and often an uncertainty estimate.
- The system looks for other players, or teams, whose ratings fall within an acceptable range.
- If no good match exists, the range gradually widens.
- When enough players are found, the game begins.
The widening step is important. A strict rule that only allows near-identical ratings would leave some players waiting a long time, particularly at the highest and lowest ends of the rating range where there are fewer people.
The trade-offs
Fair matchmaking is not one goal but several that pull against each other.
Match quality versus waiting time. Tighter rating limits give closer games but longer queues. A looser limit shortens the wait but accepts more lopsided matches. Operators have to pick a point on this curve, and the best choice depends on how many players are online at a given moment.
Quality versus connection. Players far apart geographically will have higher latency, and a game with poor connections can be unfair regardless of rating. Considerations such as tick rate and server location therefore compete with rating when choosing who plays together.
Team balance. In team games, the system must combine individual ratings into a team value. Using the average is easy, but two teams with the same average can differ a great deal if one has a single star and the other has five similar players. Systems built around TrueSkill ratings, which are designed for teams, try to address this by working with whole-team skill distributions.
Parties and friends. Players who queue together tend to coordinate better than random groups, so some systems adjust their ratings or match them against other premade groups.
Role needs. Games with set roles must balance roles as well as skill, which narrows the pool further.
Uncertainty. A new player’s rating is not reliable yet. Systems that track uncertainty can place newcomers in lower-stakes games until their rating settles, then tighten the matching as it becomes better established.
Win rate is not the target
It is tempting to assume a good matchmaker pushes every player toward a 50% win rate. That is partly true over the long run, because a rating that is accurate and a pool that is well matched produce results near even. But it does not mean every game is a coin flip, and short runs vary a lot. A sequence of losses can occur by chance, and the same logic as regression to the mean means a poor streak is likely to be followed by something more typical.
Some players suspect that systems deliberately arrange wins and losses. Public descriptions of rating-based systems do not require this, and streaks occur naturally in random sequences, which is why a belief in a hot hand, or in a rigged run of losses, is easy to form. Whether any particular game uses additional rules is not something this article can verify, so it is best to treat the idea with caution and look at long records rather than a handful of games.
Designing for more than fairness
Matchmaking has goals beyond even odds. Operators may want to reduce the chance of repeating the same opponents, keep new players away from veterans, or support a variety of play styles and modes. Some also factor in behaviour, such as reports of abusive conduct, so that those players are grouped together or kept apart from others.
These aims do not all fit neatly into a single rating. The result is that modern systems combine a skill estimate with constraints for region, role, party size and wait time, and the details differ from game to game and are often not fully public.
What to expect as a player
Rating-based matchmaking improves the average quality of games, but it cannot guarantee that any single match is balanced. Hidden factors such as form on the day, map choice and the quality of teammates sit outside the number. The helpful framing is that matchmaking is trying to keep the pool near even odds, not promising that every outcome is fair or predictable.