Beyond single-roll odds
When people talk about probability in games, they often picture a single roll: a 30% chance of this, a 70% chance of that. Team-based multiplayer games work differently. A result comes from many players’ decisions and many small random events, so the question is less “what are the odds of this one event?” and more “how widely do outcomes spread, and what shifts that spread?”
That spread is called variance, and the usual measure of it is the standard deviation. Two teams can have the same average strength and still behave very differently: one wins steadily, the other swings between brilliant and terrible.
Why bigger teams smooth things out
Imagine each player on a team independently has a chance of playing well in a given match. This is a simplification, since real performance is influenced by teammates, but it shows a useful effect.
Illustration: each of five players has a 50% chance of having a good game. The number of good performances follows a binomial distribution with five trials. The average is 5 × 0.5 = 2.5. The standard deviation is the square root of 5 × 0.5 × 0.5, which is the square root of 1.25, about 1.12.
Relative to the average, a larger group is steadier. One player alone is a 0-or-1 outcome. Five players together tend to cluster closer to the middle. This is one reason team games feel less random than one-on-one games: individual randomness partly cancels out.
The independence assumption does not hold perfectly. If players influence each other, such as when one player’s mistake ruins another’s position, outcomes can be more correlated, and the smoothing weakens.
Composition: reliable versus volatile
When choosing a team composition, there is often a trade-off between reliable and volatile options.
- A reliable composition performs near its average most of the time.
- A volatile composition has a higher peak and a lower floor.
If you are the favourite, reliability usually helps, because randomness mostly gives the other side a chance to catch up. If you are the underdog, a volatile approach can raise your chance of an upset, even if it lowers your average result. This mirrors the broader idea of risk tolerance, and it ties to game volatility as a design concept.
No composition guarantees a result. The most that can be said is that choices shift the odds and the spread, and the outcome of any single match remains uncertain.
Matchmaking and what it does to results
Matchmaking tries to pair players or teams of similar strength. Rating systems, including Elo ratings, estimate the chance that one side beats another based on a rating gap. In the standard Elo formula, a gap of 400 points means the stronger side is expected to score about 91% (10 out of 11), and a gap of 100 points gives roughly 64%. These are properties of the formula, and games using different systems will differ.
What matchmaking changes is the spread of your results. If you are always matched with close opponents, your win rate drifts toward 50% over time, even as your skill improves, because the system keeps adjusting opponents upward. This is why a flat win rate is not evidence that you have stopped improving, and why a short winning or losing run says little, as small sample sizes make short runs unreliable.
Series length and luck
Longer series reduce the role of luck. A simple illustration: suppose one side has a 55% chance of winning each game, and games are independent.
- Single game: 55%.
- Best of three: the stronger side must win at least two. The chance is 0.55² + 2 × 0.55² × 0.45. Since 0.55² = 0.3025, this is 0.3025 + 2 × 0.3025 × 0.45 = 0.3025 + 0.27225 = 0.57475, about 57.5%.
- Best of five: the stronger side needs three wins. Adding the cases of 3, 4 and 5 wins gives about 59.3%.
Each step helps the better side, but only slightly in this example. When the per-game edge is small, even a long series leaves a substantial chance of an upset. That fits the broader point that edges in team games are usually modest, and that results mix skill with luck.
Practical ways to think about it
- Look at spread, not just averages. A strategy with the same average but lower variance is safer for a favourite.
- Do not read too much into a few matches. Evaluate across many games.
- Expect noise from teammates. Results are partly outside your control, and that is normal.
- Judge decisions by process. A sound choice can lose and an unsound one can win.
Over a large number of games, the law of large numbers pulls results toward underlying strength, which is why long-term records are more informative than any single match.
Boundaries
This analysis is simplified. Real games have hidden mechanics, changing patches and human behaviour that these models do not capture. Nothing here promises a win, and numbers like 55% are illustrations, not measurements of any real game.