Two ways to say “typical”
When a player is said to “average” 18 points, 18 kills or 18 minutes per match, the reader usually hears “this is what they normally do”. That is not always what the figure means. The word average most often refers to the mean: add up all the values and divide by how many there are. The median is different. Line the values up from smallest to largest and take the one in the middle.
Both are summaries of the same data. They answer slightly different questions, and when the data is lopsided they can disagree noticeably.
A worked example
Here is an illustrative set of nine match scores for a player, in ascending order:
8, 9, 10, 10, 11, 12, 12, 13, 75
The total is 160, and 160 divided by 9 is about 17.8. So the mean is roughly 17.8.
The middle value of nine numbers is the fifth one, which is 11. The median is 11.
In eight of nine matches, this player scored 13 or less. The mean of about 17.8 is higher than eight of the nine scores. It describes no match the player actually played. The one score of 75 pulled it up. The median of 11 sits right in the cluster of ordinary games.
Neither figure is wrong. The mean tells you the total output spread evenly across matches, which is useful if you care about totals. The median tells you what a normal match looks like.
Why outliers affect the mean more
The mean uses every value and weights them by size, so one extreme number moves it. The median only cares about order, so replacing 75 with 750 would leave the median at 11 while the mean would jump enormously. This is why the median is described as resistant to outliers.
An outlier is not necessarily an error. A huge game in a video game match, a freak scoring night or a single long-range kill streak can all be real. The question is whether such a result represents what the player usually does, or whether it is closer to a rare event.
Where this shows up in sport and esports
Lopsided data is common in games. Some typical cases:
- Damage, kills or points per match, where occasional blowout games sit far above the rest.
- Salaries and prize earnings, where a few top figures pull the mean far above what most players receive.
- Match length, where a few very long games raise the mean.
- Time to complete a task, such as a speedrun segment, where rare long failures stretch the average.
In each of these, a report that gives only the mean may leave a more ordinary typical case hidden. Conversely, a report that gives only the median may hide the fact that occasional huge games exist, which matters if you care about the player’s ceiling.
When the mean is the better choice
The median is not automatically superior. If the quantity you care about is a total, such as total points over a season, the mean is the right tool because it keeps track of every contribution. If the data is fairly symmetric, with no large outliers, the mean and median will be close and either is fine. A median can also feel slippery in small samples, since it can jump by a notable amount from one value to the next.
Some practical guidelines:
- If the mean and median are close together, the data is probably fairly balanced.
- If the mean is clearly above the median, there are probably a few unusually large values pulling it up.
- If the mean is clearly below the median, there are probably a few unusually small values, for example games where a player barely took part.
Ask for both, and for the spread
The most useful habit is to look for both figures, or at least ask which one is being reported and why. Reporting only the flattering one is a form of selection, which overlaps with cherry-picked statistics. It is also helpful to know how spread out the results are, which is what the standard deviation describes, because two players with the same mean can have very different consistency.
It also matters how many games the figures come from. Medians and means based on a handful of matches can both be unstable, which ties back to small sample sizes. And when the shape of the data matters, it may help to know about probability distributions, since many game outcomes are skewed rather than bell-shaped, in contrast to the normal distribution that people often assume by default.
A note on interpretation
Even a well-chosen summary describes the past, not a promise about the future. A player who has had one enormous game has not necessarily raised their normal level, and a mean inflated by that game is not a forecast. Treat both numbers as descriptions, look at how they compare, and let a gap between them prompt a closer look at the underlying results rather than a firm conclusion.