Ursa — the Bear — is showing up in Dota 2 pubs again. According to the D2PT aggregator, at high MMR (MMR is a numerical player rating in matchmaking) the hero has a 52.1% win rate over 2,596 games. For talk of a "return to the meta," that is enough to notice the hero. But it is not enough to draw conclusions — and the difference between those two things is exactly the main skill in working with statistics.
Two numbers in one line
The report on the hero contains two numbers, and they need to be read separately.
The first is 52.1%. That is a deviation of just two points from the nominal 50%. In a normal game, where teams are formed with roughly equal strength, a hero's win rate almost always lives in the 47–53% corridor. Two points above the midpoint means "slightly above average," not "a game-breaking pick."
The second is 2,596 matches. This is the sample size, that is, the number of games the percentage was calculated on. The smaller it is, the more the percentage "wanders" due to chance. A hypothetical example to illustrate the mechanics: if you flip a coin ten times, three heads in a row will surprise no one; if you flip it ten thousand times, a skew already means something. The same logic should be applied to heroes: 2,596 games is already a decent base, but not a final verdict.
Pubs and the pro scene: two different worlds
When a hero has "come back" in public matches, it is not the same as having "come back" at a tournament. It is useful to make the comparison explicit.
Pubs at high MMR. Players don't rehearse the draft (the selection of heroes before a match) as a five-stack, communication is random, and one strong carry can pull a game out alone. A hero that needs a lot of farm and space feels comfortable here: nobody forces them to hand over resources to the team on time.
Professional play. The draft is polished, roles are fixed in advance, and the opponent prepares specifically against your favorite picks. A hero who is strong in pubs may turn out to be awkward in a tournament bracket, and vice versa.
Hence a simple correction: the number from D2PT tells you how the hero plays in the place where it was collected. You can carry it over to other conditions only if you understand how they differ.
How to read such numbers without fooling yourself
A few questions worth asking of any hero statistic before running off to try the hero in ranked:
- Which MMR bracket? The win rate of the same hero at high and medium rating is often noticeably different: where players understand timings better, mechanically difficult picks are revealed more fully.
- How many games? A small sample is a reason for caution; a large one is a reason to take a closer look.
- Does the pick rate (the share of matches in which the hero is picked) grow along with the win rate? If the win percentage is high but the hero is rarely picked, it may be an effect of "only those who play them well pick them." That doesn't mean the hero is good for everyone.
- Which role is he stronger in? The same hero in different positions produces different statistics, and a blended number hides the differences.
What remains after the number
Even a flawlessly read win rate does not replace your own games. A hero's statistics are a map of the terrain, not a route. The most practical way to use them is as a reason to test: take the hero, play a series of matches, watch the replays, and answer a specific question — is he winning for me because of the pick itself or because of how I handle him? If after a dozen games the answer is clear, the aggregator's number has done its job. If not, then the sample of your own matches is still too small to trust either.
