Overview
In developer blogs, they often point out winrates as justification when making balance changes. (E.g., Orisa's recent buff was justified by her lower winrate on Twitter)I've seen quite a few players pointing out winrate as a point to prove that a hero is or isn't broken, but I wanted to put my thoughts out there on how people use this statistic.
Statistics are great - they’re a good way to see exactly what’s being played and their effectiveness across ranks very quickly, which is especially great when balancing a game like Overwatch with millions of players. Statistics are objective and don’t lie, but they also don’t tell the whole story.
There are so many factors that go into a hero’s winrate in a team game like Overwatch, you can’t use them as the be-all and end-all stat to prove something. Imo, they’re best used as the starting point - examining a hero further to find out why their winrate might be low/high.
Factors that make winrate unreliable for balance
As an example, Kiriko in a vaccum has incredible mobility, damage potential, and utility with her suzu. And yet she has one of the lowest support winrates, why is that?
Well there’s different factors that impact winrates:
-
Mirrored vs Unmirrored: Meaning the winrate when both teams are using Kiriko, or when only one is. As a team game though, both can be skewed by the other heroes in the comp.
-
Team comp: This affects the synergy Kiriko has with her teammates, and what she can focus on in a team fight.
If it’s Kiriko with an Ana, she may have more room to take angles and deal damage, whereas with a Lucio there might be less room to damage if you need to heal more (especially if you’re playing with a resource-reliant tank like Reinhardt that takes a lot of damage). -
Rank: Heroes are going to perform differently across ranks; a bronze Kiriko is likely going to have a lot less damage and possibly more healing than a top 500 Kiriko.
Looking at her kit she has abilities with a high skill floor. Hitting headshots and using suzu at optimal times takes skill, which is why at lower ranks her winrate is lower compared to heroes like Moira who are easier and provide more value at their skill floors. -
Pick Rate + Winrate: Even at GM Kiriko has the 3rd lowest winrate (according to OverBuff which isn’t the most accurate but it’ll do), why?
Well part of this is pick rate, which is self-explanatory - how often a hero is played, but how pick rate changes depends on a few things. A hero could have recently been buffed, or players could have a good read on the meta and play a hero they know is strong. A hero could simply be fun or popular, which is why characters like Rein/Genji/Mercy have high pick rates regardless of rank.
When a hero has a high pick rate, it doesn’t guarantee that their winrate will be high or low. But the specifics of the hero can affect how we determine the relevance of that data. Kiriko has a high skill-ceiling and could probably two-tap every squishy on the enemy team, but her high points may happen only a few times a match, or never. Her consistency impacts her winrate, which is why some heroes may appear to have a higher winrate despite being “lower on tier lists”.
The best example of how pick rate impacts winrate imo is a hero like Symmetra, who has a better winrate than Sojourn in GM. Does this mean Sym is better than Sojourn? Not necessarily. Looking at other data points, she also has the 2nd lowest pick-rate in GM. Players at this rank don’t pick Sym often, but they know how effective she can be on maps like Lijang Tower, and don’t play her much on non-optimal maps. This impacts her winrate - so when players pick Sym, they know she’s going to be effective.
In-Game Statistics
Let’s look at another stat I see people misusing - KDA (kills, deaths, assists)
Moira has the 2nd highest KDA ratio of all roles in both bronze and GM, but why?
Kills aren’t just given to players who dealt the final blow, but others who contributed to that kill. If a hero was healing or supporting an ally that got a kill, they get an assist. This makes it easier for some heroes, like Moira, to get really high stats by slightly damaging an enemy that gets killed. She also has high survivability which can make her average deaths lower per game.
Similarly, high damage and healing doesn’t mean you necessarily carried your team. With Moira again, she can do lots of AoE damage/healing with her orbs, but it may not pose a threat to the enemy or sustain your tank enough. Compared to burst healing or damage, which when applied at the right time, can decide the teamfight. (A widow may have low damage, but pick off the enemy Ana that wins the fight for you)
Conclusion
When developer blogs focus on quantitative statistics but have little to no input on qualitative data, it can make people in the community think they focus too much on stats to determine balance changes.
And maybe they do, maybe they don’t; no one knows how they come to balance decisions internally. Regardless, I think developer blogs could improve their communication by talking more about balance outside of stats.
TLDR: Essentially depending how you frame it, statistics can be framed to mean whatever you want. High winrates or in-game stats don’t necessarily mean a hero is broken or someone carried, or vice-versa. Context is important in a game with as many variables as Overwatch. Also as usual, Blizzard needs to communicate better.