If people are first matched by a matchmaker that tries to get everyone to 50% winrate?
This seems like a feedback loop that produces a lot of “noise”. If people play one hero a lot, they push the winrate of that hero closer to 50% right? No matter how good that hero is.
It would be interesting to see what ranks people reach with each hero. So an average of ranks, weighted by fraction of playtime. I wonder if that results in a cleaner picture about which heroes are under/overperforming. This is then based on how high you can “climb”, or how much effect a hero has on your rank.
It does because you can have 100% winrate on genji and 0% on symmetra for example. Your total will still be 50% winrate.
Only if you never switch as when you switch you’ll have heroes you dump losses on and others you only stay on when your winning.
If you don’t ever switch if your win rate over 50% you should be climbing if your sub 50% you’ll be falling if your at 50% you’ll be at your rank.
Win rate by itself is a meaningless stat since it is heavily influenced by outside factors beyond character power such as trends in when players choose a hero and hero popularity.
The iconic example is how Symettra had the highest win rate in the game. She was generally only played first point of a 2CP defense, quickly getting switched off of if it didn’t work. That means when the system records the stats she basically gets full credit for a win, but only a small portion of a loss causing it to inflate.
Another example of player behavior driving win rates would be Cassidy. He always had a low win rate despite how strong he was. Part of this may be due to how he was often a “last hope” character for a lot of dps players due to his flashbang. This would cause him to be played disproportionately in situations where his team is already on the back foot. If this swap tends to prolonge a loss then that means he is getting significant credit for losses, despite having been an efective pick to stop the enemy team’s momentum. Stall heroes all likely have a similar downward pressure on their win rates.
Popularity is another factor that influences win rate. The more popular a hero is the more likely they will end up in a mirror match which will push the win rate to 50-50.
1 Like
yes its important bc at the end of the day it gives you information, you can contextualize however you like based on what you know about what circumstances that stat was garnered in
its valuable to people who care about stats, if they took it out we would be calculating by hand all the time (us who care about winning and how we’re performing in detail)
That’s what I meant. It wasn’t about individual stats, but how to measure a heros total power differently
the dev team probably have a few professional statisticians and data scientists working on this; how to collect meaningful data, and how to interpret it. We should probably not sweat about it when like 0.1% of the forums probably understand any of what they do if we were to look at it raw, if that