Dunno. My intuition is that private profiles probably don’t make much difference at lower ranks where sample sizes are decent. But GM? Questionable. But, they could certainly just provide full de-identified (anonymized) data through their API somehow, instead of just summary statistics.
And map data would be awesome.
P.S.
We can adjust the formula to try to account for some of the other weirdness in Overbuff’s data. Specifically, we can set the expected mirror-match win rate higher, because win rates in GM are all above 50% (which they shouldn’t be, if properly counted). I’d have to think about it harder to come up with a good value, but if we use 53% (the average reported value for supports) we get:
| Hero | Pick Rate | Win Rate |
|---|---|---|
| Ana | 88.4% | 59.3% |
| Lucio | 36.8% | 51.7% |
| Brigitte | 33.7% | 54.9% |
| Zenyatta | 30.4% | 58.2% |
| Moira | 7.9% | 51.6% |
| Mercy | 7.3% | 50.3% |
Which, I think are more reasonable numbers. It does have a different interpretation though. Specifically, that Lucio needs buffs as well.
Conveniently, I happen to think Lucio needs buffs, so there’s that. 