Rehgar has a 54.1% winrate, and his pickrate is high enough to consider that winrate accurate and characteristic of his overall performance.
Fenix actually has a lower winrate than Rehgar right now, for perspective.
Considering that he was only in the high 40s percent range before he got self-Ancestral and “Lightning Bond” back, I’d say those two buffs likely pushed his Ghost Wolf build and corresponding diver playstyle over the top, as he is now able to keep himself alive much more easily by just self-Ancestral Healing himself whenever he gets dangerously low. His playstyle has always been a more aggressive one, and enabling that playstyle likely pushed it over the top.
Now, I’m no Rehgar expert, but I’d say in general they should either up his normal healing output but take self-Ancestral back away, tone down Rising Storm’s damage scaling, or take some of his damage away in exchange for more utility, such as making Earth Shield at least partially baseline.
I’m a pretty big fan of Anub’arak, but it’s pretty obvious he’s doing too well.
Honestly, I think the easiest and most sensible way to do that would just be taking away Rewind, but when was the last time Blizzard ever did things the easy and sensible way?
First off, the dataset is large enough it can be expected to be fairly representative of reality. In the past 7 days, approximately 31-32k Storm League and over 42k QM games have been logged, which is a huge sample size. Even if we give a extremely conservative estimate that only 50% of all games end up on HotsLogs, a sample size that big is fairly reliable, and going any larger would only give back diminishing returns on increased accuracy.
Second, most games do end up on HotsLogs. It only takes one person in a match to upload that game for all players data in that game to be logged, and there are 10 people who could possibly upload that game’s data. The chances of a game you play not being uploaded are pretty low.
Thirdly, when the devs do occasionally drop or share information from their internal data, they nearly always confirm HotsLogs’ own datasets. This has happened time and time again, and to my memory they have never said anything contradicting HotsLogs’ data.