Seems that my last post just blew up, so I thought I’d add a brief caveat. DotaBuff still disagrees with itself, and it’s still unclear which set of numbers is correct (assuming they’re not both busted in different ways.) I’m not convinced their hero win rate page is even updating, since I think I recall Warlock being at 57k games a few days ago.
Over this last weekend I tried to take some hero samples of my own to test which of the results they more closely match, and the results were inconclusive. One hero kinda agreed with one set, one with the other, and both were a bit off though within the expected error given the relatively small size. I intend to try again later this week when I can get larger samples.
None of this will likely change the direction of the shifts (unless DotaBuff turns out to be completely broken in their 6.78 reporting), but it could have a significant impact on the size of some of the shifts. Either way, I’ll update again once the samples are available.