When 6.82 came out, it was a common refrain that this was the 5-carry patch, as far as pubs were concerned. And there was at least some supporting evidence of that position. As I showed last post, 6.82 made pub matches significantly longer, and if you’re expecting long games why not exclusively draft heroes that are strong in the late game?
But while 6.82 might have been the least bad patch for a 5-carry line up in recent history, I didn’t buy into this line of thinking. You still have to actually win a teamfight or at least create some trades for the kill bounty changes to take effect, and a 4th and 5th carry with no farm and a terrible laning phase is a lot less likely to contribute to those kills than a support. To me it was more likely that we would see popularity shifts within the support role, but that supports as a whole wouldn’t see a huge hit to their overall success.
Unfortunately, it was difficult to tell much from win rate shifts between patches. 6.82 had a ton of other changes, making it impossible to attribute any particular hero’s success or struggle solely to the new kill bounty formulas. Fortunately, we have a much more favorable testing environment in the recent 6.82c patch.
As you can see from the patch notes, 6.82c is a pretty simple patch. We have longer buyback cooldowns and ethereal blade change, neither of which is likely to alter hero balance dramatically. Twelve heroes received changes, mostly nerfs to the top pub win rate heroes like Omniknight and Spectre or to the top 6.82 professional bans like Brewmaster, Skywrath Mage, Death Prophet, and Terrorblade. Finally we have yet another scaling back of 6.82’s rubberband mechanic. So the logic here is pretty simple: these nerfed heroes are going to lose a chunk of win rate, so the biggest recipients of that win rate are likely the heroes that benefit the most from weaker rubberbanding.
To get these pre- and post-patch winrates (Very High only of course) I consulted DotaMax, but they unfortunately do not divide their winrates by mini-patches. To get around this, I used a Last Week search to create a slice that was purely 6.82c games. For something to compare that to, I also grabbed the 6.82 results as a whole at that point, and subtracted the 6.82c slice to create what is close enough to those desired pre- and post-patch results. You can view the raw data here.
As expected, a small group of nerfed heroes represented a lion’s share of the negative momentum:
And who benefited from all this freed up win rate?
With the exceptions of Treant Protector, as well as Chen and Earth Spirit depending on how you define things, supports were pretty much absent from the upper echelon of beneficiaries. My suspicion based off this is that the heroes most hurt by the kill bounty changes were carries and semi-carries who thrive on early game bullying or scaling and leveraging that into a win but lack the long-term scaling and utility to reliably win fair fights in a late-game scenario, as that’s a pretty fair descriptor for this entire list outside of the three exceptions I already mentioned. And if you’re wondering about conspicuous absences, Viper was just outside the cutoff point for the graphic at +0.77%.
Another conspicuous absence is Razor, but as you may recall Razor was one of the twelve heroes receiving predominantly nerfs in 6.82c. His shift was -1.05% which seems relatively minor, but what could be happening is that the win rate loss from his nerfs is being mitigated by 6.82c being a more favorable enviornment for similar reasons to the above heroes.
Incidentally, the same logic could apply to Earth Spirit. The 6.82 environment as a whole was detrimental to the hero, but this may have disguised the apparent fact that his hero specific changes appear to have been big improvements. As the rubberbanding is progressively reigned in his win rate is surging, at least in top end pub play.