Ah, I've been waiting for something like this! Really enjoy looking over statistics and drawing some plausible conclusions from them.
Perk selection % has some unexpected bits to it. I'm really surprised to see Firebug %s that low. Commando's high pick rate on higher difficulties comes as a bit of a surprise and not at the same time, considering the odd sort of person that picks it every time no matter what and misunderstands its role completely doesn't suddenly vanish when the difficulty shifts.
The rest is about what you would think though: Gunslinger is still new and has people leveling it, in addition to still having a bit too many good points al around about it. Topping the charts on everything except for HoE where it's still 2nd (Zerkwall haven, finally confirmed for good by these %s) is to be expected.
Poor Demolitions.
That Hans vs Patriarch chart is so hilariously perfect and echoes the issues with the two's balance in more ways than is immediately apparent by the numbers alone.
Han's damage numbers are a complete joke on Normal whereas the Patriarch still sort of has the ability to actually kill people, so he comes out with a slightly higher winrate.
Hard shows the horrible, HORRIBLE problem with Hans extremely well: His melee scaling skyrockets so ridiculously with the difficulty change that playing normally against him is absolutely futile. Almost no players on the difficulty are ready to face him and aren't the type that would know they have to resign to be ready and exploit him every boss round. If this isn't proof that there should have been some sort of numbers hotfix ages ago I don't know what isn't. Meanwhile the Patriarch keeps his consistent % across all difficulties because he's an actually well thought out fight that keeps players engaged at about the same rate no matter what his numbers may be scaled to.
Now, one might see Hans' win % suddenly becoming more reasonable with the jump to Suicidal and HoE as it simply being an issue with his Hard numbers for lower skill players. Not at all. These difficulties simply house the players that already know the only way to win against Hans is to exploit him to the point where he becomes a joke of a fight, as playing normally is, well, suicide. HoE players are more seasoned about this, hence the success rate being virtually the same for both bosses there as any team that was going to win on the difficulty anyways is going to be prepared to either exploit Hans or actually play the game well VS Pat and it won't make any difference which pops up. Suicidal still has some teams that might not exploit Hans, attributing to a noticeably lower win % even with it appearing somewhat reasonable on the surface.
I realize that I'm making some assumptions here, but they're based in experience and reasoning.