>For instance, if we look at Lithodora's PK Statistics, it
>looks like AP is super deadly to thief. But I don't believe
>for a second that is generally true.

It's true to the extent that APs can be duergar and/or buff APs are reasonably likely to end up with a certain item at some point in their lives. In that character's case, they were also obligated by role-play to engage with Sodsob. An otherwise identical thief who was an uncaballed pacifist would probably not have suffered quite so much.

For fun, I dug up some stats on 2012+ fort thieves that had a PK focus:

Ageryn, 528 hours, 166-41 overall, 9-4 vs. AP
Mochodin, 206 hours, 108-8 overall, 4-0 vs. AP
Vallinane, 827 hours, 79-93 overall, 2-3 vs. AP
Lithodora, 589 hours, 77-31 overall, 3-13 vs. AP
Resdrenn, 203 hours, 74-20 overall, 13-0 vs. AP
Kasir, 544 hours, 74-16 overall, 2-0 vs. AP
Thaedan, 547 hours, 72-21 overall, 1-5 vs. AP
Irineth, 236 hours, 57-18 overall, 9-1 vs. AP
Spiorra, 262 hours, 34-22 overall, 3-0 vs. AP

Seems like in terms of success against APs it helps to 1. be a thug and 2. spend a lot of time at lower ranks when the anti-paladins don't have buff weapons, and 3. not have to deal with shaapa.

Any class that can dish out a lot of melee damage, learns weapons thieve's can't parry well, can bash, and can potentially see hidden (either via an item or by being duergar) will be dangerous to a thief.

>doesn't support my feeling that assassin has moved from good
>to a bit OP (under the right circumstances, which includes a
>couple of edges).

I'm saying that for players who know what they're doing, assassin with edges still doesn't rank all that high in their list of concerns. The main exception being assassinate and, to a lesser extent, mark of the prey. And that's how I tend to evaluate whether a class/item/skill is overpowered: how hard or easy is it for a vet to neutralize (or minimize) the class's advantage, part of which may be simply avoiding fighting that class except under choice circumstances?

For the assassin class, it seems like vets aren't dying to them all that much outside of assassinate.

>I'd love to see some real analysis on the total data set, but
>that's impossible for us, so it's really up to you.

Not something I have access to at the moment. Like I said above, though, I'm not sure looking at "all the numbers" is the best way to gauge whether something is OP. If a class has a gimmick that works against newbies but hardly ever against anyone who knows what they're doing, is it really OP?