Thursday, May 5, 2011

[Bear]Oh, the joys of savage defense continued

I've talked about how Savage Defense likely would work and how good or bad it'd be, how it works on the PTR and how it continued to work there, how SD uptime works, how good and bad it is against specific fights, and then later comments on SD in cata, especially after we found out that the combat logs are probably bugged.

To be clear: I don't like the mechanic. I don't like that it is a static amount of damage in a world where almost all other mitigation mechanics are percentages. I don't like that it has weird interactions with multiple mobs (though at least that is better with 4.1). I don't like the interaction with vengeance.

But today: I really don't like how weak it is, and I really don't like the bugs in it.

I did some study on the mathematical importance of having Pulverize and having good pulverize uptime, and the results really shocked me. You can find the long topic at the wow forums, but the basic info is that I wrote a simulator that would simulate a boss attacking and dodging and then tested a few sets of values of avoidance, hit, crit, expertise to see what would happen - specifically what would happen when you assume pulverize isn't there and you lose the 9% crit.

I expected that it wouldn't be that big a deal in terms of uptime of savage defense. And boy was I wrong! Pulverize, depending on your gear level and avoidance/crit/cap levels, is equivalent to having about 5-6% block% for a paladin or warrior. In other words, you block 5-6% more of the attacks that land with pulverize active 100% of the time.

That's huge! That'd be a big, BIG bonus for a paladin or warrior. It sounds great! And for the first time I have a nice, simple way of testing various uptimes of Savage Defense in a way that is actually simulated and not entirely theoretical.

Then I dug into it some more. That 5-6% block? Well, on a boss that does 70k damage after armor (which would be one of the more harder-hitting bosses in the game currently, like Atramedes, but not as bad as Nefarian) this would be the equivalent of reducing the average attack damage (average after avoidance, blocks and not blocked) from 35.5k to 36.7k. And that's assuming a 20k shield - which as you'll see isn't a fair assumption.

Huh. That's...disappointing. And that's on something that hits every 2.0 seconds for 70k. For things that hit harder, it gets worse. For things that hit more often it gets MUCH worse; many many more missed attacks means that blocked attacks are even less frequent and less useful when they do happen. it's not as great as we thought. It's not as great as Tangedyn's spreadsheet would indicate or my prior calcs indicate or Rawr indicates. That's not the end of the world, and we never thought it was THAT good, right?

Then today there's a post on SD being 'nerfed'. And the usual folks say logs or GTFO, and that's fair, so they're actually produced. And you see things like this:

[00:55:48.039] Beradon gains Savage Defense from Beradon (Remaining: 18267)
[00:55:54.475] Beradon's Savage Defense is refreshed by Beradon (Remaining: 15009)
[00:55:56.929] Atramedes hits Beradon 54922 (A: 21286)

That's right; SD got refreshed by a later SD proc, but because vengeance was lower the SD shield it created was actually smaller. In other words, it'd be better to not have attacked during that time. That means that any streak of avoidance actually hurts mitigation down the road. Bleh.

There are countless other examples in that post about SD not working the right way with multiple mobs, but we don't know whether that's an actual bug or just issues with the combat log. It effectively makes it impossible to determine the actual usefulness of savage defense from log parses, however; either the log is telling us the truth and SD is bad, or the log is lying but we don't know how badly.

Man, do I hate this ability. So very, very much. I guess the good news is this: Savage Defense is going to be less and less relevant as we get better gear and as we face bigger mobs, so after a while it won't be worth worrying about.