It kinda is though, well, I guess not to "Stack" Block over other values, but because of Shield Block and Critical Block, it's going to be a lot more attractive to Warriors than AP is for druids. .77 Block also costs a lot less (.5 item value) itemization wise than 1.56 Strength (1.56 item value). .77 Block with Shield Mastery becomes 1 Block, and with Blessing of Kings, Abomination's Might, and Protector of the Pack this becomes 4 AP which results in +1 on Savage Defense. SD exists to give us better scaling on our weapons, and probably to silence the many complaints about Druid Tank itemization, as all "Tanking" Necks, Cloaks, and Rings come with previously worthless Strength, now it's just mostly worthless.

And I'll reiterate about Avoidance Diminishing Returns: Just because it takes more Agi/Dodge to increase your effective Dodge% doesn't mean that it's getting worse. Agility will always be our best mitigation stat (barring armor, but you can't really gem for that) until things dramatically change again, no matter how much Dodge you have (until 100%), the value of additional Dodge doesn't ever change, ie. Stop thinking of it as "Diminishing" because it really doesn't.

Interestingly enough, the dodge and agility components of Idol of Mutilation can stack with each other, meaning you can proc agi in cat form, then go bear and have both buffs for 14~ seconds. I tested it on the PTR.

This could be useful on tank-swap fights.

wrong forum location - Deleted

This is definitely not true. Agility is by far the worst stat for threat. Hit and Expertise are best. ArP, Str, and Haste are next with values (of course) dependent on gear and rotation.

To help explain what Boevis is telling you think of it this way:

If you have 0% dodge (impossible I know but for the sake of napkin math) and you add 1% dodge you are now mitigating 1% of the overall incoming damage. Say incoming DPS is 100 (ez numbers). With the 1% dodge it would be 99 DPS. Obviously a 1% decrease there as expected.

Now if you have 50% dodge and add 1% dodge you are now mitigating 2% of the overall incoming damage, not 1%. Think of it this way. Originally we are taking 100 DPS, with 50% dodge it becomes 50 DPS. With 1% more dodge that 50 DPS drops to 49 DPS, a 2% decrease in the current damage we are taking. In this way while dodge becomes harder to increase, it also becomes more powerful per point gained. It scales much like armor does in that respect.

So as Boevis says - Agility will always be our best mitigation stat, aside from armor.

I think Riprawr might be misinterpeting Nightcrowler a bit; a link to the specific post would be helpful. The ranks are closer to what Mijae describes if you look at the TPS column for the relative stat values in this post: FeralbyNight (Cat & Bear simulation tool)

In order of TPS contributions from that table (which is for a simulation with a specific near-BiS gearset):

1. Hit/Expertise (2.18)
2. Haste (1.93)
3. ArP (1.85)
4. Strength (1.75)
5. Crit (1.03)
6. Feral AP (0.88)
7. AP (0.73)
8. Agility (0.66)

I'm guessing Riprawr is looking at "Night's Choice" stat rankings. Agility is a "best" tank stat because it contributes in a laundry list of ways (some significant and others not): to mitigation (via Savage Defense and a bit of armor), avoidance (dodge, obviously), a bit of healing from iLoTP, and threat (directly via crit and, in rage-limited situations, through Primal Fury and Natural Reaction).

If the only thing you're concerned with is TPS, though, it's really low, especially when you consider that you get half as much per item budget as you do for AP. Edit: That doesn't mean ferals are going to go gemming for haste or ArP in their tank sets; short of poor execution on Hodir (tank in FrR not getting the stormcloud), I can't think of a place that an agi-gemmed feral in reasonable gear with a decent rotation is going to have threat problems in Ulduar.

Ok, after some silence on my log parsing attempts, here are the latest results. The C++ code is definitely not yet for public consumption (no interface at all), but if someone is interested let me know and I'll make it available.

The basic idea was: analyze a log file and extract two kind of information:
- hp dwell time for the tank
- damage/heal correlation

Problems first: as it turns out, it's not as easy as it seems. The logfile at times is non-causal, with weird contradictions which I've been unable to resolve. Manually, it would be possible, but procedurally.... I have no idea how to proceed. Since from the logfile I have no way to access the actual HP value for the tank (is this possible by coding an addon?) I use the heal+overheal events to resync my hp value to the maximum, but at times weird things happen. An example from a log:

t = 7/1 23:30:33.421: heal event (433, Judgement of Light from Paladin 1)
health transition from 34664/46000 to 35097/46000
t = 7/1 23:30:33.437: heal+overheal event (2091/1123, Living Seed from Druid 3)
health transition from 35097/46000 to 37188/46000 Overheal reported, but health < maxhealth??
health adjusted 37188 to 46000 (8812 / 8812)
t = 7/1 23:30:33.562: heal+overheal event (8812/1173, Nourish from Druid 3)
health transition from 46000/46000 to 54812/46000 OVER MAX HEALTH??
health adjusted 54812 to 46000 (-8812 / 0)
t = 7/1 23:30:34.062: damage event (-1581, Flame Jets from Ignis the Furnace Master)
health transition from 46000/46000 to 44419/46000

As you can see there are two consecutive heals which report heal+overheal, which is nonsensical since they occurr at two different times. I have no idea if this is related to lag across the clients and the server, with events being generated on one not necessairly able to see the others. Of course this may mess with the hp dwell time graph.

Another problem: I'm missing some healing. I trap the standard events, and I end up with some "positive correction" to be applied to the tank's health. The effect is not big, but, as above, it will mess up the hp dwell time graph. It must be some aura or something which is healing me and is not appearing in the logs. The value is too big for normal HP regeneration (for the fight below it's 11192 HP total in 6 mins).

One more problem: I have not found a way to compute the correlation among discrete events, so I transform the event sequence into a continuous function via interpolation. The question is: what kind of interpolation? Right now I'm using an ideal low-pass filter, which means sin(x)/x interpolation in time space (and a nice square function in frequency space). This explains why on the graphs you see negative healing or damage. BTW using a gaussian does not give very different results.

I'm now thinking about a different approach which consists in performing statistics on the heal events after the damage event, but I first need to get a clear idea of what I want and the some coding, so it'll not be for tomorrow

For the flashy graphics: here are the results from Ignis-25 (nice fight with no interruptions/phases). Healers are named Druid 1, Shaman 1, etc. Helistar (me) is the tank and the healing you see is ILotP. The % number is the % healed (to see for which ones looking at correlation is actually useful).



These are calculated with a 5 second filter cutoff (f_0 = 1/5 Hz).







In game time 5 seconds is quite long, so you get a general picture. The details can be seen by using a lower cutoff, here it's 1 second (i.e. 1 Hz):
You get a ton of "echo" due to the fact that damage (and healing) is roughly periodic (boss swing time).



If you want to see all the images, they are at:

Index of /wow/logparser/cutoff1sec

Index of /wow/logparser/cutoff5sec

Might the latter of these consecutive overheals add up to, say, 11192 in your example parse? Depending on how you actually sync the current health from the overheal messages, you could end up loosing some of the actual healing done due to erroneous overheal events. As to how to fix it, if that is the problem, you could convert the duplicate overheals to effective healing done.

@Isambaard: true, at the same time all fight pose this kind of problem one way or another (the only exception I can think of is Patchwerk). In any case since I want to get an idea of how to simulate statistically the incoming heals, I won't base my model on a single fight.

@Otsu: The total overheal is a lot more than 11192, I have looked manually at the heal-overheal events but it's hard to decide which ones are "right" and which ones are wrong, so I have no idea which to sum to see if the 11192 could be explained. I'll look into scripting to see if it's possible to record the tank health directly, which would solve the problem once and for all.

Newest ptr changes.

With the amount of dodge you get per point of Agi decreasing, while the amount of Parry increases I beleive that the other three tank classes will find it balances out, more or less. However, how big an impact will this have on Feral tanks? I find it hard to beleive the only avoidance stat for Feral tanks will be reduced by 15%, so I suspect I am misunderstanding this change. Can anyone explain it to me?