---

Q u o t e:

The paladin is the same. There aren't a whole lot of options for a 'max stam' warrior or paladin, because they lack options to become easily uncrittable. I'm interested in hearing what your version of a 'balanced warrior' would be though, considering the gear available at the moment.

---

IMO, they should of done 4 models for warriors/pallies. Sta/avoidance/block/balance. That way they could demostrate the effectiveness of block.

---

Q u o t e:

IMO, they should of done 4 models for warriors/pallies. Sta/avoidance/block/balance. That way they could demostrate the effectiveness of block.

---

Its not "they" Its one guy with alot of spare time... hes not perfect, and he did damn well for doing everything himself.

I found it to be a fairly interesting read.

I also, despite explanations as to why, found it somewhat off-track to not include a deep wounds warrior build in the study. At the very least, doing so would display how incredibly low-value the three points in shield specialization we give up are. The fact that we "don't necessarily need to" spec into it seems far less relevent than the reality that we DO spec into it because what we give up is so trivial. The side by side comparison could have displayed this.

It brings alot of the 3.1 changes into light. DK's are very close to equal to other tanks when their window of "I'm not going to die" cooldowns is much larger and avoidance is brought to a more manageable level.

Druid linear EH values have been brought well into the range of other tanks, in my opinion, and this shows why it happened. I think there is a real possibility that this will bring their cooldown/avoidance differences to light in comparison, though. Magic mitigation is almost certainly worth a look at this point.

As far as block goes, I'd tend to agree that the block value stat is a shackle to our ability to have it brought to its intended level of mitigation. If bosses are really going to average 20k damage per hit, we're going to find block to be incredibly trivial in its current incarnation. Some semi-random thoughts on block:

Going a different way with BV--perhaps it should be unchained from our threat generation. I realize that at current there is a variety of talents and a significant ability for both classes that utilize block value to make threat, and these seem to be the thing that really prevents it from doing what it should.

We just can't have a constant 4k BV when it means 20k shield slams with shield block up, and 4k passive DPS from damage shield on trash packs. Beyond that, though(at least on boss fights)....why can't we? If our damage shield and shield slam abilities didn't rely on block value and instead it was a simple mitigation stat, its itemization cost could be effectively doubled or tripled, and the amount contributed by stats could be increased as well. Shield Slam and Damage Shield could be given an attack power modifier to put them at functional levels.

This would have two benefits. The first is that block value could be consistently tuned to the desired mitigation level for the content that the gear drops from. The second is that ALL damaging abilities would be derived from the same source(AP) and thus could accurately be tuned without constantly having to deal with the double edged sword that is gaining damage from a mitigation stat.

The only problem I could see this on is trash. Trivializing trash is one thing, over trivializing it to the point that the healers are on follow and afk for it is quite another. Perhaps block should have a maximum percentage amount, say 50-75%, somewhere in that range, making it impossible to full block an attack.

---

Q u o t e:


Its not "they" Its one guy with alot of spare time... hes not perfect, and he did damn well for doing everything himself.

---

I know its a single person, I just used "they" for a neutral sense. I was just saying it would of been reasonable to have an example of a block set tested for comparison's sake.

---

Q u o t e:


Its not "they" Its one guy with alot of spare time... hes not perfect, and he did damn well for doing everything himself.

---

It was a good general observation, that produced a good general conclusion. Unfortunately, everyone sees a bunch of numbers and thinks they're allowed to glean these unjustifiably precise (i.e. specific) conclusions. We have to nerf this by this much, we have to change block in this way, 120 sec cooldowns for DK's will fix it (or they'll turn us into walking repair bills, depending who you ask), etc. This test doesn't support any of this, it tested whether or not there was an imbalance in damage. That is all. (And quite honestly, I don't think this test was anywhere near the first time blizzard noticed.)

These tests show that DKs don't take enough damage, and Druids are too good at taking damage. However, because of the limited scope of the tests, they can only support very general conclusions. There isn't any real data revealing why or what might be done to correct the situation. Just because you throw some speculation onto your test after you've recounted the numbers (or throw some speculation onto someone else's test after it has been posted) doesn't make any of it valid. Besides the actual conclusion the data supports, that is.

---

Q u o t e:
Cool, so attack the person rather than the obvious valid points they've made. That works, right?

Apparently it doesn't, because blizzard is objective and strives for ultimate balance and, surprise, two classes saw changes.

---

Blizzard isn't objective. Blizzard gives in to whomever whines the loudest. See: Brew of the Year achievement.

When it comes to tanking, the loudest whiner is generally the warrior.

Also, Xav, you can't complain about people attacking "the person" rather than the points. Not when you talk about "flailing, ignorant people" (post 63 of this thread).

"These tests show that DKs don't take enough damage, and Druids are too good at taking damage."

Omfg a class intended to take crushing blows and mitigate it with our old armor and hp is good at taking damage? I DIDNT KNOW THAT !1

What version of rawr is used for those character sheets? I'd like to look at them but I'm too stupid to get them working.

---

Q u o t e:
What version of rawr is used for those character sheets? I'd like to look at them but I'm too stupid to get them working.

---

Unless he used the new 2.2 Beta build of Rawr, all Rawr files should be compatible. The Beta makes it so you can't load up older files. Just get the latest Rawr and the Beta version and load them up on both. I don't rightly remember which I used to look at the files but they didn't give me problems.

Can he actually post the raw data?

Statistically, these graphs don't actually tell us anything because he didn't include rms or error. I would assume he ran a sufficient number of simulations to calculate the error, but he didn't include that number either.

I can see that he put a lot of work into it, but we really need SOME sort of error estimation to see anything from this sort of "study". Without it, the graphs are - well, worthless.

again - from what I gather he used his own simulations

and I posted a link to his jave code/jave binary in an earlier message

http://www.kamuf.org/parri/ThinkTank/

Rauxis, chosen of CAT

---

Q u o t e:
again - from what I gather he used his own simulations

and I posted a link to his jave code/jave binary in an earlier message

http://www.kamuf.org/parri/ThinkTank/

Rauxis, chosen of CAT

---

Yes, but he doesn't post the damage ranges, so trying to recreate it is hit or miss. I have been playing with the simulator with 1 change made for warriors and trying to compare it to the current data, but I am having to guess at the input values. The findings I have from my guess is in the thread posting the link.

---

Q u o t e:
again - from what I gather he used his own simulations

and I posted a link to his jave code/jave binary in an earlier message

http://www.kamuf.org/parri/ThinkTank/

Rauxis, chosen of CAT

---

This does absolutely nothing to answer my concerns or to make his results worth while. He has removed the source so I can't even recreate his results.

Let me put my concern another way. He has a "difference" of 4 seconds in his TTL calculation. He does not report the variance, standard deviation, rms, or any other form of error for his calculation. So, if I assume he has a standard deviation of 70 seconds, how meaningful is the 4 second difference between tanks?

This is a common mistake people make when trying to do something scientifically for the first time. Results without error are meaningless. Not just "non-optimal" - absolutely meaningless. I would just calculate the errors myself, but he didn't include the necessary information.

In the end, I will probably write my own simulation. It is probably far easier to start from the ground up and build it to include all the missing pieces then to modify existing source. That being said, I really would have likes some actual statistical information as these results are now being quoted as gospel in every tanking thread on these forums.

Files are still there

http://www.kamuf.org/parri/ThinkTank/Simulator.java.txt
http://www.kamuf.org/parri/ThinkTank/TankStats.java.txt
http://www.kamuf.org/parri/ThinkTank/TankBenchmarking.jar

Rauxis, chosen of CAT

but the values he used aren't, the gui has a place to put the low and high end damage, but we don't know the values he used there.

Nasty, parri asked me to point you to the updated site.

http://www.kamuf.org/parri/ThinkTank/

---

Q u o t e:


Firstly, Nastyknight. You either seriously need to learn to read or learn to add. I don't know where it is you're failing at so horribly atm. Firstly, the post itself:

"So we're going to say the boss does 22000 damage when swinging every 1.0 seconds, and scale that up to 55k damage at 2.5 seconds. Overall DTPS is the same, but distributed differently. "

Then the reminder I gave you earlier:

"Someone tell that DK who fiddles with the numbers on his copy of ThinkTank that he's using the wrong boss data information. The base damage for the first test was 22000-22001 damage. At 1.0 speed. As the swing time goes up, the damage goes up. So by 2.0 speed, the one that he's using, the boss should be swinging for a damage range of 44000-44001 damage. Not 20k-30k."

I really thought it would have been obvious first time around. But fine, people make mistakes. Second time around and you still accuse me of obfuscating figures. The only thing obfuscated here is your vision. If I say that a boss swinging at 1.0 speed is doing 22k damage per swing, and I'm going to scale it up to 2.5 swing timer and have it doing 55k per swing, surely it's obvious that the damage per swing is increased by 10% of the original damage as the swing speed is slowed by a similar amount? If it wasn't obvious, I flat out said it: "Overall DTPS is the same, but distributed differently." How much clearer can it be? 1.0 is 22k damage, 1.1 is 24200, 1.2 is 26400, etc, etc. The expected DTPS remains the same, but the distribution of the damage is altered so we can see the effect of Block on faster, weaker hits. I believe I stated that too. Please read it this time.

Sal: I'll admit straight away, I'm no statistician. I'm not even a particularly good with maths. However I can understand the value of a stated margin of error or some form of standard deviation to frame the results in, but in this case, how are they at all applicable to the TTL example? I can perhaps follow the idea that some tanks having more avoidance than others and higher cooldown uptime than others can make it less likely for them to recieve burst damage. But I already do track that. It is in the result spreadsheets which are listed for each test that I ran. Did you read them? Or is something else missing from them you want to see?

The graph was even labeled as the Lowest TTL. It's not like I have tried to deliberately pass off disingenuous figures as something else. I'm sure you are aware of the Effective Health side of the tanking argument. The idea that a series of unfortunate and moderately unlikely events can chain together and cause a tank death. Which is a case of, multiple attacks in a row not being avoided. The graph was designed to, and successfully showed just how each of the tanks faired under these circumstances. Now, while it is moderately important to note the likelyhood of this TTL within the overall results, that's not the point the graph was trying to prove. The point being made was, everyone knew that Druids had a significant EH advantage over other tanks. But the massive HP gap and lack of a worst case scenario (their block/cooldown equivalent is armour and always up) combined made them at times up to twice as resilient to unpredicted large melee damage. While coding it, I did recognise however that some deaths were simply a lot more likely than others. So there is a tracker for the frequency with which a death at that speed was recorded. It's in the spreadsheet.

I'm all for including more statistical information if you think it's needed. The results don't scare me because I'm entirely confident in my original premise. The Druid and DK nerfs were well deserved and needed. If you like, contact me on MSN at cj.mackey_uk@hotmail.co.uk. Don't bother emailing, I never check the address. If you're able to post on the US forums and have noticed this page update, link to it or paste it please =P

---

---

Q u o t e:
I don't want to derail a good discussion, which blue comments tend to do. However, I did want to offer that we did fix those Naxx block items in question.

---

Nice!

Now I probably don't have to worry about my Hero's Surrender that doesn't exist :)

Thanks GC
I'll check my Wall of Terror buff :)