This is sort of starting to grow into its own thread topic...

World of Warcraft - English (NA) Forums -> Second round of One/Two roll research

Is that good enough to go off of? Cloak of Shadows provided a good test bed for it under the assumption that PvP/PvE use the same roll system, and Cloak of Shadows acts like a negative hit debuff.

Yep, that was the Crit Gain from talents and Ice Lance coefficient test. World of Warcraft - English (NA) Forums -> So I chaincast Ice Lance on the PTR today... Irontrygress approved. I never was able to find a way to predict that crit gain though.

Plugging those points in gives an equation somewhere around y=(1/13)x+0.9, but we need more #s to get the exact m and b. x is chance to crit, y is crit reduction on the boss. So somewhere around 1% crit chance, the boss has full reduction, and for every ~13 after that, you'll see another 1% crit chance reduced. This seems to just be a straight devaluation of 8% or so on crit rating when gearing.

So, I repeated my test hit-capped against a boss level dummy with 8.48% crit rating and this time remembered to take a screenshot of my character stats.

Here's the character sheet:



Here are the results for this round of casting:



I am going to run another set of data later in the week because something is obviously different from the last set of casts.

Thank you for the data. I think it is obvious that there is some sort of crit suppression going on. I have yet to see a set of data that has a higher crit percentage than the % listed in the character sheet. It would make sense that with 20+ sets of data, 10 would be slightly higher and 10 would be slightly lower, and that doesn't seem to be the case.

Has anyone seen a HIGHER crit rate than their character sheet percentage?

This was probably answered somewhere in the thread but I didn't see anything going through most of the pages... why the 2 points in Imp. Fire Blast if it doesn't get used in the rotation?

It's a filler talent to get down the tree, since the reduced casting time on fireball has no effect on the frostfire build. However, the 3 points in Incineration actually do help a little (6% crit to scorch).

Assuming there is crit reduction, what form does it come in? I can think of a few things to look into:

- It's based on overall crit, so the source of the crit doesn't matter.
- It's based on crit rating, but not other sources of crit. If this is the case, then Molten Armor and crit talents shouldn't affect the results.
- It's partially based on Int.

Keep crit rating the same while varying crit through talents/Molten Armor should be a fairly easy experiment. Do the same but vary Int would be a little more difficult if the tester has a lot of crit gear, but not a lot of gear without crit.

Those ideas assume that crit reduction is intended by blizz, which I'm not sure Blizz would bother adding. What would be the best place to ask Blizz directly?

Not sure if it's been discussed someplace in the forum, but could I get a lil detail as to what would be better for cloak for enchant purposes: 23 haste or lightweave emboidery(I am a tailor) I've heard both scale with gear, but which is the better to use?

I think the tailor one is going to be much better in 3.1. It's changed to 250sp for 15 seconds.

Ran a crit suppression test with current raid arcane gear. Including t7 4 piece set bonus, with tweaked talents and hit adjust to hit cap (on live server);

Spec: Arcane 57/3/11 NO Arcane Potency & Molten Armor
Glyphs; AM,AB,MA
Hit Cap (exact at 289 with talents)
Paper Doll Crit Rating = 23.21%

5000 Arcane Missiles
Recount Crit = 21.1%

Prety much in line with the 2% so far. Noted that the crit % ramped up as I continued (at 2.5k casts it was 17.9%).





Hope this is of use. Can provide combatlog file if required.

The most recent test I did was 6.75 deviations off.

I'm kind of trying to characterize it now and am sort of thinking that if enough people get enough tests done, we could plot a graph of actual crit vs. expected crit and see a trend.

Sorry if I wasn't clear earlier. I was solving for a rough best fit line between those first 3 data points ("The drop measured so far is 2.5% at 23% paper doll crit. At 16% crit, Zaldinar saw a 2% drop in crit. At 8.48% paper doll crit, I saw a 1.4% drop in crit.").

I've got more tools available now, and I was slightly off. Plotting those 3 data points in excel and adding a linear best fit line gives an equation of y=0.0758x+0.7669. I added in your last post, luna's post on this page, Shaewyn's Test 3 (4 was already in it on the last page, and 1 and 2 were flawed), and that gives us the following graph of paper doll crit chance vs. observed crit reduction:



The one notable outlier currently is your last data test observing 2.3% crit reduction on an 8.48% paper doll crit chance.

This Crit suppression discussion intrigues me. I have another mage, Exactly which I am levelling, currently at level 73, about to reach 74. I thought it might be interesting to see the effects as she increases in level up to level 77. (I am leaving her at 77 for these kind of investigations)
Some results of Target dummy testing at level 73, Arcane missiles on level 70 naked, level 80 naked, level 80 geared.
Buffed with mage armour when I remembered to put it up, nothing else.
Relevant talents are Precision.
I have CombatLogs for this run if anyone is interested, but nowhere to put them.

I am about to ding to 74, would it be worthwhile doing more tests before I do that? I would prefer not to respec but will happily do that if it's felt that the information gained would be useful.
Edit: I'll get some data on the Boss level target dummy too. Might also be interesting to see what happens on the level 60 dummy. At this rate it is going to be quicket to level than to test. Oh joy.


I am seeing severe Crit suppression in this. Would postulate that it is level delta dependent, the data I hope to generate might produce the equation which can be extrapolated out to Boss level.
The whole subject has rather significant ramifications. Especially considering Crit dependency of different specs. Is this part of where Frost PvE is "ok" I wonder.

Over in the hunter forums, we did a lot of testing on crit suppression vs the boss dummies and found that it definitely does exist for physical damage. In our testing, the higher the hunter's paperdoll crit rating, the larger the amount of crit supression; therefore the effect was more noticeable with a higher crit rating. (eg, It was not noticeable at all with a crit rating of 10%, but very noticeable with a crit rate of 30-40%.)

The culprit was blocking by the boss, and it only happened when attacking form the front. There appeared to be a two-roll system where blocked shots crowded out crits; there may be a similar mechanic with crits and resists? I realize that magic damage mechanics are quite different from physical mechanics, but here is a link to our thread on the topic in case anything we did sheds light on your efforts: Focused Aim reducing Crit Return (The thread's title is a bit of a mis-nomer -- we originally thought the crit suppression might be due to problems with our hit-enhancing talent Focused Aim, but that did not seem to be the case.)

Most of the conclusions are near the end of the thread -- we did not arrive at precise numbers for crit supression as we suspended testing until 3.1, but there are preliminary results.

EDIT: I should clarify that the crit suppression I am talking about here is above and beyond the 4.8% reduction in crit physical damage receives vs a boss that is caused by the difference in levels. This crit reduction for melee physical damage was demonstrated by Vulajin in the rogue forums, and was amply confirmed to affect hunters' ranged physical damage as well. I was under the impression that there was a level-based reduction in crit vs a boss for spell damage, too, in the neighborhood of 3% -- but maybe this was never confirmed for WLK?

This might also explain why we didn't really get a lot of research on this before, if the effect is significantly smaller at, say, 30% crit rates we saw in BC vs the 60% rates we routinely see now. It would have been easier to dismiss the observations as noise.

I'm convinced you guys are on to something and yes, the next natural thing to do is significantly vary the crit rate while keeping other aspects constant, to see if we get a trend. It would also be a good idea to get at least 3 data points, to do some curve fitting and guess whether we're dealing with a linear, geometric, exponential, log-scale effect etc.

3 "tries" doesn't give us any sort of sample set to work with. Your crit on a boss will be very inflated due to a number of factors so we are trying to test this with as many samples as possible.

If someone out there has a guild that effortlessly kills Patchwerk, it might be worthwhile to strip down naked and spam AM the whole kill to see what sort of crit you get on a real live boss. Hopefully the crit suppression we have seen so far is a result of the target dummies and doesn't translate over to real bosses.

You're not gonna get much sample size over a 3 minute fight anyway. If you actually followed the discussion, you'd see you need thousands of casts to get somewhat reasonable results.

I have been following the thread and I understand the statistics involved. I also know that a targeting dummy is not a boss. Even GC has said he would rather see actual fight numbers over target dummy numbers for any performance based accusation. If anyone has a group of friends that are dedicated to theorycrafting, take a group in and do AM's for 6 mins until he enrages and do this 2 or 3 times.

Any mage can get 200+ Arcane Missile attacks in a 3 min fight. If 10 people do this twice we have a nice sample set of 4000 data points. This will not give us conclusive data, but it will give us something to compare to the data we already have to see if there is any consistency.

The reason for preferring real bosses over target dummies is not because the "boss" target dummies don't obey boss mechanics, but because the dynamics of a boss fight are almost always going to be different from standing in one place DPS'ing a stationary target that doesn't do anything. Thus, target dummy DPS tests often are inherently a somewhat flawed measure of actual DPS performance in a raid environment. For testing of specific mechanics on the other hand (like hit rate, crit suppression, or how various buffs add/multiply), the target dummy should be fine.

The best you are going to get are the target dummies. You need a static target that you can make a control test from. Raid Bosses are terrible because it can take months because you need the same raid to do the exact same instance multiple times.

As some extra detail on my tests, I fight until my mage is 100% OOM. This typically yields 80 frostfire bolts each round.

Also, I don't use target dummies to determine my raid DPS, but my minimal raid dps.

target dummies usually show your best NON-raid dps, unless you have a bunch of buddies tossing all the buffs and debuffs.