Hmmm. Has there been any testing regarding the two-roll mechanic of yellow damage attacks and crit depression at all? Are we even sure that yellow damage is actually on a two-roll now, or was it simply the crit depression kicking in back when the testing was done?
Do we have an idea how level difference affects crit depression (i.e. is there a "jump" just like for the miss rate)?


I cannot remember exactly anymore, but as far as I do, the two-roll mechanic was tested with mobs that place a -hit debuff on you (like the trash mobs before C'Thun). I think those were above level 60 (61 or 62), so if crit depression isn't only in effect with +3 mobs, tests against these mobs would have been affected too, effectifly skewing any data due to false assumptions.

This is the thread where the two-roll theorey was discussed. Interestingly, there were some talk about a one-roll mechanic with some sort of cap, which was dropped due to crit depression not being known (or even considered) at that time.

Also, there has been quite some changes during these 3 years since the initial testing (e.g. the dodge rate was assumed to be around 5.x%, expertise was introduced, miss rate was lowered), so how sure are we actually that our current working theory is still up to date?

Unless I'm missing something obvious, a single noncrit Mutilate hit with the Professor's Choking Gas (-75% hit) up from a remotely decently geared rogue should be enough to disprove one-roll yellow hits, yes? In which case, I present the following snippet of combat log from one of our Putricide attempts last week:

There's a bunch of other stuff in there as well, of course, but these are sort of the relevant 4 lines. So... yeah. I'm sticking with two-roll for now.

Hm, unless I'm missing something obvious now, that only proves that either two-roll or crit depression was kicking in (or both, of course).
It doesn't really show which of those, does it?

It's been shown in the crit reduction thread that it's possible to achieve 100% crit on special attacks - as in, it doesn't have the same enforced 4.8% regular hit that white attacks do, via testing of Ambush/Overpower above 104.8% crit rate. So assuming your crit is high enough to push all regular hits off the table (which, with Mutilate and a 75% miss debuff, it is), your attack table would consist of nothing but crit and miss. Hence a single hit disproves the one-roll theory.

I might be wrong here, but if we're strictly talking about white damage, I'm fairly certain some of your assumptions are incorrect.

1) You use the term "Glancing etc (zero from behind and exp capped)". Considering the make up of your listed combat table, I assume "etc" covers Dodge, Block and Parry. While it's correct that standing behind the boss at 26 Expertise will eliminate Dodge, Block and Parry from the combat table, it'll have no effect on Glancing Blows. When it comes to white damage, you'll always experience Glancing Blows.

2) Your combat table entails both a 4,8% crit reduction and 4,8% guaranteed hits. As per the discussion in this thread, notably post 12 and the posts that follow, I was under the impression that the the Forced Crit Conversion was to be understood as 4,8% of your crits being converted into hits - and not understood as your chance to crit being reduced by 4,8% prior to building the combat roll table.

Thus, if we gear all the way to the Crit Cap, reach 26 Expertise, reach 722 Hit Rating and hit the boss mob from behind, our crit chance would 100%-24% (Glancing Blows)-4,8%(Forced Crit Conversion) = 71,2%. In this light, we do have to gear all the way to the cap in order to "only" suffer the 4,8% Forced Crit Conversion. Using the same assumptions regarding Hit Rating and Expertise, if your gear brings you to, say, 5% below the cap, your crit chance would effectively be 95%-24%(Glancing Blows)-4,8%(Forced Crit Conversion) = 66,2%.

As much as anything, this is to figure out if I've actually understood the concept of the Crit Cap correctly. Can anyone confirm or refute the points I've made?

The glancing line is poorly phrased - Glancing is a fixed 24%, but you can (and usually do) zero out the other effects (or close) through positioning and expertise.

I'm pretty sure that everyone is actually saying the same thing here - the language of crit capping tends to be very confusing, and everyone explains it a bit differently, but usually they come back to the same thing in the end. So to be perfectly clear on the situation for anyone still confused:

1) Your actual, observed, crit rate on bosses is 4.8% lower than would be predicted by your gear, spec, etc. That is, if your tooltip crit is 54.8% and there are no crit-increasing debuffs on the boss, you will in practice only crit 50% of the time.

2) No matter how much crit you have, you will always have regular hits - not crits, not glances, not anything else - for 4.8% of the hit table.

There are many ways of explaining this - see this post for a comparison of two of them - but those are the essential facts. And what this means in practice is that you are crit capped when your theoretical crit rate (tooltip + crit debuffs on the boss), which we abbreviate C, satisfies the equation 100-D-M-G = C; where D, M, and G represent your dodge, miss, and glancing chance respectively. At this point, your hit table consists of C-4.8 crit, 4.8 hit, and D+M+G other stuff, for a total of 100%. If you have C<100-D-M-G, you are not crit capped. If you have C>100-D-M-G, you're losing white crit to the crit cap. And if you're at C=100-D-M-G exactly, you're exactly at the crit cap - further crit does nothing for your white attacks, but you're not wasting any of the crit you have, either.

That's exactly what it means (see the thread I've linked). The problem is though that we are talking about yellow damage, so you cannot simply afk while testing, you have to actively press buttons all the time. And with the relatively small difference in %, the sample size cannot be too small either (I'm not good enough at statistic to tell you how much attacks we'd actually need, but I assume at least a thousand?).


Yet, testing this now is way easier than it used to be, before the implementation of training dummies.

I do apologise, I think my thought processes went along the lines of 'things that aren't hits or crits, some of which can be zeroed, but whose actual values don't matter over much'. My point was more clarifying the role of the 4.8 value in calculations.

Actually if you have a G15 keyboard its not too hard to test. You can just make a macro that pushes your SS+stopattack macro every 4 seconds and let it run overnight. That would be 900 sinister strikes per hour, so like 7,200 in an overnight log. The only danger is that some random guy puts debuffs on your dummy while you are not watching.

Here are the results of some basic testing I performed this evening.

First, to test under the assumption that yellows work the same as white hits:
Talent Calculator - World of Warcraft
Talent spec used

Google Docs image link of my stats
I have zero hit, which means yellows will have an 8% miss chance. (No points in Precision either)
Zero expertise means 6.5% dodge chance.
47.43% paper doll crit rate (including Malice, Master of Anatomy, Guru's Elixir, Drums of Forgotten Kings and Drums of the Wild)
5% CQC crit (since it is weapon-type specific, will not be included in the paper doll)
50% Improved Ambush crit

That should put me on a total of 116.93 combined crit/miss/dodge, or well over the 104.8 required.

World of Logs - Real Time Raid Analysis - Click on Damage by Spell
That then shows me hitting, twice, with an Ambush.

Testing two-roll system:
Here, hit and expertise are immaterial. All I need is a crit rating over 104.8 and watch for a hit. I use the same talents as listed above.

Google docs image - stats
Minor gear changes put me on 49.97% paper doll (including crit food and elixir, and the non-elixir buffs listed above)
5% CQC
50% Improved Ambush

Total: 104.97%

World of Logs - Real Time Raid Analysis - Click on Damage by Spell
That then shows me hitting again with an Ambush.

So, even if Yellows are on a two-roll system, the crit depression either applies and is not removable, as with whites, or it is larger than 4.8%.

I believe that this is in contradiction to mentioned (but never linked) tests with overpower way back when?

The remaining tests would then be the following two:
Test 1 - Hit Capped, Expertise Capped, crit rating of say 50%ish, keep doing any yellow attack until we have enough sample hits to say with confidence what the crit depression is. This can then be used to retest crit depression removal speculations if it does prove to be larger than 4.8%.
Test 2 - Zero hit, zero expertise, again a crit rating of say 50%ish, and then using the difference between these results and the above test to predict whether yellows are one or two roll attacks.

Anyone with more stats knowledge able to calculate the number of hits required for a reasonably confident result set? And would it be favourable to select some other crit rating?

If I've made any mistakes, please point them out.

CQC appears in the paper doll. So in your second test (for instance), you have 49.97 paper doll + 50% for Improved Ambush, -4.8% crit reduction, for a 95.17% crit rate; hence, per the two roll theory, of attacks that hit, you should see 4.83% regular hits. So the fact that you see one is, well, not even remotely surprising.

Similarly, in your first test, per the two roll system you're at 92.63%, so we predreict, based on your dodge and miss rates, as a percentage of all attacks:

6.5% dodges
8% misses
.855*.9263 = 79.2% crits
.855*.0737 = 6.30% hits

So again, you should still see some regular hits.

Basically, you haven't demonstrated anything that isn't 100% consistent with the standing theory.

Searching through this thread I couldn't find the answer, so I have a question regarding the assumed 27% miss rate vs L83 mobs.

Autoattacking on the boss dummy with MH weapon equipped and OH empty, with 576 hit rating I was unable to get a white miss attack after multiple 500+ quick parses.

Alternatively, with my OH equipped and MH empty, I was observing white misses almost immediately and averaging the expected 4.5% miss rate (5/5 precision).

So is the 27% dual-wield miss rate only occur on the offhand? Or am I completely missing something simple here? If that is so, what is the miss rate for mainhand? Does this affect the weight of talents for combat such as combat potency, or affect mutilate for which dagger goes in MH/OH?

With a MH equipped and no OH, you're not dual-wielding. So your miss rate is only 8%.

I recall from my testing a long time ago that if you have a MH with no OH, it counts as single-wielding, but if you have a OH with no MH, it counts as dual-wielding.

Just did a simple crit cap test on a heroic dummy:

Character stats: hit=0, expertise=0/0, crit=50.94%

Results:
Total auto swings: 10956
Crit: 4718 (43.1%)
Miss: 3002 (27.4%)
Glance: 2548 (23.3%)
Dodge: 688 (6.3%)

Well, I expected ~4.8% regular hits but found zero.

Need other people in US or EU servers to verify this.

Maybe the enforced "4.8% of your attacks must be hits" is not in effect untill your Miss+Dodge < 4.8%

Just asking to make sure, but did you attack from behind the target? If not, check your logs for (partial) blocks. I've seen people in other threads forgetting this, the hits were nicely hidden by blocks.

Well there aren't any parries, and with 0 expertise he should've received around 13-15%. So he was attacking from behind.

Additionally, the 0 hits encountered complies with the newer findings you're referring to, namely that there isn't a fixed conversion of 4.8% crits to hits.