sp00n

Of course, but Therasiv was also at 365 weapon skill, so he should face the same crit rate as against an even level mob. At least if the formula for crit is similar to hit, where level difference itself doesn't matter, only weapon skill to defense skill ratio.

 

• Aldriana

Yes and no. The weapon skill should make it equivalent to him attacking a same-level mob, meaning the the crit% with 365 skill and a level 73 opponent "should" be equivalent to attacking a level 70 mob with 350 skill. The difference being: his tooltip crit rate reflects the increase in crit chance from the extra weapon skill, so at the very least we should be looking at his base crit chance *not* including the bonus from weapon skill (i.e. 5.37% rather than 5.97%). And, weapon skill being the psychotic stat bonus that it is, it's entirely possible than something totally different and bizarre is happening instead.

 

Herb

If this indeed implies a 2-roll theory for white attacks, wouldn't all partial blocks count towards the "successful 'normal' attack" number as well (as they would for yellow attacks, since blocked crits were one of the strongest hints towards the theory in the first place)?

The parser used doesn't differentiate between partial and full blocks, and wouldn't even notice blocked crits, as far as I can tell - in any case, the ratio of "critable* : total" attacks would rise, and so would the relative crit rate, up to

3933 hits + 396 block + 196 crits = 4525 successfull 'normal' attacks ("critable")
(4525 / 8682) * 5.25% = 2.74% crit rate or 237.5625 crits

That would challenge the "not very far off" statement a bit.


*) "critable" as in: attacks that could or did crit under the 2-roll-theory (sp00n's "successful 'normal' attacks)

 

sp00n

As far as I can remember, blocked crits didn't play a major role in the theory of the two roll theory. I do remember that someone wondered why his special attacks had a lower crit rate than expected.
I just found the thread, it's here: http://elitistjerks.com/f33/t9104-backstab_two_rolls/


However, yesterday over 2555 attacks on Morogrim with 350 weapon skill, my crit rate for white attacks was already too high for a two roll mechanic to be at work, even considering all possible procs (dual mongoose), so this seems to be not case, at least not now.
Could need further testing in a controlled environment of course.

So we are back to either a) combat mechanics were different on the PTR or b) as Aldriana proposed, crit has a (totally) different mechanic than hit in regard of weapon skill.

 

◊ Rob

Hey. I am not that great at parsing logs and doing math, but I happen to have an interesting setup right now. I'm an Enhancement Shaman and I've got [Belt of One-Hundred Deaths], so my only +weapon skill comes from that -- and it only applies to my main hand, since I MH the MG Pummeler and OH the MG Cleaver. Would you guys want a log from my next raid to examine?

 

• Aldriana

I don't think that would be particularly informative, myself, since there's no real way to distinguish between attacks made with MH vs OH, so it'd more or less be a log of "this is a bunch of attacks, some made at 350, some made at 356" which is... hard to do much with.

 

◊ Rob

I could get you something on a farm boss with a blue mace from my bank, if it will be useful.

 

Karmon

I managed to get some more data for 353 weaponskill, dualwielding daggers, precision.

According to the modified wiki formula you need 351 hitrating to never misss with 353 skill
and 5% precision:
(24 + (365-353)*8/30 - 5)*205/13 = 350.076 (the -5 is for precision)

So I ran Letheras with 353 skill and 351 hitrating.
Out of 2542 autoattacks no misses. So far so good.
Yesterday I run Karazhan with 350 skill to see if I could get a miss.
Out of 3989 autoattack also no miss.
This is not yet enough data to refute the modified wiki therory.
350 hitrating is only 0.005% less than the expected value so I need about 20000 hits to see one miss.

What it does show, that the new formula that sp00n proposes cannot be true.
According to that forumula you would need 354 hitrating to never miss
(25 + (365-353)*0.2 - 5)*205/13 = 353.53
The difference between 350 hitrating and 354 hitrating is about 0.2%. With about 6500 attacks
one should see a miss if that formula is true.

 

sp00n

Well you should have seen a miss with 6531, true, but randomness has no memory, so you could just have been unlucky. Or maybe there was a moonkin with his improved faerie fire (can you check?).

However, it is more likely the way you see it, and from my own testing I also suspect 0.2% to be too low at the moment. The difference should be high enough to have encountered a miss so far.

 

Karmon

No moonkin or draeneipala in my groups.
I doublchecked each time for hit buff/debuff

 

sp00n

Ok, then it seems quite unlikely to be 0.2%.
Did I mention that I like 'nice' numbers and that always the 'nice' formulas are disproved?

 

Karmon

Yes you did
But look at the Agi to Crit conversion for different classes.. No nice numbers at all

 

Punscho

Agi per crit was nice at level 60.

 

Systim

Now that we've confirmed the CM post regarding the effect 4 weapon skill has on miss, dodge and parry, would it be prudent to assume that dodge and parry keep the same ratios against miss at 5 weapon skill and up?

Ex., 4 weapon skill grants: .8% anti-miss, .4% anti-dodge, and 2.4% anti-parry against bosses.
5 weapon skill grants 3% anti-miss, thus 1.5% anti-dodge, and 9% anti-parry.

Could that be correct?

 

• Aldriana

It could, but it seems unlikely, given that the paladin testing indicated no discernable effect on parry chance with 15 weapon skill, so getting 9% from 5 weapon skill seems out of the question.

 

Beliandra

Just had a look at the numbers from a night of Gruul attempts and finally a kill. With 360 dagger skill (Weapon Expertise and nothing else boosting it), my dodges were 7 out of 191 for Backstab (3.66%) and 77 out of 1461 for normal melee (5.27%)

Is there a reason why Backstab would get dodged a lot less? Or do you reckon that would just be the smaller sample size and a bit of luck?

Anyway, the total is 84 dodges out of 1652, or 5.08%.

 

sp00n

Erm, it was not confirmed. It cannot be 0.2% per point of weapon skill, else either Karmon or me would have seen a miss by now.

Myself, I'm going to check for a 0.4% miss rate. Another nice number.
With 351 hit rating and 2 skill, I shouldn't see any misses if it was at 0.4%, and if it is 0.267%, I'd see one miss after about 450 attacks.


Well yes, 191 special attacks is quite low to determine an average number. The math wizzards in here can probably tell you by how much.

 

Gruntle

Yup, 7 dodges in 191 Backstabs is just luck, it's still consistent with a 5% dodge rate. The upper 95% confidence interval for this sample is 5.82%.

 

Gruntle

Well, for now I see no reason to think that the PTR mechanics are different based only on the paladin test. Since no other detailed tests of crit/dodge/parry vs weapon skill has been done the paladin tests are so far the only thing we can use to look at this. Therefore I believe that your alternative b) is what we should work with. If we find that some tests done on a live server to give very different results I guess the PTR really could have used a different mechanic.

As to finding out how crit works with level difference and weapon skill. Are we sure that specials and white attacks use the same crit mechanic? It would mess up the statistics if they were different, e.g. rogues and warriors use way more crittable specials than paladins do (I think?).

Based on the crits by Therasiv and Beaglej a tentative two-component theory of crits could be:
c_obs = c_tooltip - (defense - weapon skill) * 0.02 - (mob lvl - player lvl) * 0.9

This fits with both Therasiv's and Beaglej's results but ofc need to be tested a lot more. The best way of testing this would be to do a lot of swings with a low base crit rate (remember that a lower number of "successes" combined with a lot of swings make the statistics better) at different lvl differences and weapon skill.

edit: made a silly math error
edit2: and corrected a mistake in the formula

 

sp00n

No. Specials use a two-roll mechanic, and therefore an attack can only crit if it is not missed, dodged or parried (and maybe blocked, but I cannot exactly remember if blocked crits appear on mobs as well or only in PvP).

 

Karmon

Well special attacks follow a two roll system. So you need to adjust the observed critrate
with the hits:
c_obs = (1-dodge-miss) * real_crit (whatever real_crit is computed)

Sure that it is (weapon skill - defense)?

Ok lets assume the formula is as given and lets for the moment assume there are no +skill items.
Then it is possible to rearrange the formula as follows:
c_obs = c_tooltip + (defense - weapon skill) * 0.02 - (defense - weapon skill)*0.9/5
c_obs = c_tooltip + (defense - weapon skill) *(0.02-0.18)
c_obs = c_tooltip - (defense - weapon skill) *(0.18-0.02)
c_obs = c_tooltip - (defense - weapon skill) *0.16

looks much nicer to me anyway
Hm given that (1-dodge-miss) is about 0.95 it is entire possible that the factor 0.16 is lower e.g 0.14. And that is 0.04 + 0.1 per skill point.....

 

koaschten

This "coincidence" is just worth highlighting.

 

Gruntle

Doh, yeah it should be (defense - weapon skill) ofc, will edit my post.

Well you can't really assume that there are no +skill items. The formula you give here is something completely else than the one I give. The 0.16 per w.skill difference will give you an incorrect observed crit value for Therasiv. He has 0 in weapon skill difference, i.e. he should have an observed crit rate exactly equal to the tooltip crit, but he hasn't, it's way lower. I tried to present a theory which fitted with the data from both Beaglej and Therasiv (and it does).

The problem with crit is discussed a few posts up. The formula you derive here does not fit with the data I'm sorry to say.

The two-roll mechanic for specials, has that really been proven?

Another thing which is not included in the formula is nonlinear effects on crit by weapon skill, it's there in the +hit mechanic, I wouldn't be surprised if crit also has some kind of nonlinear dependence.

 

Karmon

Well if the original formula is (defense - skill) the 0.16 is wrong and this formula will result in different values.

Your original forumula accounts for leveldifference in two ways:
A "finegrain" based on defens-skill and a "coarse grain" based on level difference.
I personaly dislike that idea so this variation based on defense-skill alone.
Seems my idea does not fit the data. Too bad

AFAIK yes.

 

Gruntle

Well I agree that it would be pointless to include both a lvl difference and a defense - weapon skill term if one was just a multiple of 5 with the other. My reasoning is that the lvl difference can never be affected, i.e. you will always get lvldiff*0.9 less crits regardless of how much weapon skill you have. Hope that made it clearer, sorry for getting it the wrong way around with def and weapon skill.

I'd like to see a link to a proper test of the two-roll mechanic, the only one I could find was a test done with Backstab before TBC (which, what I could see didn't include any effects of weaponskill on the crit rate). It seems to me that this theory is just one of these that everyone talks about but is never really proven