With respect to the effect of Weapon Skill on +hit, I'd like to understand better the "proof" that Crezax's post has been invalidated.

I slogged through as many pages of the Rogue DPS Spreadsheet thread as I could. I didn't see the evidence there.

I DID see evidence that, against an equal level mob, one point of Weapon Skill will give .1% +hit. But I don't see this as a refutation of Crezax's post, since his post specifically stated that it applies against a mob 3 levels higher. In fact, if anything, this new testing proves that our "old" understanding of how Weapon Skill affects +hit (i.e., a linear .04% +Hit per Skill level) is, indeed, no longer valid. The fact that the old system has been debunked should open our minds to the fact that a new model has been implemented. And we should try to figure it out, ignoring past assumptions as much as possible.

If you read the Wowwiki formulation (I know, I know, no flames please... just bear with me), the base miss rate for a level 70 character against other mobs when wielding a 2H would be:

v. Level 70: 5% to miss
v. Level 71: 5.5% to miss
v. Level 72: 6.0% to miss
v. Level 73: 8.0% to miss

Basically, wowwiki asserts (and, btw, this does not contradict Crezax's post), that if the difference between your Weapon Skill and the opponents Defense Skill is 10 points or less, than the formula for miss rate is: 5% + (Defense Skill - Weapon Skill)*.1%

However, if the difference between your Weapon Skill and the opponents Defense skill is MORE than 10 points than the formula for miss rate is: 5% + (Defense Skill - Weapon Skill)*.2%.

----------------------

All the discussion I've read so far on these forums (and I confess I haven't read everything) seems to want to find a single formula for converting Weapon Skill into +Hit. And the tacit assumption is that there is a linear relationship between Weapon Skill and +Hit. But I think that tacit assumption should be exposed and either confirmed or denied.

For me personally, I am perfectly willing to believe that the conversion rate is NOT linear. In fact, it would make sense to me that Blizzard would make a greater difference between Weapon Skill and Defense Skill have a more profound (i.e., non-linear) impact on missing. Indeed, Blizzard has already implemented such a non-linear approach when it comes to miss rates for spells. For spells, the different miss rates against higher level mobs are as follows:

v. Level 70: 4% base miss rate
v. Level 71: 5% base miss rate
v. Level 72: 6% base miss rate
v. Level 73: 17% base miss rate

Now I know that there are a lot of difference between spell damage and physical damage, so there are lots of ways to assert that this is an apples-and-oranges comparison. But the underlying fact that Blizzard employed a NON-LINEAR approach to factoring miss rates against higher level mobs remains true.

And, if they did it for spell damage, we should at least be open-minded that they may have done it for physical damage as well.

If so, then it is possible that BOTH of the following are correct: (i) our observed data that each point of Weapon Skill against a Level 70 mob equals an additional .1% hit, AND (ii) Crezax's post that says against a Level 73 mob a point of Weapon Skill is equal to an additional .2% hit.

All feedback on this post is welcome (hehe, even the Wowwiki flames!)

Draenei Paladin racial: Thottbot World of Warcraft: Heroic Presence ? I suppose they would have been in the same group though.

Sp00n (or anyone),

Can you please direct me to the proof that Crezax was wrong? (especially if there are multiple proofs of this.)

I've scanned these forums as best as I can -- and I've seen the data that suggests that each point of Weapon Skill gives +.1% hit against an equal level mob.

But I haven't yet seen data that proves actual hit rate (or reduction in miss rate) for mobs 3 levels higher. And Crezax's post was specific to mobs 3 levels higher.

Apologies if this point has been proven out and I've just missed the data.

Thanks.

I disproved that in another thread around here somewhere. vs a 73 mob

Here it is

http://elitistjerks.com/377783-post49.html
and
http://elitistjerks.com/378234-post35.html

What would you think about testing the miss rate on the immune mobs in shadowmoon with a level 67 rogue? I don't see why it wouldn't work the same way as a level 70 vs 73/boss. And if it's a rogue you can run some 100k attacks naked and then some 100k more with 360 weapon skill. Just make sure you max out your weapon skill with daggers (which is fastest I assume).

If anyone got a rogue in range this would be useful I think.

Good reference. Thank you. Unfortunately, the WWS data was expired (when I tried clicking on the links). But I still hear your point.

I'm gonna try a test of my own.

I'm a Dwarf Hunter. Which means I have a +5 Weapon Skill to Gun. According to the Wowwiki theory (and the Crezax theory for that matter), I should therefore only need +6% to hit to never miss.

So, I've intentionally gimped my +hit rating (by swapping out my +hit gems and putting in some +crit ones instead). My Hit Rating in DPS gear is now 100 (6.34%) and in my mana regen set (which I uses on long boss fights) its 98 (6.21%). I'm gonna run like this for a week, including raids to Gruul's and Kara and see what happens.

Yes, but we have confounding effects here. The weaponskill boost cuts you down (functionally) to a one level difference, if it's based on a weaponskill/defence comparison. If there are nonlinear effects going on, that single data point isn't really generalisable. What it tells you is that 10 points of weaponskill give an average of 0.1% to hit each, giver certain assumptions about base miss rates, and about how the hit calculation is made.

However, the latter is actually the point at issue! One of the major holes in our understanding right now is relating the 1H calculation to the 2H calculation.

In the 1H calculation, there is a term for the absolute level difference and *also* a term for the weaponskill/defence comparison. These two of course are inevitably confounded to some extent in all our measurements.

What we need for a full understanding is:

* A set of parses at 1/2/3 level difference, with no added weaponskill. This gives us the confounded effect of level difference and weaponskill/defence comparison.

* A set of parses at 1/2/3 level difference with some added weaponskill. This gives us pure measurements of the effect of weaponskill, which we use to deconvolute the previous data.


It looks like absolutely the best way to do this would be to take three rogues of level 57/58/59 down to Blasted Lands and have them beat on a level 60 Servant. Then take one point out of weapon expertise and repeat the exercise. Then take the second point out of weapon expertise and repeat it again.

Does anybody have suitable alts they could use to do this?

Songster, the servants are not even close to level 70...
The corruptors in Shadowmoon valley, however, are.

Brainfart, edited. Of course I meant the level 60 servants, and of course I meant level 57/58/59 rogues.

I can't think of any other decent way to get long parses with 3/2/1 levels of difference, without needing a whole infrastructure of healers behind you, and an endless supply of identically-levelled mobs. Not 100% convinced by the corruptors in Shadowmoon, if I remember right the behaviour is different - they don't move, don't attack, don't anything until the totems are down, so I'm not sure they act as real level 70 mobs.

But those test always had weponskill either as talent or on items!

I made a test as a 41/20/0 mutilate spec a few days back against Hydross.

The missrate against this boss was about 5.5 % (278 hitrating without any weapon skill items/talents).

Yesterday I respecced to a 15/41/5 but with only 1 point in Weaponsexpertise and the missrate
agains Morogrim was in line with the "usual wisdom". Unfortunatly i forgot to acitvate combatlog.
I only had Recap running .

I will try to push my rep at lower city to get Shapeshifter's Signet. Then I will repeat the test with 41/20/0 + that ring.

At the moment I tend to beleive that there is a discontinuity with +hit depending on difference between
weaponskill and bossdefense.

Well, if 308 + Talent + Weaponskill means that you don't miss against level 73 mobs anymore, the following conclusions can be made:
(308/15.769)+5= 24.532 to hit.
With this amount and the pre TBC assumption of a miss rate of 24% + 0.2% per level = 24.6% against a level 73 mob, you would very rarely miss.
310 hit rating would be sufficient to never miss a single swing (24.659 tohit).

Obviously this wasn't the case, and level 73 mobs are therefore presumed to have a 25.5% chance to be missed, because it plays so nicely with 24% + 0.5% per level difference and 0.1% per weapon skill.
You are right that we can't be sure that 1 weapon skill does indeed give 0.1% chance to hit in any cases against all mobs, so what we'd really need is:
Somebody to never miss against a level 73 mob without any weapon skill.

If the miss chance is at 25.5, then (25.5-5)*15.769= 323.2645 hit rating is required to never miss.
Or rather, with 323 hit rating you would have a very low chance to miss, and with 324 you shouldn't see it ever.

So after that we have determined the actual miss rate against a level 73 mob, and now we can more easily determine the amount of hit 1 point of weapon skill gives (or rather, 5 points and 10 points).

Do the same test with 1 point in weapon expertise, and with (25.5-5-(5*0.1))*15.769= 315.38 => 316 hit rating. If you miss with 316 hit rating, 5 points in weapon skill (or 10 skill below the assumed defense value of the mob) does not give you 0.1% to hit, but less.

This is a very long and tedious process which I certainly won't take part in.
I suspect you could do the tests with level 57-60 rogues, you'd have to adapt the hit rating conversion though.

So, suppose we say Crezax is right, and further clarify that he meant the *first 4 points of +skill*.

If I'm rocking 367 skill (weap ex, Fang of Vashj, Shoulderpads of the stranger), my skill is above the defense of a level 73 raid boss. First 5 points are giving me .2% each, next 10 are .1% each, last 2 are? The old .04?

How about dodge and parry? I wonder if I should ditch the stranger shoulders and pick up T5.

Same here, I'm starting to regret picking them up if all this is true.

The 27% against lvl 73 seems to be ruled out by the data I posted here.

278 hitrating would imply 4.37% miss chance with 27% base miss.
113hits in 2023 attempts gives 5.586% miss with stddev of 0.2297.
The 5% confidence interval for normaldistributed samples with these values is
[4.584% 6.586%] (mean - 1.96*stddev/sqrt(N), mean + 1.96*stddev/sqrt(N))

I am not sure if we realy can use the confidence interval for normaldistributed data, so this is to be taken with care.

Unfortunatley I dont know how to find the interval for uniform or binomnal distributed data.
Maybe someone can enlighten me.

Well I did tonight. 4 hours fun with Morogrim and Magtheridon.
316 hit, dualwielding daggers, 15/41/5 with 1 talent point in WeaponsExpertise

546 crits, 958 hits, 0 blocks, 524 glances, 0 misses, 134 dodges, 13 parries,
2175 total attacks
696007 damage, 320.003218 damage per swing
272574 hit damage, 284.524008 hit damage per swing
111788 glance damage, 213.335878 glance damage per swing
25.103% crits
44.046% hits
0.000% blocks
24.092% glancings
0.000% misses
6.161% dodges
0.598% parries

So with one point in wepons expertise the values for 25.5% base miss and 0.1% hit/skill
seems to be accurate.
But with 278 hitrating and no +skill I had a much higher missrate than predicted.

And besides that I am not sure if +skill on items and weapons expertise behave identical.
I need to get hold of that Shapeshifter's Signet with +20daggerrating

For large data sets - and 2023 trials certainly qualifies - a normal distribution is a very good model for the probability distribution. So I don't think you need to worry about that aspect.

I know you're gonna hate me for saying this. But this test could also be cited as "proof" of the Crezax/Wowwiki formulation. Here's how:

Your one point in Weapon Expertise gives you +5 Weapon Skill, so, when you are fighting these mobs, it is as though you are fighting a Level 72 mob. Your base miss rate against a Level 72 mob is 25% (according to the Wowwiki formula).

You have 20.04% +hit from your gear and +5% to hit from Precision, or a total of 25.04%. Which is exactly the right amount you need to witness zero misses, just as you have done...

Look, I'm still not sure I believe Wowwiki either -- I'm just trying to point out that this last data run of yours could be interpreted in another way.

OK, here are the early returns from my test so far. I'm going to continue to run this test, but I thought the early returns were interesting enough to share.

I ran Gruul's this evening equipped as I described above (i.e., I never had more than 6.34% +hit gear equipped).

Here's the results:

Steady Shot: 358 hits, 99 crits, 0 misses
Auto Shot: 382 hits, 104 crits, 0 misses
Arcane Shot: 19 hits, 8 crits, 0 misses
Multi-Shot: 3 hits, 4 crits, 0 misses
Serpent Sting: 4 hits, 0 crits, 0 misses

Total: 766 hits, 215 crits, 0 misses

So, in total, I fired 981 shots and never missed a single one.

Now, 981 shots is not a huge sample size. But according to the "old pre-BC" theory, my miss rate against a Level 73 mob should have been at least 2.06%. (8.6% bass miss rate - 6.34% - .04%*5).

And even according to the new ".1% per skill level" theory, my miss rate should have been at least 1.76% (8.6% - 6.34% - .1%*5).

So, I should have seen something like 17 to 20 misses, give or take. Instead, I had ZERO.

And even if you assume a base miss rate of 8% (instead of 8.6%), you'd still expect to see something like 11 to 14 misses.

Now, I'll continue to run this test -- we definitely need a bigger sample size. But the early results definitely confirm that the "old pre-BC" theory of +hit is no longer valid. And they also seem to suggest that the ".1% hit per skill level" theory is also not valid. So far, the only "theory" that could explain these test results is the "Wowwiki/Crezax" theory...

This is still a small sample set. More data to come.

Olgas, you didnt happen to have a moonkin with imp faerie fire did you?

With regards to melee hit, I am 0.10% away from hit cap on a lvl 73 mob, with 296 hit, precisions, and 16 skill. On roughly 40 archimonde runs, I missed, on average, 0.11%. That, to me, is pointing at things working how we think they work.
I do see the hit cap at 25.5% vs a lvl 73, dual wielding.

Once again this is a situation where you already *have* weaponskill, so you'll be subtracting out the exact effect we're trying to detect.

Do you have enough hit rating (324) to run this again (oh noes!) with 0 points in WE?
Did you at least succeed on those two bosses?

Yeah I think I can replace 2 glinting noble topaz with 2 rigid dawnstones, netting in 324 hit....
Sadly not . Morogrim was our second night out with We managed 64% but healing for the MT was thin.

Magtheridon we had a streak of "human errors" After 3 7% wipes we called it a day. But we already
down him easily in previous weeks. It is always something with those guys that have to press the cubes :/
It is not our favorite boss at the moment..