Olgas

Our Magtheridon raid ended earlier tonight because the instance servers all crashed. So, I was only able to get in 586 shots in against Level 73 mobs before our raid was called off. But I still didn't see any misses.

So, that now brings my total sample size to 2909 shots on bosses, with zero misses.

I technically should probably push for an even larger sample size, but I'm feeling too impatient. So, I've already re-geared myself with exactly 141 hit rating (8.94%). I'm hopefully gonna be in our Gruul's run tomorrow night -- and I'm praying I see a miss.

 

ZiggyX

This was recently changed but I'm not sure what it does now. I heard rumors that say it's .02 hit/crit, and less chance for boss to dodge/parry/block, and lowers defense by a little (not sure how much). Can anybody verify this?
Also, 4 Wpn Skill Rating=1 Wpn Skill? I forget these things.
And is weapon skill ONLY useful on bosses or ANY mobs above level 70? Heard rumors here too saying it only comes into effect on bosses.

 

Gruntle

Well, people are trying to figure out how it works now, there are no definite answers yet but 0.02 to hit per skill is definitely wrong, the effect of weapon skill is almost certainly nonlinear (i.e. the 5 first points are more worth than the next 10+). For crit, testing (see post 61) showed that weapon skill gives you
but this might very well also be nonlinear (this has not been tested).

Weapon skill does not lower defense skill, what it does is that it makes the difference (boss defense - player weapon skill) become smaller. This difference is what causes the extra misses and lower crits vs >70 mobs (and might give the mobs a higher dodge/parry/block rate as well, but not sure about that).

I suggest that you try to scan through the thread before asking questions, the current status in miss chance testing is summarized on page 15.

 

sp00n

Note that currently I'm running with 354 hit so at some point I should see a miss according to the wowwiki theory. I first like to see if the value per point is above 0.267%. If I see a miss, I'm going to proceed directly to 356 hit rating, with which I shouldn't see a miss for wowwiki, but a miss with the 0.2% theory.

 

ZiggyX

Unbelievable, if this is true I am going to make a post on the WoW forums telling blizz to wake the fuck up. You can't put a stat on a piece of gear and not let us know exactly what it does....that's just ridiculous.

 

sp00n

Well that's what the whole thread is about. Finding out the values weapon skill provides. Blizzard has never been known to provide much information (despite Crezax' fabuluos posting), something that irks me for quite some time now.

Basically you're right, they're tossing e.g. a belt with 25 weapon skill rating to us, and never let us know what exactly it does. I just don't think they will provide us with the information we need.

 

sp00n

Olgas, maybe you're lucky enough to find a [Marksman Bands] in the AH, or if you haven't done the quest already, get the [Expedition Repeater].
4 skill rating is 1.015 skill, enough to test at least.

 

Olgas

So, I raided tonight with exactly 141 hit rating equipped (8.94%). It was a short raid on Gruul's, which we cleared pretty quickly. And I died early on the High King fight so my sample size was not all that large -- only 474 shots.

But I got lucky -- one of those shots was a MISS ! And it only takes 1 miss to prove the point.

I had done almost 3000 shots on bosses without a miss with 9.00% (142) equipped.
But, then, I did experience a miss on a boss with 8.94% (141) equipped.

As far as I am concerned, I feel this test is done. This is as close to proving that the exact base miss rate on a Level 73 is 9.00%/28.00% as I think we can come.

I mean I guess you could argue that the base miss rate could be between 8.94% and 9.00%, but I'm not sure what the point of that argument would be since you can't equip gear to get you between those numbers anyway.

Or I guess you could argue that my 3000 shot sample size at 9.00% was simply not large enough. However, when you combine this test with the results of Karmon's test in post #318 (3140 attacks, 28.02% hit equipped, zero misses), I'd say we have now collectively aggregated enough data to definitively conclude that the base miss rate on a Level 73 boss is 9.00% (28.00% if dual wielding).

Now, we just have to figure out the effect of Weapon Skill difference >10 and we'll have "solved" the whole puzzle.

(BTW, I'm leaving for a one week vacation starting tomorrow morning -- so I'll be offline for a bit. Can't wait to see what our tests show when I return!)

 

sp00n

Heh, I was saying the same, turned out you guys have been lazy then. :P

 

Gruntle

I agree, I think we have more or less nailed the 9% base miss chance. Using your 1 miss in 474 shots we get an observed miss rate (correcting for 141 hit rating): 9.15. The confidence interval (calculated by using Exact Confidence Interval for a Proportion) of this value is [8.99,9.24].

You were really lucky to get a miss, the very best strategy to minimize the intervals is actually have a very big sample *but* also to setup the test to get few misses (the intervals will get smaller if you have fewer statistical "successes").

Using the same calculator for 0 misses in 2909 shots (correcting for 142 hit rating), I get an upper 95% confidence limit of 9.13 (not sure how it works to calculate a stdev for 0 successes but this calculator can do it).

Conclusion is that we can be 95% certain that the base miss chance lies somewhere between 8.99 and 9.13 (combining both tests like this might not be statistically correct, but hey, were not doing rocket science here).

Great work with all the log parsing and fiddling around with +hit! Thanks a lot for doing it.

 

rooppa

1st time posting here but i have read through this post with alot of curiosity.

now that i belive it has been established with 95% certaintly that a level 73 boss has a miss rate of 28%..

would it be possible for a test to be conducted were the person does not have ANY + hit what so ever or any WE or wepons skill points..

this should produce a miss rate on the boss (depending on the character and talent build) with a default miss rate extreamly close to 28%.

but this would also give a rate for dodge and parry depending on if the target was attacked from the front or rear.

once this "flat" rate has been establised.. then altering the talent build to incorperate wepons skill talent points (possibly a mace specced rogue increaseing there talanet points 1/5 each test) would then indicate a alteration on miss/dodge on each of the points that the rogue gained.

in this sense the amount of dodge which "may" be pushed from the table may be able to be worked out for each point of wepons skill gained. thus giving a more definatly value to WE and wepons skill points.

sorry if this is not making sense..and i know that this may be difficult to test as getting rogue to a) loose so much in terms of DPS for the cause of testing may make there guild angry b) alter there talent build over a course of several raid in order to test 1-5 points in the mace spec talent to see the results on each of run and the dodge pushed out of the table (if any)

/edit

thought of a possible test situation for this..

good old snake boss in ZG..

warrior tanking ZG boss but with "low" level items thus not hitting "that" hard and extending the fight.. with no hit rating nor wepons skill points the flat rate's should be indicated.. then simply repeating the test by resetting the boss once he hit's phase 2 with a increasing amount of wepons skill points.. the difference in alterations to parry/dodge should then be visable and seeing if any dodge is pushed from the table due to the wepons skill points.

possible viable wepons in the test would be the default "white" wepons that are purchased from pretty much every vendor in a capitol city.. as they are "white" wepons then more and 1 is able to be packed into bags and thus no time breaks for broken wepons..

obviously this could be tested with a orc/human and altering the wepons involved to incorperate there racial ability wepons. this "should" therefore indicate the benefit of 5 points of wepons skill rather clearly and any effect on dodge/parry

 

• Aldriana

So, the problem with that approach is that to accurately measure a miss rate on the order of 28% requires a truly monumental number of swings to get the variance under control. To distinguish between a base miss rate of 28% and 27.5% requires tens of thousands of swings, and to actually get it to sufficient accuracy for what we're trying to figure out could easily take over 100k swings, which is just not practical.

What works a whole heck of a lot better is to figure out "how much +hit gear do I need to no longer miss", which can generally be determined by a few thousand swings at each of a couple different speeds, which works out to be much less work on the whole.

What you really want to do is for each of a number of different weapon skills, start with some amount of hit below the expected cap, attack until you miss, and then slowly increase hit until you stop missing, with the goal to be finding an amount of +hit N such that with hit rating N you miss at least once, but with hit rating N+1 you go at least 5k hits without a miss, thus allowing the base miss rate to be reverse-engineered.

This, of course, takes a certain amount of time as well, particularly due to the need to test against higher-level mobs, which makes it harder to collect data than it otherwise might be.

 

Petersen

I would assume the same applies to a "weapon swing experiment" to determine the effect of weapon skill on the enemy's chance to dodge and chance to parry. That is to say it would require an exorbitant number of swings to be able to say with reasonable confidence what the "baseline" dodge and parry chances are, then again to say how +5 weapon skill affects them, and so on increasing weapon skill.

That being said by what method may we determine how a boss' chance to parry and dodge are reduced from having more weapon skill?

 

• Aldriana

The problem is less severe when measuring stuff around 5% than it is around 25%, but yes, it's still an issue. However, there are a couple easy workarounds.

The challenge with hit is that in order to get a sensible data set, you need to have your hit rating and weapon skill even for the entire test.

Dodge, on the other hand, is only affected by weapon skill, so as long as weapon skill remains constant all your other stats can change.

Additionally, while a number of bosses have abilities that increase your miss rate, I'm not aware of any abilities that affect their dodge rate.

What this means is, unlike hit chance where very controlled tests are needed, it should be possible to just grab WWS from a bunch of different rogues killing a bunch of different bosses, and simply index it by weapon skill. Hence is should be possible to accumulate data a lot more rapidly than for hit. The one caution is to make sure to only take data from fights where all opponents are the same level, but there are certainly plenty of boss fights where that is true.

So, the way I'd approach this is: have people log combats were all opponents are of the same level (so, for instance: Hydross or Tidewalker will work assuming you never attack any adds. Bosses like Void Reaver and Gruul where there's no adds are even better. Magtheridon is fine, since his adds are boss level as well. And so on). Whenever you record such a combat, post in this thread total number of attacks, number of dodges, and your weapon skill. Since we can combine data sets across multiple testers, we should be able to get up into the tens of thousands of data points with relative ease, combined with the slightly less outrageous variance should allow us to get a reasonable set of data without undue effort.

Parry is a little harder since bosses only parry from the front; data for this can really only be taken by main tanks, since everyone else will generally be hitting the sides. But, if people want to test that as well, that's certainly an option.

 

Kytrarewn

This might prove useful:

Gruul's with Whispering Blade of Slaying MH/Merc Glad's dagger OH and 363 skill:

2% dodge-rate, out of about 400 total melee attacks.

Gruul's with Edge of Cosmos MH/Merc glad's dagger OH and 360 skill:
4% dodge-rate out of about 400 total melee attacks.

ΜΟΛΩΝ ΛΑΒΕ


Jesus don't want me in a sunbeam
Sunbeams are always made on me
Don't expect me to cry, for all the reasons I'm gonna die
Don't ever ask your kick of me.

 

• Aldriana

Actually, it occurred to me that the real way to test this would be to just have someone write a mod that records what level mob you're attacking, total attacks, and how many of those attacks were dodged. Just have it accumulate data and sort by the mob's level (i.e., you'd get a display to the effect of "against level 73 mobs, you attacked X times and Y were dodged. Against level 72 mobs, etc.")

Then you'd just need a couple people to run this in the background for a week or two without changing their weapon skill, and you would soon have all the data you could ever want.

 

koaschten

You could even read out hit and weapon skill from the char screen to get more accurate data. After all you could have more variety in hit and weaponskill and wouldnt need to make sure you dont accidently switch a piece of gear due to an upgrade and all data is "poluted" and yes, i would run that mod if someone creates it.

 

sp00n

I was just looking at some of the data posted here in terms of crit and dodge and so also on the original postings by XP-Dolphin with the tests on the PTR.

Let me quote it again:

Now I was confused by the very low crit rate they have in this fight, until I did some math:
For Beaglej with 5.25% crit:
3933 hits + 196 crits = 4129 successful 'normal' attacks
4129/8682 total attacks = 0.476% rate of normal attacks
(4129/8682) * 5.25 = 2.49% crit rate or 216.7725 crits

Not very far from 2.258% and 169 crits.

And for Therasiv with 5.97% crit:
4845 hits + 327 crits = 5172 successful 'normal' attacks
4845/9975 total attacks = 0.486% rate of normal attacks
(4845/9975) * 5.97 = 3.095% crit rate or 308.7684 crits

Not very far away either.


Now that we take into account that Therasiv had 15 weapon skill more than Beaglej and that Blizzard stated that weapon skill would increase the crit (rating?) against mobs higher in level, it seems reasonable that Therasiv would have had a higher crit rate than 5.97% against a level 73 mob.

Let's take a look, for Beaglej (350 skill) it would have been
(5.25-0.5)*(4129/8682)/100*8682= 196.1275
So 0.5% less crit rate than on his character screen to meet the results in his test.

And for Therasiv (365 skill)
(5.97+0.36)*(4845/9975)/100*9975= 327.3876
0.36% more crit to match the results.



Unfortunately, the kind of math I just did contradicts the current theory for white attacks. It requires a two roll mechanic to be at work for white hits as well!
We know that yellow / special attacks are on this two roll mechanic (so that an attack first has to be a hit to be able to crit), but for white attacks we assumed it to be a table based one roll mechanic.

The thing is, I just cannot explain otherwise why Therasiv would have had only a 3.287% crit rate, although he was at the same weapon skill number like the boss had in defense skill. He should have been way more closer to 5%, but a 1.713% difference in almost 10,000 attacks seems very, very unlikely to me.


So we are facing the following situation:
a) We have to rethink our formula for white attacks as well. We would need a controlled test with a stable crit rate for this (so no raid and no procs).
b) Blizzard was testing on PTR another weapon skill formula, and has since changed it again, thus (unfortunately) making Beaglej's and Therasiv's test outdated now.


At the moment, I lean towards the latter explanation. Beaglej, with 350 skill, should have been at a 9% miss chance (pity that the character screen screenshots aren't available anymore to double check!), but is reported with only 7.971%, also more than 1% difference.
My guess is, Blizzard really used the PTR to actually test things, and the test fell within a time span where they were testing a different weapon skill formula / combat mechanics overall.
With almost 10,000 attacks, you shouldn't see so many differences.

 

Gruntle

Beaglej is a Dranei Paladin, he has +1% hit aura (and I think Therasiv was under this influence as well). We already discussed this a few pages ago.

The low crit is a bit trickier though, I agree that Therasiv should have been at a 5% crit rate with 365 skill (the upper 95% c.l. is 3.64%). Can't really think of anything that would cause it to be as low as this... Crit could possibly have some added level dependence (i.e. not related to weapon skill - defense)?

 

• Aldriana

One thing to keep in mind is that, as a level 73 mob, Venoxis would have 365 defense, which presumably would lower the crit chance of people attacking him by some amount. Now, I might not have guessed that "some amount" was roughly 1% per level, but, on the other hand, I can't really rule out the possibility either, particularly since the difference between their crit rates is fairly close to what is predicted in theory.

 

saedo

Well I dunno if this helps anyone but here's some "data" I got. Since last week I've been running with 304 hit rating and 362 weapon skill. Didn't see a miss til 2 nights ago on Leotheras. Here's the WWS of when that occurred.

Saedo - WWS

Not a large sample yet for myself, but it's the first and only miss thus far.

 

Punscho

What would be the most extra weapon skill you could achieve? I believe dagger skills is the best candidate. Some examples:

10 skill: Rogue talents.
Items and ratings:
[Circlet of Restless Dreams]: 14
[Shoulderpads of the Stranger]: 10
[Gloves of the Searing Grip]: 18, [Grips of Deftness]: 15, [Aged Core Leather Gloves]: 12
[Belt of One-Hundred Deaths]: 25, [Mugger's Belt]: 12
[Shapeshifter's Signet]: 20
[Fang of Vashj]: 21
[Whispering Blade of Slaying] 15

At this level of bonus skill we should be able to notice a differance in avoidance and crit more clearly if it exists.

Even if we subtract some of the more exotic items one could build up a pretty solid extra amount of weapon skill making testing so much easier.

 

Danther

Edit: Log had inconsistencies, so I don't want to submit it.

 

• Aldriana

Unsurpising. According to current theory your miss rate should be .022% with that gear, so you should see 1 miss every 5000 attacks or so.

Thing to be cautious of is that there are nonlinearities in the value of weapon skill, so knowing what happens up around 380 weapon skill isn't actually very interesting since it doesn't necessarily tell you anything about weapon skill at 350-365, which is the far more interesting range as it's where most people who are raiding are.

 

Punscho

True, but at the moment we require a huge amount of swings to conclude there even exist a decreased avoidance rate and/or an increased crit rate. Sure the reasonable values of 355-365 are the most interesting from a practical standpoint but an even higher value would be a good starting point for further research. If we see something interesting we could lower the values and try to deduct if it's linear or not.