If you're parsing directly from the combat logs, you're fine following this as a guide. If you're going through the pages of WWS logs be careful. The individual character pages can either show you all the attacks made from a target or all the attacks made to a target, but not both. Therefore when going through them scroll down to check if all those attacks were against the same target. If not, it's possible the person was behind the target, throwing the parry data off, or was hitting a non-73 mob, throwing all the data off. You'll either have to go manually through the WWS combat log page and remove the bad entries or skip the fight data for that person altogether. Furthermore, there has been discussion (I'm not sure if it has been tested and proven) over whether mobs can dodge or parry while casting or channeling. There is no way to tell through WWS alone which or how many attacks were made during that time.

With that in mind, here's what I found for each boss: [EDIT: Reminder - this only applies if you're pulling data from the WWS stat pages. If you plan to use log files, ignore this list.]

- Attunmen: MT data clean, OT data not
- Moroes: only clean data is the MTs
- Maiden, Curator, and Prince should all be clean
- Aran: No one's in tank gear, you should always be behind him if you're melee, and he's always casting. Don't even bother.
- Netherspite: there was some talk over if he has a different hit table than other bosses
- Nightbane: clear out attacks against the skeletons and you're good

- Gruul: Same as Sp00n
- Magtheridon: The channelers and infernals are not level 73, and every tank has one, plus what Sp00n said. Not good.

- Hydross: Same as Sp00n, plus add troubles. The two main tanks will be fine dodgewise.
- Lurker: Spout and adds. Unless your MT doesn't touch adds all the data will be iffy.
- Morogrim: "could be the bast fight" for the MT only
- Karathress: Same as Sp00n
- Leotheras: Same as Sp00n, plus the adds in the beginning
- Lady Vashj: Adds, adds, and more adds! I avoided her altogether.

- Voidreaver: Same as Sp00n, plus the occasional slamming the ground channel, but the guaranteed lack of adds makes this the easiest fight to get data from.
- Al'ar: Same as Sp00n, but only for the three tanks that stay on Al'ar the whole time.
- Solarian: more adds == more pain
- Kael'Thas: Every tank has an add at some point. Probably just as bad as Vashj for using WWS pages.

I am extremely interested in finding out what this information means for tanks. I alwalys look for "ideal gear sets" and such for tanking, and one thing that I absolutely dislike with a passion while tanking is avoidance streaks from my attacks. When I heard weapon skill reduces the mob's chance to dodge and possibly parry your attacks, I was really interested. I have the Mallet of the Tides and am a Human, so my weapon skill is 358. I also have some +hit on my gear, 48 rating, for 3.04%. My warrior (Xav) is just my alt, but I still want to help out however I can.

I could do old world content vs Boss Mobs to further testing, as I've already done some stuff several times and fairly frequently (Naxx clears, BWL, MC, world dragons). Since there's Boss mobs in there, I figure they'd work fine.

A boss I think would be extremely Ideal for this would be Golemagg, since you can have everyone stand in front of him so whenever he goes to use his pyro, he doesn't turn around or anything. Then it'd just be tank and spank for however long I want.

So, please tell me anything I should/can do to help out.
The Armory

That's my warrior's armory - if a different weapon or something is equipped, just assume Mallet of the Tides is what I'd be using.

I really want to know exactly the benefit of +weapon skill and if I was right to originally think Mallet, Gauntlets of Enforcement, Human bonus, and about 5% hit is the best possible setup you could hope for as a tank.

10 pages ago in this same post there was a discussion about what bosses were the best for "vs boss" testing. Golemagg is a no-no due to dog respawns and some other stuff I cannot remember of.
Two paladins and a druid did a looooooooong non-conclussive(on dodge/parry) test on venoxis, so maybe you can start there.

@Leguaris
Concerning adds, if you are using the combatlog itself, you can filter it only for the actual boss' name, so these aren't a problem.

And Golemagg should be fine if you can keep the dogs busy AND use the combatlog.

I've uploaded a simple script calculating your required hit rating based on your weapon skill:
Weapon Skill

I would just have another person offtank the dogs. And we'd kill magmadar first so that core hounds don't respawn, either.

So.. is there anything I could help with?

What would be considered a good sample size? I have the brutalizer and the gauntlets of enforcement and I would be more than happy to help out.

Would 900-1000 be sufficient? Or are we looking for something like 2-3k+?

Last night on archimonde we had a few wipes, so I was able to get a screenshot of my melee attack window from recount.

355 weapon skill (rising tide/merc gladiator oh)
16.76% hit (13.76% hit + 3% precision)

With those two stats I had a 5.9% miss rate on archimonde with around 900 auto-attack swings.

IMAGE DUMP

Small bug: decimals work. The script should round the weapon skill value to the lowest integer, shouldn't it?

Good point, corrected.


@GrungMag
To determine such rates, we need huge sample sizes. 1000 is not nearly enough, I think 5000 or even 10000 would be required.

Assuming I'm doing statistics properly for an event with a 5.00% chance of occurring per swing, to get a 95% confidence interval of +- 0.1%, you'd need almost 20,000 swings. To get it down to 0.04%, you'd need over 1,000,000 swings. That's why we'll probably have to accumulate data from multiple tanks.

Yeah you need a huge sample to nail it down to the degree we all want, that's why completing the "nice bosses to collect data from" list is important. Then people can supply us with logs.

I started a list waaay back in this thread but it's only up til ssc so hyjal/bt people, think about it next raid.

Been lurking and watching this thread for a while, not sure if this would help but there's a level 73 elite demon in Nagrand (Reth'hedron the Subduer - NPCs - World of Warcraft) that may be useful for testing. He has some sort of rain of fire ability, not sure if that'd mess things up or not.

That's the thing about the WWS website though: you can't just download the combat log behind the webstats. The closest thing you could possibly do is to go though the log pages (such as this one) and copy-paste the words into a .txt file, which would take way more time than going though the WWS stat pages like I was talking about (example). Yes, there is a filtration system, but you can only apply one filter. You need two to limit it to both a boss and a player. Even then you still lack a way to directly download the filtered logfile.

If Nahked, or whomever, wants to do that, or for that matter has access to the source combat files, all the more power to them, you can go ahead and ignore what I said. If not, that is where my post comes into play.

Ok, I copied 1500 lines from a Voidreaver combatlog, all between "Void Reaver gains Pounding" and "Pounding fades from Void Reaver". Not a single dodge in between these lines, so at least the channeled Pounding spell prevents him from dodging.
Also no single occurrence of parry ("parries") there.

Over 1379 attacks my dodge rate was 5.66%. 2 tries, every minute a Cloak of Shadows during Pounding and sometimes running out too late or coming back too early while he was still channeling.

This might complicate things a bit. I suppose the low dodge rate against Venoxis was also due to the casting preventing dodges (and parries, so the "true" parry rate would hover around 12-14%).


// Edit
Leotheras is also a bad fight, since he does not change his name during his phases. Only on phase 3, where he splits up, the demon form has a different name.
2.85% dodge there in 386 attacks. Might be luck, or again the casting of Chaos Blast. Unfortunately I cannot see whether he was casting or not while a dodge or parry occurs, since there is no "gain" or "fade" message.

Rholr - WWS

That is a combatlog of mostly Gorefiend wipes, and the tank has 358 weapon skill (Mallet of the Tides).


The raw combatlog is 20 megs (has trash clearing in it).

A real combatlog is much preferrable to a wws compilation.

If you are running windows (Linux/Os X guys already got it) you could download a GREP tool (like Grep for Windows) and filter the logs on just the player you're interested in. In that case we need to know there's nothing funny about the boss such as casting/channeling spells or something that will affect the avoidance rate since we won't be able to distinguish when that event happen. But to be honest, getting data from bosses you need to remove parts of the combat log from is not very effective to begin with. If you just extract the data for your tank the log gets much smaller. And when posting data please remember to state the actual number of avoided swings and not just the percentage.

Specials are on a two roll system (supposedly, still haven't heard of another theory to also explain the issues), meaning that only landed attacks can crit.
Therefore, if you are standing in front of the mobs, you're facing miss dodge and parry, which can easily accumulate to 25% without any weapon skill and +hit. And 12%*(1-0.25) = only 9% crit rate for your special attacks.
Also, looking at 18 normal melee attacks, you can hardly expect a consistent rate with that number.


Now, Patchwerk would have been a perfect sample, but it is possible that he has special mechanics. As far as I can remember, Patchwerk cannot do crushing blows, and it might be that he cannot parry/dodge while performing hateful strike (which is, basically, always).
Taking a closer look at your WWS, I think that there isn't anything different on them.
The sample size is a bit low, but you had e.g. 11% parry on Heroic Strike, 14% for Devestate and even 22% for normal melee attacks.
So I think the fight seems ok.

Patchwerk with 3 tanks (3 different weapon skill), 3 dps (also with different weaponskill) and healers to keep the tanks up might be interesting.

I'd first like to say fantastic job to everyone who's been working on this project! It is by far the most complete research on the mechanics of weapon skill that I've ever seen, and has been getting solid reviews from even Druids..

I was curious as to where the model is as to determining the mechanics of weapon skill in relation to parry/dodge chance, and to ask what would help with this particular research. It's something that I've been interested in for some time, and I do have a little bit of data (I don't know if it would be helpful, however).

From swing tests I performed on my Druid (auto-attacks with zero +hit or +feral combat skill), I do have a few numbers for level 71 and 72 mobs. From parsing just under 7000 swings on level 71 mobs, I confirmed the 5.50% miss chance (for 5 weapon skill difference), and found an identical chance to be parried (5.50%). Repeating this test on level 72 mobs, although I only had time to parse 2500 swings, I again confirmed the 6.00% miss chance (10 weapon skill difference), and it appeared that the chance to be parried was again 6.00%. Dodge sadly was all over the board.. quite clearly each mob type has a significantly different chance to dodge, and so that data was worthless.
I didn't get the chance to test these scenarios with weapon skill yet.. but would this test be useful?

Doesn't he have a hard enrage ? I recall some freaks kiting him to death several times so they added an ae poison bolt of doom or something.

I was thinking that for testing dodge, you could have 2 rogues in the same raid on the same boss at the same time and do a comparative difference to figure out the effects on dodge. This would take out all the complications of the dodge chance changes during casting. I'm not sure how accurate that would be, but it could be something to consider. 2 Rogues full time on Magtheridon could provide pretty good results if it works.

The variance on such comparative measures tends to be too large to get good results without having truly epic data sets. Also, the effect of spell casting would still matter. I suspect all you'd get out of this is "weapon skill reduces dodge chance by a decent amount", which isn't really that helpful.

Yep.

Might be time for another 5 man Loatheb, though.