Rallik

For the sake of completeness, I did some short testing today to lock down the boss armor value precisely at 10643.

Test method:
1) Decide if armor is less than or equal to 10643 or greater than or equal to 10644. This was done by finding a borderline level of AP such that if armor was 10643, the possible damage values for BT shifts down by 1 compared to 10644. The AP value I used for this test was 3214, giving an armorless BT damage of 1607.

1607*15232.5/(15232.5+10643) = 946.016, crit = 1892.031
1607*15232.5/(15232.5+10644) = 945.979, crit = 1891.958

So, if there is either a 947 hit or a 1893 crit, the armor value must be less than or equal to 10643. If there is a 945 hit or a 1891 crit, the armor value must be greater than or equal to 10644.

After a few dozen BT's landing for 946 and 1892, I saw a 947 hit, meaning the armor must be less than or equal to 10643.

2) Decide if armor is less than or equal to 10642 or greater than or equal to 10643 by the same method as above. The AP value I used for this test was 3044, giving an armorless BT damage of 1522.

1522*15232.5/(15232.5+10642) = 896.012, crit = 1792.024
1522*15232.5/(15232.5+10643) = 895.977, crit = 1791.955

After a few dozen BTs, I saw a crit for 1791, which means the armor value must be greater than or equal to 10643.

Since the armor value must be less than or equal to 10643 and must be greater than or equal to 10643, it can only possibly be 10643.

 

Mavanas

I have a hypothesis that the odd term B/(A+B)*xy in Rallic's formula could be nothin else but 1/2*0.81*xy, where 0.81x is the already reported scaling of arpen buffs (battle stance/gear/mace spec).

So the full formula becomes:
,


where

I have tested this formula on Hellord's data from the previous page and was able to match Effective BT Dmg in every single case. I requested data from Rallic as well so hopefully we can test it on his dataset as too.

 

Rallik

All you actually did was change the B/(A+B) term which is ~.41 to .5. The "x" I used is simply the sum of your character's arpen buffs(which you multiply by .81 if you're including the penalty). If you just call x the sum of arpen buffs without the penalty, and label z as .81x, then the formula I used would be the exact same as the change you suggested except with B/(A+B) * xy instead of .5 * xy. They certainly aren't the same, and it's easy enough to point out a case where it won't be accurate. Downloading Hellord's sheet, the bstance/sunder/ff sample has 67.5745% damage. Using .41 as the xy term modifier, you get an expected 67.567% damage, which predicts 1451-1452 damage, exactly as obserbed. Using .5 as the xy term modifier, you get an expected 67.512% damage, which predicts 1450-1451 damage, which is off.

 

Mavanas

No I am replacing B/(A+B) which is approximately 0.41, with 0.81*(1/2)=0.405.

I was looking at Hellord's table in post 363 on the previous page and was able to match every single number there using this modification of your formula and it's not very surprising because 0.405 is not that far off from 0.41. The main difference is that

,

is cleaner and reduced armor formula does not contain B/(B+A) term, which is not intuitive. The A term is a paramater of damage reduction and it did not seem logical to me to include it in the calculation of the reduced armor.

 

Rallik

You're misreading the original formula and substituting incorrectly. If you want to include the 81% penalty to x, you replace x with .81x, giving:



If you replace .81x with "z" for whatever reason, that still just makes it:



.81 * .41 ~= .3321 if you multiply it out. Using .5, it's obviously .5 * .81, which is .405 and entirely different. 1/2 is a nice round number and more "intuitive" but it just doesn't work, as in the example I gave above. Hellord's table in the post had only a single example where this change affects the formula at all(since only battle+sunder would have a non-zero xy term), and the only reason it checks out is because the values of x and y are too low and the precision of his test(rounding to the nearest integer) isn't high enough to shift the damage range by 1. If you download the spreadsheet a few posts below, you'll find it doesn't match, as in the example I listed from it. If you don't like the B/(A+B) term, just call it .41, it doesn't really matter, and might not be based on B or A at all, but it certainly isn't just 1/2.

 

Mavanas

You are absolutely right, both .3321 and .405 produced same results in the very first table of Hellord. However, in further tests, 0.81*(1/2) produced inaccurate results. While at it, I checked that 0.81*B/(A+B) matched his data in all but 4 cases and the difference there was no more than 0.6 damage or less than 0.05%.

You should still try to carry out a test where either A or B value is different (lvl 60 dummy or a warrior without a shield that is getting healed) to make sure that the B/(B+A) factor is intentional.

 

Rallik

Yeah, it's most likely not based on B/(A+B), was just an assumption when it came out to be nearly identical to it in the tests I saw and I was trying to find some explanation for the strange stacking effect. From further testing, I'm actually positive that it isn't just constant now, and gets quite a bit higher when y gets higher. With 5xSunder/Shattering Throw, it goes up to .5-.6. It's hard to pin it down exactly when I can only get off 2 BTs every 5 mins, and the constant can shift by as much as .1 or more with a difference of only a couple armor in the calculation, so it requires a large sample size.

~.41 is an accurate estimate for now for Sunder+FF, which is the most common debuffs up the majority of time, but I'd like to figure out why it changes when y changes or if there's some other interaction between buffs and multiple debuffs that can explain the change when 5xSunder/ST brings y up to .36.

 

dysent

Patch 3.1 - PTR Build 9767 dropped today with very few listed changes and apparently was bugfix focused.

Anyone been able to log on and check to see if the -19% ArP bug has been fixed? I'll do it once I'm home from work if noone confirms/denies before then.


[edit:
1744 BT base, 186 ArP, 1081 zerk BT, 1120 battle

expected zerk BT= 1095, expected battle BT = 1145
expected zerk BT with 19% penalty = 1081, expected battle BT with 19% penalty = 1120.

Numbers are STILL bugged.

Suggest someone post about it again and ask them if they actually intend to buff ArP or not. I'm still banned, so I can't :\ ]

 

• landsoul

Is anyone able to pin down glancing blow reduction ranges on bosses I'd love to get my hands on that. I have a 45% to 25% reduction range on glancing blows currently and I'm questioning it. All you'd need to do is get some small range daggers and high AP and record a large set of glancing blow sizes. Prefer to do this on PTR.

Visit my Youtube Channel for an increasing selection of warrior videos, including Undermanning, PvP, and LK Raids:
http://www.youtube.com/user/LandsoulWoW

 

drumbum

Around a year ago it was determined by some of the rogue theorycrafters that glancing blows average at a 25% damage reduction, not 35%. However it wasn't clear to me in your post whether you are asking what the currently accepted value is (I believe it to still be 25%) or if you were asking whether there is evidence of it changing in 3.1.

I tried to find a post that detailed the testing but I couldn't pin it down.

 

• Aldriana

I believe the rogue testing indicated that it was 20-30% reduction, for 25% average. However, I think that testing was in 2.3 or 2.4, so it's probably worth rechecking.

 

Rallik

While I was doing crit depression testing with a few thousand hits, I saw glances as ~15-35%, with an average of ~25% reduction: Retesting hit table assumptions

 

Althor

I incorporated this new Armor damage reduction formula into SimulationCraft tonight but I'm left the question as to how do Serrated Blades figure in?

It appears to still be a static armor reduction amount rather than a percentage or a rating.

 

Rallik

ArPen bug of 81% of tooltip value still exists on live.

 

dysent

World of Warcraft - English (NA) Forums -> ArP still bugged. Battle stance gives 8.1%...

Please contribute your own figures or math.

Let's at least see if we can't get a response. It's reasonable to request one.

 

Lollersk8er

I don't think it's a bug. ArP would be too strong at 100% effect, so it's most likely a simple nerf.

 

RPZip

So the patch note and started purpose of buffing ArPen rating by 25% was some kind of clever ruse? Not to mention the bugged tooltip values.

The thread on the official forums is much more likely to get a response on the Damage Dealing forums, which has the advantage of being less spammed and actually read by someone other than a Moderator, i.e. Ghostcrawler.

 

dysent

Agree. The Damage Dealing forums were locked at the time and I was upset/frustrated and got impatient. General forum was a bad call on my part.

Would appreciate if, after we verify that the emergency maintenance today did not fix the problem, someone posts on Damage Dealing forums and links to it here so we can go there and bump it until a response or change comes. Reposting same post in multiple places risks another ban, or I'd do it.

The reason we're calling it a bug is that the observed value and the tooltip value disagree, and the tooltip value represents the buff in the patch notes, while the observed values are almost the same as 3.09 (or less, in case of static/talent buffs).

Blizzard said they wanted to do one thing (buff ArP), then did something else (sidegraded/nerfed ArP). We just want to know which one is intended and which one is a bug.

 

hellord

116% Armor Pen only giving 66%!?!?!?! | DPS | WoW Blue Tracker | World of Raids

GC Announce they'll reveal ArP formula.

What cap should be that?
If I have 30% ArP the cap is 2461.725 of .. what?
Does this mean your relative arp cap dependends on the arp you have?

From the sound of this reply I guess arp is "working as intended".

ArP Whore

 

RPZip

Working through the math now, but the reply is interesting.

 

• Aldriana

So, upon preliminary testing, this makes sense - (10643+15232.5)/3 is .8104 times 10643, which is where our factor of .81 in the above formulas comes from. The rest of the formula wierdness seems to come from the fact that Sunder/Expose/Faerie Fire/etc. seem to also lower the cap, as well as reducing the armor value.

So, for instance, if I drop a 5-point Expose Armor on the boss-level target dummy (which has 10643 armor, base), it's armor is reduced to .8 * 10643 = 8514.4 armor, and the cap is reduced to (8514.4+15232.5)/3 = 7915.633333 armor. So, when I attack it with 171 armor penetration rating on (=13.8841% reduction), the armor value is reduced by 7915.633333* .138841 = 1099.015 armor to 7415.385; thus, a gouge with 3521 AP (base damage: 814.451) should do 740.41 * 15232.5 / (15232.5+7415.385) = 547.783 damage, which matches the observed value of 547-548 nicely.

Or at least, it almost does. The problem is that a crit with Lethality, Prey on the Weak, and Relentless Earthsiege Diamond would be expected to do (2 * 1.03 * 1.2 - 1) * 1.3 + 1 = 2.9136 times as much, which would work out to 1596.02 damage, but the observed value is 1595-1596. So there's apparently some internal rounding in one or more steps such that this formula isn't exactly right (or the boss armor value is actually 10643.5, or some such).

Regardless: as a first cut at the formula, modulo minor rounding issues, seems to be something like



for a typical boss mob. Note that the ((1-y)B+A)/3 term is really the minimum of that and B(1-y), but for a typical boss mob it will usually be the term as written.

Edit: The rogue Serrated Blades talent appears to apply after Sunder/Faerie Fire and similarly reduces the cap.

 

nightcrowler

Aldriana should be something wrong in your formula because as GC said:

For a level 83, C=16635. And you are using the level 80 value, so probably we should change Armor value to compensate it.

(Or GC made a typo?).

 

Rallik

GC is just wrong about that. The half armor value uses the player's level, not the target's level.

 

Mavanas

Since I had a spreadsheet handy from previous testing, I plugged in Aldriana's formula in Hellord's data (4 datasets). It was mostly exact matches. Maximum discrepancy was 0.05% which resulted in 1 damage difference between expected BT damage and observed BT damage in 6 out of 32 cases. The discrepancy could be due to rounding of armor, or because reported BT damage was not a whole number as indicated in the dataset.

Edit: What bothers me is that in testing the formula they suggest not having mace spec and stance in addition to armorpen as they can complicate things. If mace spec and battle stance are in fact additive with armor pen from gear, there should be no complications testing the formula.

Also GC says "Another way of understanding that is we multiply the percentage in the tooltip times the minimum of the two values: the cap, and the amount of armor on the target after all other modifiers," which makes me wonder if mace spec and battle stance are also part of these modifiers.

 

• Aldriana

More to the point, the formula I have written gives the right answers, and what GC said doesn't.