On a sidenote, which addon did you use for the SoS-testing, Zoey?

The Screenshot I posted is from "Recap (Hawksy) 3.86", i used SWStats and a /combatlog to verify the
numbers, and recap picked it all up.

Additionally, i ran another test today, this time i went for Level 45 Mobs in Feralas:
Hulking Feral Scar Yeti and Enraged Feral Scar Yeti.

Again, my tests are based on the following assumptions:
- One-Roll attack table, as described here: http://www.wowwiki.com/Attack_table
- Mob Base miss chance is 5%
- Mob Base crit chance is 5%
- Defense reduces the chance to be hit and crit by 0.04% per point
- Defense increses the chance to dodge and parry by 0.04% per point
- Weapon skill directly counters Defense
- Mob Weapon skill is solely based on the Mob's level, using the formula: WeaponSkill = MobLevel * 5

Based on those assumptions we'd expect the following outcome when a Level 70 characer with maxed out
Base Defense Skill (350) is fighting a level 45 Mob:

Weaponskill/Defense difference: 350 - (45 * 5) = 125



Against a Level 70 Mob, my unbuffed values are:
The 125 points of difference from fighting Level 45 Mobs would result in the following expected values:

Now my Test setup, its pretty much the same as the one i've been using in SoS:
- 11 Mobs, 4 "Enraged Feral Scars" and 7 "Hulking Feral Scars"
- due to the "Enrage" Buff that the "Enraged Ferals Scars" gain at low health all of the "Enraged Feral
Scars" were kept on high health (>75%)
- Since i was casting heals and its impossible to parry while casting (does the same apply to dodge? I
assume it does, at least thats what my numbers would suggest, but a confirmation would be nice) the
values for Parry (and maybe Dodge) are irrelevant
- Buffs used: Blessing of Wisdom, Righteous Fury, Concentration Aura, +Spelldamage Trinkets

The test lasted roughly 1h 9m, you can download the full CombatLog here: http://files.filefront.com//;7736223;;/
During the time of the test 19.390 attacks were recorded:



Note that recap shows "Misses: 3.692", which in reality are "Hits avoided (as in Misses+Parries+Dodges).

As i've stated the lower parrry and dodge chances (and thus the increased number of normal hits) are
likley a result of casting heals on myself.

What still remains is the recorded critchance of 1.14%, which is not only significantly higher than
the expected 0% but also significantly higher than the 0.3% i got out of my Murloc Test (see above,
lvl 42 Mobs).

So where are all the reports of tanks getting critted by level 73 Mobs? There are some, but not nearly
as many as my result would suggest. My theory is that it is because of Shieldblock. A tank who's up
against level 73 mobs will try to keep ShieldBlock up for as much as possible, so after avoidance
(which, following the one-roll Attack table theory would lower the crit rates in the first place) he
might up with 75% of the incoming attacks being blocked (and thus unable to crit).

The reason of why people report higher critchances from dualwielding mobs is their attackspeed, they
simply use up the Shieldblock charges faster than normal mobs, thus increasing the chance to actually
get a critical hit coming through significantly.

In conclusion, there is something fishy about the Defense = crit immunity thing, at least the formulas
that are wide spread and currently used to calculate the odds of the Attacktable are inaccurate.

(PS: i'm far from an experienced Theorycrafter, so there is still the possibility that my assumption
or calculations are flawed, i tried my best to doublecheck everything and provide you the data samples
to verify/invalidate my results)

Small follow-up:

Would the decresed Dodge+Parry chance i've been suffering from due to casting also increase
the Mob's chance to land a crit? Or would those just be converted into normal hits?

My chance to Parry or Dodge is 49.1% lower than expected ... what could cause this besides
the casting issue? The majority of the Mobs were in front of me 100% of the time (assuming
parry works in an 180° angle in front of the character.

I am assuming you were getting hit while casting as well and during that time all your avoidance is set to zero (only misses can happen). Do you think that would skew the test results? Crit may not happening on a player that is just standing there and not casting.

Good question, it certainly skewed the results, lowering Parry and Dodge. I'm not sure about
the Crit outcome tho.

They did. After checking my Combatlog there were crits while i wasnt casting anything.

I hardly keep up Shield Block all the time on raid trash or in heroics (depends on how hard it hits, importance of threat, lazyness etc), and yet I have never been crit except by aforementioned mobs.

I have never been crit while tanking.

However, I have 490 defense, and I'm not spamming holy light on myself while tanking level 45 mobs.

I'm not blaming you, but -- yikes, the text you quoted is painful just to read for any computer scientist.

Briefly, there are plenty of opportunities for rounding errors in floating point arithmetic (starting with the classic example of floating point addition not being associative), but if you want to speculate about rounding errors in the game engine's implementation, it's probably best to ignore the above paragraphs in its entirety, starting with outright stupidity such as "The zero is never perfectly zero". It does not describe any model of computer arithmethic I know of.

It's not wrong as much as it's just complete gibberish. It's like the author heard that floats are imprecise and then banged his head on the keyboard.

It's also painful for any systems engineer who does aerospace work. I concur in the "bullshit" verdict.