Maybe there's some connection to the base dodge for the different classes, f.e. as seen on WoWWiki?
This formula seems so complex, they either must have based it on something already present or worked really hard to create it.

Class	Base (D)odge	AGI:D Ratio	D Rating:D Ratio
Druid	-1.87%	14.7059	18.9231
Hunter	-5.45%	25	18.9231
Mage	3.45%	25	18.9231
Paladin	0.65%	25	18.9231
Priest	3.18%	25	18.9231
Rogue	-0.59%	20	18.9231
Shaman	1.67%	25	18.9231
Warlock	2.03%	25	18.9231
Warrior	0.75%	30	18.9231

I did some testing on my level 80 premade DK, and here are the results:

88.12902089

46.10419548

0.99356050

Dodge Per Agi=0.01413434

Note: The parry rating from base strength does not suffer from DR.

What's the base dodge chance as a death knight at level 80? I tried to google it but haven't found
any useful information yet.

May i suggest that you in your example pick dodge and parry numbers that added up are greater then 3,9% since below that you actually get more then expected. It really confused me when the "diminished" number was greater then the starting number.

Do you know the Dodge per Agi ratio for lvl 80 Paladins?

No, as you can see, there aren't and lv80 Paladin data here.

The feral WotLK thread didn't seem too interested, so I'm posting this back here. Does the math seem correct to you?

The formulas I believe apply to feral druids at 80 (based on this thread) are as follows:

PreDR dodge: x = .024 * agil_from_gear + .025414 * dodge_rating + .0081326 * defenserating_from_gear
PostDR dodge: x' = (c*x) / (x + c*k) = 1/(.008555 + .972/x)
Actual dodge: base + talents + x' = 7.039 + 10 + x'

Net avoidance (boss):
A = (miss + scorpid + 0.0081326 * def_rating) + (17.039 + x') = (4.4 + 3 + .0081326 * def_rating) + (17.039 + 1/(.00855505 + .972/x))
~= 24.44 + .008133*defr + 1/(.008555 + .972/(0.024*agil + .02541*dodge_rating + .008133*defr))

In particular, let +defense and +agil = 0 and dodge rating = z. Then net avoidance A = 24.44 + 1/(.00855 + .972/(.02541z)),
and A = 24.44 + 1/(.008555 + 38.25/z).
dA/dx = 38.25 / (.008555*z + 38.25)^2

That last function goes from 0.0261 at z=0 to 0.01746 at z=1000, to 0.01248 at z=2000, and 0.007287 at z=4000.
That function represents the return on dodge rating at that level of avoidance (which is proportional to the return on agility). This derivative is nearly linear in the range from 0 to 4000 (visibly not so, but the curvature is light),

Net Mitigation Delta is a separate matter: It's calculated relative to avoidance, (1% dodge at 98% avoidance is worth more than 1% dodge at 25% avoidance). The A_factor is the amount of net mitigation adding one dodge rating to your gear would give you. For example, with no gear on (24.4% avoidance), adding one dodge rating reduces total damage taken over a long fight by 0.0346%.
The function is A' * 100 / (100 - A) .

Dodge rating	Net Avoidance	dA/dx	A_factor
0	24%	0.0261437908497	0.034600040828
200	29%	0.0239529354539	0.0339492518355
400	34%	0.0220263608907	0.033392890557
600	38%	0.0203232087395	0.032922903337
800	42%	0.0188102218843	0.0325327843664
1000	45%	0.0174601085875	0.0322173364858
1200	49%	0.0162503030586	0.0319724898597
1400	52%	0.0151620163779	0.0317951661714
1600	55%	0.0141795025616	0.0316831795826
1800	57%	0.013289485785	0.0316351683612
2000	60%	0.0124807095543	0.0316505531157
2200	62%	0.0117435790311	0.0317295192189
2400	65%	0.0110698751461	0.0318730224188
2600	67%	0.0104525244932	0.0320828179552
2800	69%	0.00988541290422	0.0323615148525
3000	71%	0.00936323347922	0.0327126585668
3200	73%	0.00888136198092	0.0331408469762
3400	74%	0.0084357541047	0.0336518870068
3600	76%	0.00802286034203	0.0342530022563
3800	78%	0.00763955507654	0.0349531061834
4000	79%	0.00728307725685	0.0357631613656
4200	81%	0.00695098053549	0.0366966538702
4400	82%	0.00664109118615	0.0377702243403
4600	83%	0.00635147244282	0.0390045162271
4800	84%	0.00608039416474	0.0404253304338

As you can see, returns go from diminishing to increasing somewhere around 60% avoidance, or 2000 dodge rating/2100 agility..

Edit: Updated to use the more recent numbers Whitetooth gave us below.

I did some updates for the druid values today, not sure how much this will affect your math.

PreDR dodge: x = .024 * agil_from_gear + .025414 * dodge_rating + .0081326 * defense_rating_from_gear
PostDR dodge: x' = (c*x) / (x + c*k) = 1/(.008555 + .972/x)

It'd move the inflection point somewhat, but I was ignoring agility and defense for the majority of it anyway. I'll go back and fix the post though.


The way they *should* be doing it is to calculate the net avoidance the way things are on live (including miss. Not sure if recalculating on the fly when scorpid is up would be an issue for them). Call that value x - then apply the formula A = 100 - 100 * exp(-x/100) to get the post-DR avoidance value from it. (Not sure of how to allocate what portion of the total comes from parry and what from dodge, but keeping the preDR ratio would probably work.

This has the effect of 1% worth of dodge rating ALWAYS reducing net damage intake by 1% - getting from 50% to 51% dodge takes twice as much as getting from 0% to 1%.

Does this imply we no longer convert defense rating to defense first? If not, has the correct rounding been accounted for?

Also, with a new agility to dodge conversion is there a new level based formula to get this value?

I certainly ignored rounding error and discretization error - most of these equations do, I believe.

Do we have any idea where avoidance chance adjustment from attacker level plays into this?

In TBC, if I've got 50% dodge on the character sheet, and fight a lvl73 mob, I actually end up dodging 49.4% of the attacks, due to the 3 level difference reducing my effective dodge by 0.6%.

In WotLK, does that 0.6% get taken off before or after DR is applied? I don't see any way we can accurately test this, however. Perhaps Blizz should add attacker dummies next to the target dummies, which just hit you for 1000 base damage per hit, rather fast, of both 80 and 83...

I'm amused by my mental image of thousands of lowbie corpses in a sprawling circle around the PunchBot2000.

It looks like it is 47.06 agi per dodge for 80 Paladins.

On beta, with 92 agi I have 5.03% dodge, and with 100 agi I have 5.20% dodge. According to the tooltip this is before diminishing returns, so with some simple math you get 47.06.

What is odd though, is that with 100 agi and no defense or dodge rating gear I have 5.20% dodge. Which means I would have 3.08% base dodge, which is significantly higher then at 70.

The "(Before diminishing returns)" in the dodge tooltip is a bit misleading, it only meant for the "Dodge rating of X adds X.XX% Dodge" part, but the total Dodge is the value after DR, this is the same for Defense and Parry.

If you would like to contribute some lv80 paladin data, this post may help.

I assume everyone understands DR as the more you have the less you get, but what people can't easily see from the formula is that you actually start out with more then what you have on live, the example was meant to show this property of the formula. That way people won't really get confused when they find that out themselves.

From this table, we got for level 80 paladins:
Dodge/Agi: 0.0192
Agi/1%Dodge: 52.083333
Base Agi: 3.2685

I copied 8 premades to 2 accounts, and was able to gather data for 8 classes at level 80.
The death knight dodge/agi values I had were strange and seemed wrong, so I did some more tests and found that the K value used in calculating parry and dodge for death knights were not the same.
Only DKs seem to behave this way, I double checked warriors/paladins and K values were the same for dodge and parry.

Here are the results:

							Lv80 Dodge/Agi	Lv80 Agi/1%Dodge	Base Dodge
Warrior	0.956	47.003525	0.021275	0.956	88.129021	0.011347	0.0136	73.5294	3.4636
Paladin	0.956	47.003525	0.021275	0.956	88.129021	0.011347	0.0192	52.0833	3.2685
Hunter	0.988	145.560408	0.006870	0.988	145.560408	0.006870	0.0133	75.1880	-5.4500
Rogue	0.988	145.560408	0.006870	0.988	145.560408	0.006870	0.0241	41.4938	-0.5900
Priest	0	0	0	0.983	150.375940	0.006650	0.0192	52.0833	3.1830
Deathknight	0.9935605	46.104195	0.021690	0.956	88.129021	0.011347	0.0136	73.5294	0.7580
Shaman	0	0	0	0.988	145.560408	0.006870	0.0192	52.0833	1.6750
Mage	0	0	0	0.983	150.375940	0.006650	0.0195	51.2821	3.4575
Warlock	0	0	0	0.983	150.375940	0.006650	0.0192	52.0833	2.0350
Druid	0	0	0	0.972	116.890707	0.008555	0.0240	41.6667	4.9510

Edit: I got a 3ed account, calculating hunter/rogue stats now.
Edit2: Added new Hunter and Rogue data.

Thats a good idea, I totally forgot about those vendor gems. I tried the AH but could only find Agi pieces. So you still need paladin data with defense/dodge/parry?

He wrote "8 classes done, 2 to go: Hunters and Rogues" so I think he is good to go with Paladins .


However, you only get 4 transfers per account, so that is why he is missing two classes.

I'm good with paladin data now, thanks, do you have any transfers left for hunter or rogue?

Edit: nvm, I borrowed a 3ed account, doing the last 2 classes now

First post updated with final(hopefully) results for all classes.