It deserves to be said, regardless of normal thread policy:

well done. Thank you.

The expertise tooltip on the character sheet shows expertise percentage being rounded down to the nearest .25%, when in reality testing shows that the tooltip is wrong and the percentage is based on exact expertise rating with no rounding.

I tested this with Recount on a training dummy using 212 expertise rating. The tooltip displays 212 expertise rating as 6.25%. I compiled a sample of 20000 swings and from those 20000 swings only 7 of them were dodged. That is a .035% chance to be dodged. If the 6.25% on the tooltip was accurate, the test should instead show approximately .25% chance to be dodged.


Link to Recount test data:
http://i39.tinypic.com/nbuam8.jpg

Well if you calculate the expertise rating you have got (212) into a percentage (divide by number given in opening post) you end up with 6,46538...% and if you substract it from 6,5% you end up with 0,0346% which is quite the value you showed.

I'm trying this for myself but you seem to have made quite a discovery. All those tricky math evaluating expertise is going to be wasted work.

My testing is not quite at the level as yours but it points in the same direction.

Not to put a damper on your analysis, but expertise not truncating was discovered and verified in another thread recently:

http://elitistjerks.com/1092985-post73.html
http://elitistjerks.com/1093112-post75.html

3.1.0.9614 BASE_REGEN values

Level	BASE_REGEN	Level	BASE_REGEN	Level	BASE_REGEN	Level	BASE_REGEN
1	0.020979	21	0.011840	41	0.008235	61	0.006421
2	0.020515	22	0.011494	42	0.008113	62	0.006314
3	0.020079	23	0.011292	43	0.008018	63	0.006175
4	0.019516	24	0.010990	44	0.007906	64	0.006072
5	0.018997	25	0.010761	45	0.007798	65	0.005981
6	0.018646	26	0.010546	46	0.007713	66	0.005885
7	0.018314	27	0.010321	47	0.007612	67	0.005791
8	0.017997	28	0.010151	48	0.007524	68	0.005732
9	0.017584	29	0.009949	49	0.007430	69	0.005668
10	0.017197	30	0.009740	50	0.007340	70	0.005596
11	0.016551	31	0.009597	51	0.007268	71	0.005316
12	0.015729	32	0.009425	52	0.007184	72	0.005049
13	0.015229	33	0.009278	53	0.007116	73	0.004796
14	0.014580	34	0.009123	54	0.007029	74	0.004555
15	0.014008	35	0.008974	55	0.006945	75	0.004327
16	0.013650	36	0.008847	56	0.006884	76	0.004110
17	0.013175	37	0.008698	57	0.006805	77	0.003903
18	0.012832	38	0.008581	58	0.006747	78	0.003708
19	0.012475	39	0.008457	59	0.006667	79	0.003522
20	0.012073	40	0.008338	60	0.006600	80	0.003345

White, is there already a ratingbuster that runs on the PTR using the new constant? would be awesome :P

PTR boss value somewhere around 10650, range of minus 20 to plus 280. Medium confidence level. Used bloodthirst tests with varying attack power. Also ran data with 5 sunder armors = 80% armor and various levels of armor penetration. Armor penetration effects are applied after percentage reductions from sunder armor, and assumedly faerie fire.

Needs confirmation via beast lore or whatever.

Landsoul and I spent about 5 hours last night trying to nail down the new arpen mechanics. (Bloodthirst makes for much more accurate testing, since it's doing thousands of damage, and has no damage range, less margin of error)

New boss armor value is between 10635 and 10650. Looks like 10645.

Sunder and Faerie Fire are multiplicative with each other, for sure. Sunder is -20% armor on the boss, Faerie Fire is -5%, both are -24%.

ArPenRating is acting odd. I have word from a friend of a friend that how it works on PTR at the moment is totally a bug, and will be fixed, but if you care how it works right now...
It's not multiplicative, and it's not additive. It's actually even worse than multiplicative, using some formulae we couldn't figure out after slamming our faces into this huge pile of numbers for 5 hours, if it's using the same rating to % conversion.

If you assume, however, that it is multiplicative, and they just changed the conversion ratio, it works... except the conversion ratio is different depending on what debuffs are on the target. On live, it's 15.3. On PTR it's 19 with no debuffs, 16.6 with Sunder, and 16.0 with Sunder+FF. You'd think that that meant it was additive, but that's off, in the wrong direction, to be additive.

This makes a lot of sense when I think back of the hours I spent trying to figure out why my hunter shot calculations were off. With a fixed damage ranged weapon I was trying to determine the damage done at various AP levels and the % of mitigation kept changing..

I also have to check at what point the pet special abilities that affect armor are applied.

I've tested on a rogue with Mace Specialization* ("Your attacks with maces ignore up to 15% of your opponent's armor."), and got rather weird results as well: for the level 60 and 70 dummies, Mace Specialization yielded roughly 15% less armor, and roughly 32% together with Expose Armor (85% x 80% = 68%), pretty much spot on. For the level 80 and boss level dummy, however, Mace Specialization was only roughly 12% on its own - but still roughly 32% together with Expose Armor.

I can confirm both boss-level armor, and multiplicative stacking of Major (SA/EA) and Minor (CoW/FF) armor debuffs.


*) via Gouge at fixed AP, Maces and Daggers with weapon skill 400; "roughly" as used above is +/-0.5%

No, it hasn't been that, it really was bugged before. As soon as Landsoul is on tonight, we'll redo our tests.

Also, looks like everyone gets 25% more Expertise from Expertise Rating. (confirmed with Druids, Rogues, Warriors, Shamans, and DKs)