Wow, that makes me wonder if wound is somehow different. I have a total of over 4,000 instant and deadly procs between 1.83 and 2.13 without a single crit.

If it matters, I am not in beta. My testing is on the live US servers.

Well I'm using Wound Poison so I'm applying Master Poisoner, that's the only thing I could think of.

I have master poisoner up the whole time as well. I wonder if Blizz raised the crit reduction with the maintenance this week.

In that case yours is probably the correct one, however I'm certain that there was no outside buff introduced since I was watching it by myself watching it and thinking I didn't expect a crit with that little crit%. I'll test it again tomorrow when servers get back up, though.

My data got erased, so I started again with [Heartpierce], [Steel Bladebreaker] one agi-crit gem and crit food: 2.18% spell crit (100rating), and got crits in medium sized samples:
IP/DP

WP/WP

I've encountered a crit with 98 crit rating (2.13%). (I also saw one with 100 rating.)
There was a warlock near, however I specifically checked for any occurrences for Shadow Mastery in my combat log, and they were none - indeed the string "shadow" didn't even turn up once.




I'm currently testing with 95 crit rating (2.07%) and have yet to see a crit after 5047 procs.


// Edit
Added combat log if anyone wants to double check.

// Edit #2
So 7700 procs later I haven't seen a crit with 96 crit rating (2.09%). Need a break, maybe I'll continue testing with 97 or 98 again (in case my previous finding was a fluke, which I doubt however) later.

9900 procs with no crits implies that the actual proc rate is 99% likely to be below .0465%; hence, these tests are consistent with a crit reduction somewhere between 2.066% and 2.135%. So it's "about" 2.1% according to most testing we've done.

Yeah, I just finished my own testing in lieu of sp00n's.

97 crit, 8378 proccs, no crits.
98 crit, 844 proccs, 1 crit.




Edit: Added results from a crit with 98 crit rating.

Not sure if it's been discussed here, but there is a bug when using double wound poison where the proc from the other weapon's poison will always cancel savage combat, hence the low uptime.

EDIT: Upon a quick test this also seems to happen with only one weapon coated with wound poison. Every proc cancels savage combat and it takes quite a while for it to reapply. This problem goes away though when using deadly poison on the other weapon. Whenever deadly poison gets applied to the mob savage combat appears and stays there regardless of following wound poison or deadly poison procs.

I instructed a friend with beta acces to do some tests. He used a lvl85 premade, stripped off trinkets and two dirks with 5988AP, no buffs nor debuffs on the Exodar dummy. Goal was to run three tests with different talent choices, checking the combat log ingame; test1:no talents, only the assassination specialization taken: Mutilate and Assassin's Resolve; test2: three points in Lethality; test3: fill up the assassination tree and take 3 points in Opportunity. My mate took (as I told him) about 2-3 different values for each cell in this table

	MH hits	OH hits	MH crits	OH crits
No talents	1099-1101	668-669	2198-2201	1336-1338
Lethality	1099-1101	668-669	2527-2531	1537-1539
Opportunity	1429-1431	869-870	3288-3290	1998-2000

Using the formula (normalized weapon damage+201)*1.5*Assassin's Resolve*armor reduction Assassin's Resolve being 1.15 and armor reduction: 26070/(26070+11977); and modifying it by

	MH hits	OH hits	MH crits	OH crits
No talents	1.0*1.0	1.0*1.0 and 0.5 weapon damage	2.0*1.0	2.0*1.0 and 0.5 weapon damage
Lethality	1.0*1.0	1.0*1.0 and 0.5 weapon damage	2.3*1.0	2.3*1.0 and 0.5 weapon damage
Opportunity	1.0*1.3	1.0*1.3 and 0.5 weapon damage	2.3*1.3	2.3*1.3 and 0.5 weapon damage

We get the following (note that the two values account for the high and low-end of dirk's damage):

	MH hits	OH hits	MH crits	OH crits
opp=0 leth=0	1098,19-1100,56	667,89-669,07	2196,39-2201,11	1335,77-1338,14
opp=0 leth=3	1098,19-1100,56	667,89-669,07	2525,84-2531,28	1536,14-1538,86
opp=3 leth=3	1427,65-1430,72	868,25-869,79	3283,6-3290,67	1996,98-2000,51

Every value from the tests fits into the expected outcomes. Testing ended when a warrior approached and I promptly told my friend to stop in fear of sunders and physical vulnerabilities applied; that may explain the not so well populated data on crits with maxed opportunity and lethality.
I think we can conclude the damage formula is pretty much as expected: both hands get full bonus, lethality adds additively to the crit bonus (230% damage that is) and opportunity adds multiplicatively to the whole formula. For completeness in this post, I may add the 20% from poisoned target that adds multiplicatively too as sugested in my previous testings.
To wrap it up: (1+crit?*(1+lethality+meta))*opportunity*(normalized weapon damage*MH?+201)*1.5*1.15*1.2*26070/(26070+11977) where crit? should be 0 on hit and 1 on crit, and MH? a 1-0.5 binary value to account for main hand or offhand mutilates (I took the liberty to throw the meta in there since we seem to be getting the same +3%crit damage kind)
I must admit the tests show a sift towards the expected high ends; I don't know if this is due to lack of data or the inconsistency I find in dirk's tooltip (1-3 damage/1.6 speed should render 1.25dps and not 1.4), or something I completely missunderstood in here. Is there anything else I could include in this?

>edited a thousand times for clarification
>edit 2: It's worth noting that Mutilate no longer can be taken with DW specialization. The DW penaly seems to be consistently applied to the normalized weapon damage and not the 181, 201 at 85, part of the formula (and this also applies to other fromulas arranged in the same fashion).
Also, I'd like to know about the oddities Aldriana refers to when talking about Opportunity. I remember some speculation about it not being applied when attacking from the front (due to mutilate being the only attack affected by the talent that could be used from the front after BC), but from what we've seen I think that was ruled out.