Quick summary
1. Bosses have 2.1 magic crit reduction.
2. Dark Conviction is active only if you have weapon equipped. Ebon Plaguebringer is always active.
---

While a melee crit depression rate of 4.8 against bosses is well known, I did not find such a number for magic.
The biggest topic on this subject I could find had speculations on the figure being from 0 to 3, but again no hard data.
Vontre's DK sheet for example shows basic magic crit chance as +0.91.

All of the tests performed on the boss dummy in Darnassus given how it's a place no one ever visits and won't corrupt the data with possible debuffs.

Test 1 (data removed, appears to be corrupt)

Test 2
Gear: 108 hit rating (4.12), 199 crit rating (4.33)
Talents: 4 in dark conviction, 3 in rime
IT = 3851 casts, 511 miss (13.27), 567 crit (14.72)
DC+BB = 6800 casts, 871 miss (12.8), 122 crit (1.79)

(19.33-X)*0.8712=14.72; X=2.43 [6.43 if DC works]
(4.33-X)*0.8712=1.79; X=2.27 [6.27 if DC works]

Test 3
Gear: 108 hit rating (4.12), 199 crit rating (4.33)
Talents: 2 in dark conviction, 3 in ebon plaguebringer, 3 in virulence
DC: 4370 casts 423 miss 195 crit
IT: 5733 casts 560 miss 259 crit
BB: 5708 casts 543 miss 296 crit
All: 15811 casts, 1526 miss (9.65), 750 crit (4.74)

(4.33+3-X)*0.9012=4.74; X=2.07 [if only EP works, 4.07 if both do]

Test 4
Gear: 108 hit rating (4.12), 199 crit rating (4.33)
Talents: none
DC: 1758 44 crit
IT: 2650 46 crit
BB: 2610 52 crit
All: 7018 142 crit (2.02)

(4.33-X)*0.8712=2.02; X=2.01

Test 5
Gear: 236 hit rating (9), 539 crit rating (11.74)
Talents: none
IT: 5659 477 miss 482 crit
DC: 3982 335 miss 348 crit
BB: 5622 429 miss 522 crit
All: 15263 1241 miss (8.13) 1352 crit (8.85)

(11.74-X)*0.92=8.85; X=2.12

His sample sizes should be large enough to remove any excessive variation. Also, the number of casts in the first trial represents roughly 8 full hours of using Icy Touch.

Anyway, first off there's a flaw in your logic. If there's a hidden crit modifier on bosses, which there definitely is, it will decrease your crit rate, not your hit-modified rate. So, for example, with Icy Touch in the first trial:

Theoretical Crit Rate: 19.7%
Hit Rate: 83%
Actual Crit Rate: 12.98%

Rather than saying:
19.7 * .83 - 12.98 = variation

You want to say

(19.7 - variation) * .83 = 12.98.

If we do this in the first trial, working backwards

variation = 19.7 - (12.98 / .83) = 4.06
variation = 4.7 - (.04 / .83) = 4.65

In the second trial:
23.33 - (14.72/.87) = 6.41
8.33 - (1.79/.87) = 6.27

Third:
9.33 - (4.74/.9) = 4.06



Two things to take away:
1) The numbers are still too different from each other. This isn't complete, something is missing, but my gut tells me we have everything we need to solve it.

2) If 4.8% crit is removed, you would not have a single critical strike during the BB/DC phase of your 1st trial, unless there's always a possibility to crit.

Following from where I left off: We can deduce that crit is not reduced by a percentage, but by a flat value, because of trial one; It'd have to be a relatively low percentage for the first portion and a huge percentage for the second portion. Doesn't hold up. So we're definitely looking for the effect of a flat value somewhere.

Lets consider the possibility that crits cannot miss; This removes the division from our calculation, turning it into
Actual - Expected = Variation

Edit: Is this right? It feels wrong, but I don't have time to sit down and think about it anymore right now >_>

Lets see what happens.


First Trial:
19.7 - 12.98 = 6.72
4.7 - .04 = 4.66

In the second trial:
23.33 - 14.72 = 8.61
8.33 - 1.79 = 6.54

Third:
9.33 - 4.74 = 4.59




These numbers don't feel anywhere near as clean. The almost-exactly-4% chance reduction cropping up twice in my last post haunts me.

Are you taking into account the "one roll" system? Getting more hits doesn't get you more crits. So, testing for crit depression shouldn't involve HIT gear at all. Take a single magic attack (IT is a good one for us) and get a sample for both boss mobs and non-boss mobs. Compare. Test again with more (or less) crit and do the same thing. Compare. At the same time, you can even check for crit depression on melee.

Did you chaperon your DK for the entirety of the tests? Given the size of the samples, I image that the tests were at least in part automated. It is not all that uncommon for players to go to Darnassus for dummy tests. In 15 hours worth of testing there your test results might have very well been skewed by Improved Scorch, Winter's Chill, Focus Magic, Totem of Wrath, Heart of the Crusader, or another DK putting diseases on one of the surrounding non-boss dummies.

The Darnassus boss dummies are well within BB range of many other non-boss dummies. Since BB works off other DK's diseases as well, you may very well be critting off another DK's diseases on non-boss dummies around the boss dummy you used for the test while you were AFK. If you still have the combat logs from the tests, you may want to reformat your results with only IT numbers and DC numbers, taking out the most likely source of data corruption (with the next most likely source a distant second in Totem of Wrath). Do future tests in Shaman/Paladin/Mageless Ebon Hold, counting only IT and DC.

Dark Conviction not adding to spell crit would certainly explain a lot though, as my test numbers never quite match my theoretical numbers for builds with it.

No, we're saying that against a boss, your chance to crit is reduced by some small amount.

Sorry, pre-coffee post.

I'm going to parse some icy touch spam and edit it in here for more data ASAP.

As for Dark Conviction not working or even showing up in the char window - before you edited the original post, you mentioned that the only gear you were wearing was the stuff without durability. That would mean no weapon, something that seems to be a requirement for DC to work. (see Dark Conviction - Spell - World of Warcraft)

Sarkus, you are right, tests 2 and 3 were performed without melee weapon, and that tooltip, wow. Thanks.
Once we'll establish the final number on magic crit depression I'll retest with and without weapon to confirm.

Meanwhile I got a crit wearing 2.16 after 1200 casts and now running 2.05 for 2100 casts without a crit and counting.

Heirloom weapons don't have durability, which makes them ideal for testing.