Well, it was asserted on the last page that the proc rate is 7% per some datamined information - I can neither confirm nor deny that. What testing we do have seems to generally support that, however. So while I will never oppose additional testing, I don't think its particularly urgent to verify its behavior, either.

Personally, I think the mechanics most in need of testing are further verification of the legendary set bonus proc rate, plus additional verification of the proc mechanics of Vial of Shadows. And perhaps Starcatcher Compass as well, as I haven't seen any positive verification of its ICD yet.

If you can find two target dummies within Blade Flurry range, it should be a fairly simple matter to just autoattack with Blade Flurry up for 10 or 15 minutes with a cancelaura macro and see how the number of procs compares to the number of attacks. Not sure off the top of my head what cities that might be true of.

Just did an 18m 42s test with 2234 haste, 1004 mastery, and no haste or mastery procs on the gear.

1732 Blade Flurry Hits
1560 Melee Hits (790 Hits, 316 Crits, 454 Glances)
173 Main Gauche Hits (148 Hits, 25 Crits)
210 Misses
166 Suffering Procs

And just in case it makes any difference, 443 Burning Wounds ticks

So, it looks like it's still about 10%

That's around 5%, which means it does not proc from blade flurry. I know that's almost definitely what you meant, but saying "it's still about 10%" pretty strongly indicates that blade flurry has a chance to proc it as well and it might cause confusion.

I have heroic Vial of Shadows. If you can describe how exactly it should be tested, I will do it and share results and/or combat logs.

Yeah, I should have worded it better in that it's still roughly 10% of the direct melee hits only, but that's what posting late at night gets ya.

A complete test would be two-fold:

For starters we need to figure (confirm?) the damage formula. The working model we have is [base + modifier * AP] where base is the damage you see in the tooltip when unequiped and the modifier is one of (0.797, 0.9, 1.016) deppendant on the version (lfr, nr, hc). Just noting down the damage ranges (or averages) of the equiped trinket at a couple different stages of AP should suffice. This correlates with the one I had in ptr but it could have changed ever since; if it doesn't fit, we need to either find different modifiers or a different formula.

Ultimately it'd be nice to have a log with a handful of procs to pin down the 'real' damage (tooltips come with errors sometimes), and the rate at which it crits. All tests should be performed with the same AP and crit modifiers (no landslides or other procs) and the same armor on the target. Note that the damage should be mitigated by armor (~.71 with sunders or ~.68 without). These kind of tests are a bit tricky, because the damage range is very large; if the sample is big enough we can find the largest and lowest hit (or crit) and see if they fit inside the assumption or we could use averages in much the same fashion. The main problem is the icd (25''), which would produce about 140 procs in 1h of testing (not too many) but it's a start. It's also worth noting that tests done with the least amount of AP and crit are better (the variance will be much lower) and without poisons, if combat specced, to avoid savage combat.

I posted them to WOL <log1> <log2> <log3>. The expected tooltip ranges would be 16040,81-22861,81, 24461,42-31282,42 and 31514,49-38335,49; there are discrepancies in the higher ends that I would blame on (rounding?) issues on the tooltips we're given: I suspect the modifier is 1.02 instead of 1.016 and the tooltips are off by some amount. From the combat logs I'd expect these ranges for .68 armor mitigated hits: 16633,76-21272,04, 21429,85-26068,13, 10907,75-15546,03; we're getting, again, hits that are off the charts by ~150 damage in the higher ends, but also, the whole test seems to be shifted by that same amount (not so much the 2363AP one): they lack the lowest hits in a ~120 range and the average suffers from that too.

All in all, the model seems to be correct, but the exact figures are somewhat off (we're talking about ~6dps difference) and I'm open to suggestions to fix that, but overall I'm willing to call it good.

Well, for the discrepancies on observed values, its important to remember that armor mitigation against an undebuffed target dummy is not 68% but 26070 / (26070 + 11977) = 68.52...% So, for instance, for the highest-AP set, the expectation based on our current understanding would be 31514.49 * .6852... = 21593.89 to 38335.49 * .6852... = 26267.67, which aligns really pretty well with the 3rd parse, which (by brief examination) seems to range from 21706 to 26222. The tooltip discrepancies still need to be investigated, but at least we have good agreement between the tooltips and the in-game testing.

Any chance you could run a short test for FoK? Would be nice to see if it would be that powerful on for example Yor'sahj.

"Brief" Fangs of the Father testing. I'll probably do a few more in the following days. I did it as Sub, and didn't really follow much of a rotation. I messed around with 1CP finishers and energy management while trying to see if I could game the proc.

Right now, it seems like a decent assumption that every spec uses the same formula to proc the finisher portion, otherwise it begs to ask why they would give Combat a reduced proc chance a stack of agi. This should definitely be double checked in the future, but since Sub has the highest chance to proc a stack, it would probably be helpful if most tests were done as Sub so as to not waste time.

World of Logs - Real Time Raid Analysis
Proc'ed at stacks: 43, 39, 34, 43, 32, 37, 43, 47, 36, 31, 44, 45, 35, 34, 41, 34, 38, 32, 43, 30, 45, 34, 48, 47, 45, 34, 47, 38, 32, 41, 37, 34, 37, 40, 44, 39, 31, 43, 30, 31, 48, 39, 47, 30, 42, 46, 31, 39, 41, 45, 30
Average stack: ~38.92
Range: 31-48
Mode: 34

I plotted it out, but with such a small sample size I saw no real pattern, though the distribution looks somewhat flat given the limited data. It wasn't like the 45+ stack procs spikes were double the size of the 30-35 stack ones.

Unfortunately, these take a lot of time so I won't be able to add a lot quickly, but I'll try and do some more soon.

Update, added another 80 or so procs from tonight, still chaotic distribution.

Mean	39.442
Median	39.5
Mode	30
Occurances	240

Update (Jan 25th): Another 110 procs, looks smoother, but still the same general trends.