I can attest to this. I had the Hourgless trinket as soon as i hit 70 one week after release and haven't noticed a double proc, even thought I've been watching for it.

i can confirm that
i have the trinket for about 2 weeks and never had a double proc at all

In the days after I got my Hourglass, I started noticing myself that there appeared to be a cooldown window on it before it could proc again. This was before I saw this post. I strongly feel that this is the case. The best example I've found of this is having it proc when I pull a mob with thrown, and it will never proc again before the mob dies.

If there is truely and invisible cooldown, then someone needs to find out exactly how long it is. Best way would be to beat on the servants in BL for an extended time(go afk with heroism or something) and run a combat log. Parse through the log and figure out the shortest time between procs and that would be our best guess on this cooldown. Why blizzard would make an invisible cooldown on something like this instead of fixing it to a balanced procrate is beyond me, but if the cooldown idea holds up, the procrate of the trinket could actually be 10%, and would be trivial to model that way.

There is definitely a cooldown, I got the trinket about ten days ago and have a buffwatch bar, I would have noticed a back to back proc.

So I turned on procwatch and recap and hit the servants in blasted lands and tried a few things.

One, I tested with my armor on and off, to see what the difference in buff uptime would be (was curious if blizzard had modeled this trinket PPM style so stacking crit wouldnt help.)

The result after a bit of combat was noticable... I didnt bother to record or calculate the data for w/gear on and high critrate because I believed that the crit was helping it proc more often after all and did not need that information.

I had a 5% uptime with 12% crit, 948 hits+crits. This was with the gear off. The chance on crit for this period was 7% (at 22% crit I had roughly the same chance on crit.)

Obviously with raid buffs I would have a little over 10% uptime.

I never saw it proc within less than a minute of itself, but didn't lock down a cooldown after all, because I havent figured out a way of gathering the time between procs to use pf's suggested method.

If the trinket was PPM style modeling, changing your gear would make a noticable difference, PPM is used to define the base % procrate for weapons with different speeds so that they all would have similar dps (the side effect of this is that instant attacks use the same value and thus slower weapons become favored). If they trinket has an invisible cooldown, changing your critrate would make a much smaller impact.

The best way to test this is to /combatlog and then view the results afterwards. Then parse a lot of attacks with a decent critrate and only look at the "you gain buff" "you lose buff" lines. Note the smallest interval between "you gain buff" and also double check the you lose buffs to make sure it was not a double proc. Technically, the best way to do it would be to take an average time between buffs noting your crit % and then using statistics to back into a better number, but that requires more effort than I want to spend and simply noting the smallest time between buffs should be fairly accurate if you have enough samples.

I could write a PHP parser for this (if you give me enough motivation ;)).


// Edit
Done. If you send me your combatlog, I can run the parser through it.

I scanned the last few pages and didn't see a report of it, but the +2 dmg ring enchant does nothing for me in either the Excel or OO formats. Sorry if that's a known issue.

post #549

Anyway, does anyone have any solid data on the hourglass trinket and the supposed invis cooldown? Would like to get that modeled up correctly and release my final version, (unless I leave a major bug of course) and then hopefully someone else can continue adding the new items and minor bug fixes.

If someone supplies me with a combatlog, I could find out the minimum time between procs and also the crit/proc ratio. Yet, I still need such a combatlog, since I do not have the trinket.

I'll see what I can do about putting my Hourglass to work I suppose.

Hm, is it possible that the sheet still uses the level 60 base hp stats? When entering my gear I see signifcant difference between the health I have and the values the sheet gives.

Well, by quick inspection I'm seeing the quickest time between two procs being 45 and 47 seconds.

Total Procs: 21

Shortest Time Between Procs: 45 seconds (35 seconds between fade of buff and next proc)

Intervals sorted from lowest to highest on 21 procs:

45
48
48
50
53
53
54
56
57
58
59
61
64
72
74
76
78
81
84
99
117

(In seconds between procs)


Edit: Did some further data parsing and if you operate on the assumption that the proc has a 45 second cooldown, then you can start counting "eligible" crits that could cause the proc. When you do this, you find that outside of the 45 second cooldown window there are 186 eligible crits that result in the 21 procs listed above = 11.3% proc rate. So the thottbot: http://www.thottbot.com/?sp=33648, info is correct it's just incomplete in not mentioning the hidden 45 second cooldown.

Edit #2: Preliminary rough calculations then show that assuming, SnD:

(Combat Daggers) 1.8 Spd MH, 1.3 Spd OH, 30% Crit = Avg time between procs = 61.2 Seconds (avg of 16.2 seconds to proc once it's eligible)

(Combat Daggers) 1.8 Spd MH, 1.6 Spd OH, 20% Crit = Avg time between procs = 70.61 Seconds

That means that even under the best of conditions (example 1) the proc on the trinket translates to:

10 Seconds/61.2 Seconds * 300 AP = 49 AP.

However it's still almost 2x better as Bladefist's Breadth as the proc for 300 AP over 10 seconds once every ~65 seconds would be way better than the 200 AP for 15 seconds every 120 seconds. The big benefit of Bladefist's Breadth being able to consciously maximize the usage. With Unraveller you could easily proc right before a mob dies and have it tick while you're transitioning to another mob.

If there were no CD on the proc then it'd be:

10/16.2 Seconds & 300 AP = 185 AP

And admittedly a trinket with 32 Crit Rating and 185 AP would be pretty broken.

I just noticed this also. It doesn't matter what 2 pieces of Netherblade you equip or what you're replacing there's a huge dps drop when having 2 pieces equiped. I thought I was going crazy when I switched between 1 vastly inferior piece to Netherblade (any piece) to get the 2 set bonus and there would be a 30-40dps drop. Glad to see I'm not just losing my mind on this one and that someone else has noticed this also.

ty for the log bleedo. Anyway, useful info from his log if someone wants to parse it for me as I suck at programming now is:

Total crits in the log.
Crits per second: totalcrits/1456

Assuming the true proc-rate is 10%, we can then use the crits/sec number to find estimated time to get the buff after cooldown expires as: 10/critsec

Using the average time between buffs (66 seconds from lace's data), we can subtract the expected time to get a buff and we should end up with the invisible cooldown which we would expect to be close to 45 seconds. Also, after finding this value, use the minimum of this value and 45 to determine the following:

Average # of crits after cooldown(as described above) expired to proc buff. This should be fairly close to 10, and would be of good use to double check our understanding of the trinket.

How would Abacus of Violent Odds compare to the crit trinket? Personally I feel it gives more dps for me with blacksmith mace but havnt been able to parse any real numbers.

After using 30 seconds as the invis cooldown time, I came up with an average 11.36 number of crits to proc the buff. This should be within a standard deviation of the 10% from thottbot. If anyone wants to do the math on that to back me up, go ahead, but my statistics is fairly rusty at this point and i'm not cracking a book to see the formulas for all the distributions. I'll use 10% and 30 seconds as the invis cooldown in the model on the spreadsheet. Very odd that blizzard chose to model a trinket such as this. Maybe its to keep it from scaling out of hand with +haste and +crit in the future, but it should be made clear this is how it works instead of lie about it in the description.


With the latest patch, the abacus is a very strong trinket. for sustained dps, it is probably one of the strongest on-use trinkets, and I would probably say it is the strongest for a combat potency build. Also, if haste rating works exactly how i think it does now (ie all haste ratings are added together instead of multiplied), stacking haste rating means each +haste you get is weaker than the one before (whereas in the past it was stronger). Still, +haste rating is the best raiding rogue stat for sustained fights without question.


Not only possible, its very likely. If you could, give me how much hp/stam you have naked and geared and i'll try to make the amount more accurate. (also, its likely that dodge is off as well, so if anyone wants to check the base dodge%, feel free)

I'm curious as to why you're going to use 30 seconds as the invisible cooldown number when the shortest time between procs was noted as 45 (35 if you start the cooldown when the buff fades.)

Edit: Oh i think i see the reasoning.

Crits per second:
581/1456= 0.399

Expected # of crits until a proc given a 10% procrate = 10.
Expected time this takes: 10/(crits/sec) = 10/.399 = 25.06.

Average Time between crits from parsing: ~66.
Estimated Cooldown based on this data: 66-25 (-10 if you count buff) = 31

I highly doubt blizzard would choose 31 seconds as their cooldown timer, so I am using 30 seconds (or 40 from when the buff procs) based on statistical data. If we had a much larger sample, using the shortest time between procs would be a decent data point, but I feel (based on the data I have), that 30 seconds is a better approximation, and will use that unless someone wants do testing so we have a larger data set.

Collected data confirms your suspicions:

Before the patch:
2.70 / 2.00 (Drakefist / Nightmace)
1.41 / 1.04 (SND, BF, Drakefist proc, 2pc Assassination)


After the patch:
2.70 / 2.00 (Drakefist / Nightmace)
1.29 / 0.96 (SND, BF, Drakefist proc, 2pc Assassination)


Also the tooltip (as of the latest patch) displays:

Drakefist proc - Haste Rating 200 (19.02% Haste)
2pc DS3 proc - Haste Rating 160 (15.22% Haste)
Drakefist + 2pc DS3 = Haste 360 (34.24% haste)

The new attack speeds = 2.01 / 1.49
The old attack speeds = 1.97 / 1.4


Something to note:

- Procs that give +Haste stack additively now (as in, it uses the total Haste% rather than each one by itself)
- Abilities that increase attack speed (Slice and Dice, Blade Flurry, etc.) as still using the old formula for stacking. Hence why my new attack speeds are actually lower now than they were before the patch. Essentially, the buff to Haste Rating is slightly outpacing the old method of stacking all Haste effects at the low end. Still need to run math to find out where you start to the two methods even out again.


I assumed that since they were buffing +Haste that they must've made a change to the way it stacks, otherwise it could pretty easily get a little overpowered. It certainly appears now that it just adds up the total Haste percentage / rating and calculates it that way (A quick 2.70/1.34 confirms it) Note: Abilities that increase attack speed are not included, they still stack in the old multiplicative fashion. I can only see this change really affecting players that go all-out stacking Haste, in the order of six or more haste buffs where the stacking nature of +Haste worked in your favor. With the new changes in place, players only picking up a handful of haste procs and effects will actually see lower attack speeds, which I think is a polite way of Blizzard saying "We like that you're stacking these as we intended, but we no longer want to reward you for giving up other things just so you can get more +haste".

I dont think they changed how haste stacks with the recent patch. I think they changed how haste stacked when they added haste ratings.

Hastes that are not given with ratings should still stack multiplicatively. Hastes that are determined from haste ratings will probably now stack additively because of how they implemented haste rating (add all of chars haste rating and determine haste %). So 158 haste rating is 15% haste. Add 158 more haste rating and we have 30% haste instead of 1.15*1.15 = 1.3225 or 32.25% haste which would have been given before 2.0. It is only a minor change since SnD/BF should still interact multiplicatively.