You guys really should move the Windfury Cooldown discussion to that thread at this point, especially since there hasn't been a single post indicative of haste affecting the cooldown or anything concrete that the cooldown has been changed. A few outliers really doesn't prove anything.

The testing from the last PTR seemed to indicate that the cooldown actually works more like 3 seconds +/- 0.3 seconds in actual practice which is well within what people are showing here.

No, please.

I'd like to keep all the enhancement based 3.1 discussion in one spot. If someone believes a ninja change to the WF CD occurred, this is the place to discuss it. If the WF CD is confirmed, then a WF CD thread may be the place to discuss it - but more likely it will just be a new consideration for enhancement simulation discussion.

In summary, I don't want to see threads pop up to discuss very specific and speculative changes in 3.1 as it relates to enhance.

To those who want to discuss WF:
The burden of proof is on the tester to prove the 3.0 second cooldown has been modified. If you aren't testing this in a reproduceable manner, I dont' expect to see you posting your feelings on it. If you can't reproduce your results, keep trying and don't clutter up our discussions until you can reproduce it.

When in doubt, stfu about windfury changes. When I see people posting about this, I often think of the common "they made jboots slower" crap from EQ after every patch.

Well there's already a WF CD thread is what Rounce was pointing out. [Enh] Windfury Weapon Cooldown
Wouldn't be a new thread popping up.

When there's a confirmed WF change, have at it. Right now it's just people speculating that the WF cooldown somehow changed, and we all realize that lag and combat logging can make it appear that things have changed when they have not. Until we get some sort of broad agreement that the WF CD has changed, and how it has changed, there's little need to theorycraft it.

"Onyxia deep breathes more after patch"


Edit: The last poster on this issue indicated that, of 144 procs, 3 were under the WF 3.0 second cooldown. I don't find that compelling enough to assume that there is an unreported change to the mechanic and, thus, to fire up a theorycraft effort outside this thread. We should spend out time proving the CD has been modified, not re-evaluating how WF stacks up against other imbues with the assumption that the CD has changed in some unquantifiable way.

Didn't see it mentioned on mmo: They added all three totems that can drop in Naxx25 to the Emblem of Valor vendor (25 emblems each) and I guess, removed them also from the loot tables.

I'm now doing more tests on the wf cooldown and will also improve my parser with more output.

Edit: The new version with much more useful output is available: http://wow.nuclear-warriors.de/wflogparser.rar
Is also fixed the bug, that it didn't detect offhand windfury procs.

And here is an output of a test I just did on the PTR with 2.5/1.4 WF/WF and WF Totem:

So you see, while there are some procs under 3.0 seconds, the largest group is between 3.1 and 3.7 seconds. That speaks against my CD change theory. But I will do some more tests.

Depends entirely on your weapon speed and average haste. If you normally are haste to ~1.5-1.6 then the majority of your WF procs are going to be in the 3.0-3.2 range assuming the CD hasn't been cut in half. If the assumption is the CD scales with haste you'd have to stack a bunch of it to properly test it. It's hard to say from that data if flurry or WF totem might affect the CD, but if they did the expected CD time would still be outside the normal swing time unless other haste effects were also in effect, making the effective CD still ~3sec for the majority of the cases.

Get one of those 3.0 axes from the blue chicks, a 1.5speed weapon, or a 3.2+ speed 2H and put WF on that and try it naked/untalented no WF totem, then drop WF totem, then talent and get flurry but no WF totem, then drop WF totem, then put on gear with haste, and see how each of those affects your WF procs with a 3.0 or divisible weapon. Maybe try talentless and only haste from gear as well.

For 1H testing a 1.6 speed weapon might be better yet to get a guarantee that hit #2 is outside the WF CD. And by effective 3sec CD I mean a 2.6 weapon hasted down to 1.5 is ~73% haste, and if all that haste is applied to WF CD then the CD is still going to be 1.73 which would still make back to back WF procs on a 1H impossible (save the blue chick axe), but would eliminate the haste scaling issue where too much haste can block 2 (sub 1.5) or even 3 (sub 1.0) attacks from proccing WF.

So I am attempting to test the windfury cooldown and potential changes to it on the test server at the moment. I'll have a combat log for this once I am done testing. All of these are done with only points spent in resto talents, and completely ungeared except for an expertise trinket with no proc unless otherwise stated.

Back to back procs have been seen with the following combinations:
3.0 speed staff with no haste.
3.0 speed staff (2.95 effective) with 55 haste rating, or 1.68% haste.
3.0 speed staff (2.85 effective) with 169 haste rating, or 5.15% haste.

My question is the following: is it enough to see one back to back proc and move on to the next test?

Edited to add 2.85. Like I said, I'll post this entire combat log when I'm done. I'm doing a full 30 minutes of testing with each setup. The only one I don't have logged is the base 3.0 speed.

I'd say no due to server lag and other unknown factors, plus you really need a lot more haste than just enough to get it barely below the 3.0 mark. See two posts above yours for what I recommend doing.

For a definitive answer you have to witness a statistically significant number of procs (under 3 secs) before moving on.

Also the second test that you propose would not garner any new data as 2.95 is within the currently acknowledged margin of error on the 3sec cooldown.

Well I'm slowly moving to more and more haste. How would having a higher ping increase your likelihood of getting back to back procs? I'm not working with "time between procs" in the combat log. I'm using a single weapon with windfury with set amounts of haste, and seeing if I get back-to-back procs.

Looks about right other than the fact that one is a savings and the other is an actual return though in the end it works out the same. IMO we shouldn't be pushing for any changes to UR thinking that 2 points in ISS would make up for it. The one point in UR gives ~620mp5 assuming a 13k raid buffed mana pool (13000 / 105 * 5 = 619, 105 cause lets face it, for the 15 seconds UR is up, your mana pool is full). Leave UR alone and use ISS for PvP and 10mans and 5mans where you don't have a paladin with JoW and nobody is bringing replenishment, and you're going OOM for 15-20 seconds waiting for UR to come off CD.

Right, so at what point could we consider it to be a counterexample of significance? 2.8 seconds? 2.6?

Me personally I'd say at least 5% haste from gear, 10% would be ideal, and a long duration test showing multiple back to back procs to prove it's not a "one of" thing but repeatable. Also like I said above throw in WF totem and Flurry in various ways in separate tests and see if there are any back to backs with those various combinations to verify if all haste would hypothetically affect that CD or just certain kinds of haste. Heroism testing would be super too but would require assistance from a group of same faction shammy's to chain it for you and at least one cross faction friend to murder you after each buff wore off so you could get rezed and go again.

Another test and it doesn't look good:

WF/WF with two 2.2 speed synced weapons, 1.21% haste, WF Totem, only auto-attack:

With flurry I have 1.34 attack-speed, without 1.81. If haste (including WF Totem) would lower the CD, we should see a large number of procs at 2.6 (1.34 * 2). But the big number is at 4.0 (1.34 * 3), with some at 3.2 (1.34 + 1.81) and 3.6 (1.81 * 2). There is still the posibility that only non-specific haste from items, procs or Bloodlust lowers the cooldown, but I wont be able to test this today, but maybe tomorrow.

Take a look at the Windfury Cooldown thread. The most coherent theory we could come up with is that there is some calculating factor occurring on the server that adds a period of inaccuracy/give to the cooldown. So even if the cooldown is exactly 3 seconds there is enough give in the system to occasionally see procs that are less then 3 seconds apart and there will also be times when the cooldown will interfere with procs that are more then 3 seconds apart.

The only way I was able to consistently get a proper proc rate (ie no interference from the cooldown) was with weapon swing times over 3.3 seconds. The original concept was that it had something to do with latency between client and server but that was disproven with a test involving pulling the network connection so the server was working without any info from the client.

So my current theory is that there is a cycle that the server follows and it only examines input from you on set intervals (0.3 seconds being such an interval). So it gathers client info over 0.3 seconds and then looks for the proc and checks against the current cooldown so the only way to never clip the cooldown was to have an interval between attacks longer then that processor interval. I have no clue how to confirm/deny this theory but it fits the evidence and keeps me from bashing my head into my keyboard until both are bloody.

If the theory is right that would mean the WF cooldown would actually be more like 3 +/- 0.3 seconds which means WF would fall onto a bell curve which means the outliers are just that, outliers in the extreme end of the curve. (there is a pretty graph in the thread that demonstrates the concept better then I can explain).

Nothing I've seen so far disproves that theory, but since it is only a theory feel free to poke holes in it.

By the way, that parser is really nice. I wish you could talk to the guy who does procodile and show him how to get his mod to recognize that even though WF is two hits it really is only 1 proc so it would actually give accurate PPM and cooldown info.

If you posted the combat log of the test we could look at the actual time for the 2.6 and 2.8 second gap attacks and see if there was significant lag between the swing and the windfury proc. Or it could illumintate the exact combination of flurry that might produce that situation.

A lot of data is/ has been gathered on WF CD v’s haste.
I think it would be an idea to approach this from the other end as well.
Hitting a dummy with a slow weapon should produce 18% WF proc’s, if the cd is affected by haste then the proc rate should remain unaffected by haste.

Rounced has proposed a +/- 0.3 sec tolerance on the 3 sec cd, so another test would be to use a 2h at 3.4 or greater no haste / talents.

Hit the lvl 60 dummy for x. WF hits should be x * 0.2 wf proc rate. (0.22 glyphed)

Then haste the weapon to 2.6 or less, hit the dummy again.

The number of WF proc's should be 16% if the cd is unaffected by haste
(Hits – (no. of hits within WF cd ) * wf proc rate
(x – (x * 0.2)) * 0.2 = 16% (17.16% glyphed)

I’m not sure how much testing would be required to avoid RNG.

e) made it more readable

Why are you using 0.18 as the proc rate? If you remove the cooldown from the situation by using a 3.4 speed weapon or slower you should have a 20% proc rate (22% with the glyph).

I did exactly this kind of test in this post.

Didn't use the combat log or the parser, just had recount running cause all I was interested in was the proc rate.

Had a 22.2% proc rate with zero haste, and just over 18% proc rate with Flurry and/or WFT. All tests had the WF Glyph. Only did 1000 hits in each test, but gives you a sort of feel for it.

Those tests generally would imply that the ICD is not affected by haste, as the values are the expected ones - 22% with no hits during ICD and 18% with one hit during ICD. However, due to the existence of flurry to haste the weapon the tests becomes unclear. It is quite possible that, assuming an ICD change by Blizz due to haste, flurry does not affect it. So only haste from gear would decrease the ICD. Also flurry acts non linearly on haste, so it's hard to draw conclusions.


There should be a very easy way to test it, and I will try to do it tonight when I get home, but more data would always help.

Test 1 - baseline:

Take any weapon you want, 0 talents, 0 haste, no WF totem and start hitting a level 60 target dummy. Run this until you get a fair bit of WF procs (I'd say 500 should be the minimum). Run it through Sentinel's program and graph the results. We expect some form of normal curve (at 500 it might not look like a perfect normal, but it should be enough to find where the normal is centered)

Test 2 - hasted

Take a slower weapon (preferably over 3.0) and haste it until its' speed is equal to the baseline weapon speed. (this haste would be preferably done without WF totem). Do all the above again.

Look at the two graphs - if the normal's centers are near each other then there is no gain from haste in ICD. If the test 2 center is lower than test 1 then the haste affects the ICD.

Then we can do some other tests like try to haste using WF totem + gear haste to obtain test 1 weapon speed, and see if WF/gear affects the ICD.

Something small that I neglected to mention, the arena lava lash totem is a 10 second duration on PTR as opposed to the 6 on live.

Point for point, SF is very slightly better (4 mp5), but I.SS has 2 ranks. 100% chance of 20% base mana at 2/2.