[Melee Combat] How does Flurry Work? - Page 2

Lactose · 06/06/07, 9:00 PM

What I'm getting:

Flurry is attached per weapon (off-hand vs main hand), lasting for 3 swings on said weapon.
While Flurry is active (on either off-hand, main hand or both), both weapons get hasted.
A crit reset its weapon's duration to 3 swings.
When no weapon has a Flurry attached to it, the Flurry effect fades.

slant · 06/06/07, 9:14 PM

No. If you proc flurry mainhand, offhand attacks definitely consume charges.

I still think my theory stands up.

Disquette · 06/06/07, 9:22 PM

Originally Posted by Vulajin

Well, let's start with what we know about Flurry:

- The length of time a weapon swing will take is determined after the previous swing with that weapon.
- When you crit, you gain a buff with three charges.
- When you swing while Flurry is active, the time-to-next-swing is only 70% of the normal length.

Looking at the upper yellow region in your screenshot, let's see what's going on. First of all, you MH crit at 542.656 and proc three fresh Flurry charges. At the same time, you were already in Flurry, so your next MH swing gets the haste, and your ongoing OH swing was already hasted.

Now, your OH hits at 544.187. Because Flurry is up, your next OH swing gets hasted. You should have 2 charges of Flurry left.

Your MH hits at 544.671. Flurry is up, so your next MH swing gets hasted. You should have 1 charge of Flurry left.

Your OH hits again at 546.312. Flurry is up, so your next OH swing is again hasted. You should be out of Flurry charges.

Here's where weird stuff happens. Your MH hits again at 546.875 and gets hasted. Now you're at -1 Flurry charges. O.o

Your OH hits at 548.5. Its next swing is not hasted (as you can see the next OH swing comes with a 2.625 delay at 551.125).

Your MH hits at 548.718. Its next swing is not hasted (as you can see the next MH swing comes with a 2.594 delay at 551.312).

Finally, at 549.296, over two seconds after the last hasted hit (and three seconds after it should have been used up), Flurry wears off.

Now, my potential explanation for this is that Flurry doesn't wear off until the last Flurry-effected swing connects. However, it also seems to be affecting one more hit than it should. The "extra" hasted swing (by the MH at 546.875) doesn't seem to come soon enough after the third hasted swing to be able to be considered a lag effect. In addition, it's not likely that the first hasted swing incorrectly used a leftover charge of Flurry from before the 542.656 MH crit, because the first swing after that crit was over a second and a half later, plenty of time for the new Flurry effect to replace the old.

Very nice explanation and walkthrough of the combat process, thank you. I think it fits with what Pater wrote as well.

I'll have more of the excel file (well, screenshots and the whole .xls) for viewing tomorrow.

I wonder if it's as simple as saying "flurry grants you 4 charges instead of 3", and that the rest of the confusing aspect came from me not interpreting the data correctly. Ah well, on to the g/f's house, I'll resume thoughts on this tomorrow :p

Disquette · 06/06/07, 9:24 PM

Originally Posted by slant

You may have parried before that 1.25s swing, too.

Well aren't you a smart cookie!

6/6 11:06:58.781  You hit Servant of Allistarj for 290.
6/6 11:06:59.062  You hit Servant of Allistarj for 104.
6/6 11:06:59.156  Servant of Allistarj attacks. You parry. <-------
6/6 11:07:00.031  You hit Servant of Allistarj for 269.
6/6 11:07:00.234  You gain 60 health from Healing Stream Totem VI's Healing Stream.
6/6 11:07:01.046  Servant of Allistarj misses you.
6/6 11:07:01.187  You hit Servant of Allistarj for 101.
6/6 11:07:02.140  Flurry fades from you.

thanks - sometimes i miss the rather obvious in my attempts to cut down on the noise in the combat log ;-)

Xerophyte · 06/06/07, 9:29 PM

The 3 charges per hand theory doesn't appear to hold water either. From the provided log.

6/6 11:07:38.703  You hit Servant of Allistarj for 95.    - OH
6/6 11:07:39.515  You crit Servant of Allistarj for 488.  - MH
6/6 11:07:40.484  You gain Flurry.
6/6 11:07:41.203  You hit Servant of Allistarj for 102.   - OH, 2.500 delay
6/6 11:07:41.703  You hit Servant of Allistarj for 261.   - MH, 2.188 delay
6/6 11:07:43.437  You hit Servant of Allistarj for 106.   - OH, 2.234 delay
6/6 11:07:43.750  You hit Servant of Allistarj for 281.   - MH, 2.047 delay
6/6 11:07:44.109  Flurry fades from you.
6/6 11:07:46.031  You crit Servant of Allistarj for 206.  - OH, 2.594 delay
6/6 11:07:46.546  You crit Servant of Allistarj for 539.  - MH, 2.796 delay

4 swings in Flurry, 2 with each hand, all noticeably hasted. In this case the result is pretty consistent with Flurry lasting 3 autoswings; the buff was in fact queued for faiding after the OH hit at 43.437 and the last MH hit just got to tag along for the ride.

My best guess is that what you're looking at is the Flurry "stack" being both slow to renew to 3 after a crit mid-Flurry and slow to fade when it's supposed to. If you crit while at 1 charge remaining then the game takes a second of thinking time before getting you back up to 3 charges, so you get a free extension of your old Flurry. As a Flurry is expended the game takes a second of thinking time to actually fade it, giving another free extension. The two effects combine to make the tail end of any Flurry streak considerably longer than it has a right to.

A good way to discount the above possibility is find an isolated Flurry that lasts for 5-6 hits.

Lactose · 06/07/07, 4:36 AM

Indeed slant, I forgot about the fact that off-hand consumes main hand procs (and vice versa). In my defense, it was about 3 AM when I wrote it :P
The thing about Flurry sometimes lasting a swing too much though, l think can be credited to delay in applying / removing the Flurry buff.
For example, you can see instanced when the Flurry gain is almost a full second after a crit, and an "extra" flurried hit being ~0.3 seconds after when Flurry should have disappeared.
At the very least, I don't think this delay should be neglected.

Universal · 06/07/07, 5:16 PM

Originally Posted by slant

You may have parried before that 1.25s swing, too.

This is something I haven't heard of before. Does a parry alter your swing timer in some way? I looked through the log and couldn't find anything consistent. If there is some effect, do some of the irregularities vanish if we assume such swings don't consume a flurry charge?

thorin5 · 06/07/07, 5:41 PM

Actually, if the attack hasted from the parry doesn't, for some reason, consume a flurry charge then it would explain that specific instance where it appears to be 4 attacks hasted by flurry.

slant · 06/07/07, 6:17 PM

Yes, after a parry your next swing is hasted by 40%. No clue if this stacks with or consumes flurry, although everybody assumes it does both.

It's not just lasting one swing too many; Disquette had six flurried attacks after a single proc. There's definitely something else going on.

I still tend to think my explanation is the most likely due to occam's razor; each hand procs separate buffs, simple.

Lactose · 06/07/07, 6:20 PM

I agree with most of it, excepting

If an attack procs flurry, it doesn't consume a charge.

I believe it resets that hand's stack, instead of just not consuming a charge.
Any logs confirming / disproving this?

Wandre · 06/07/07, 6:24 PM

Perhaps if a swing cooldown starts while your flurry buff is active but before the last charge is consumed it will continue at it's "hasted" rate.

This would prove more likely as this happens with spellcasting and slow. If I cast a spell while slow is up but slow then dissipates before the spell is over I still retain the slowed casting time.

That would explain an extra attack from an opposit hand. I'll put good money on the fact that you'll never see 2 swings from the same hand connect at a faster speed when you see 4 hasted hits. Maybe using 2 same speed weapons would help prove this theory.

Edit: This would also mean that you will always score 3 attacks when using only 1 weapon but generally score 4 attacks at hasted speed with dual wield as long as your weapons start a swing cooldown before the last charge is used.

Disquette · 06/07/07, 6:29 PM

Originally Posted by Wandre

Perhaps if a swing cooldown starts while your flurry buff is active but before the last charge is consumed it will continue at it's "hasted" rate.

I like this theory.

Slant, I'm just having trouble with reconciling what you saw (the strings of six, as I highlighted) when the single-hit versions only show 4 hasted attacks. I'm going to get more data after lowering my crit rate. There's too much noise in the data currently, but with a 5% crit rate or so it should be easier to identify different scenarios clearly.

Wandre · 06/07/07, 6:40 PM

If this is true then a swing cooldown in progress would never be altered, ergo the first swing to be hasted would be any swing cooldown that starts after the crit lands.

So yellow attack crits would normally proc flurry and the white attack that follows would have to connect once before it's hasted cooldown began. This could actually reduce the effectiveness of Flurry if you crit yellow attacks right after a white attack fired. That would probably be the cases of 3 hasted hits rather than 4 when using 2 weps.

White hit (start normal cooldown)
Instant Yellow crit (Flurry 3 charges)
White hit (start hasted cooldown and eat charge)

That's an example of what would happen if you yellow crit with an attack. Heroic strike would actually be a better attack because it uses the swing timer and you'd achieve the 4 hasted swings when DWing. Attacks like Stormstrike and Hamstring which fall outside the swing based attacks would eat charges prematurely.

slant · 06/07/07, 7:35 PM

Wait, you show 4 hasted attacks with only one weapon equipped? So much for my theory, then. I guess it was just client/server latency.

Universal · 06/08/07, 12:38 AM

I am pretty sure that slant's 2 procs theory isn't the explanation. For one thing, the example explanation shows OH crits refreshing the MH counter. I tried 15ish rules for how to decrement from the different procs, but none of them were consistent with both the long sequences Disquette originally posted. Furthermore, consider this sequence:

6/6 11:10:10.890  You crit Servant of Allistarj for 512.
6/6 11:10:11.171  You crit Servant of Allistarj for 202.
6/6 11:10:11.671  You gain 18 health from Healing Stream Totem VI's Healing Stream.
6/6 11:10:11.968  Servant of Allistarj misses you.
6/6 11:10:12.781  You hit Servant of Allistarj for 268.
6/6 11:10:13.296  You hit Servant of Allistarj for 102.
6/6 11:10:13.765  You gain 18 health from Healing Stream Totem VI's Healing Stream.
6/6 11:10:13.968  Servant of Allistarj hits you for 81.
6/6 11:10:14.890  You hit Servant of Allistarj for 279.
6/6 11:10:15.703  You hit Servant of Allistarj for 109.
6/6 11:10:15.703  You gain 18 health from Healing Stream Totem VI's Healing Stream.
6/6 11:10:15.984  Flurry fades from you.
6/6 11:10:16.000  Servant of Allistarj attacks. You dodge.
6/6 11:10:17.296  You hit Servant of Allistarj for 269.
6/6 11:10:17.687  You gain 18 health from Healing Stream Totem VI's Healing Stream.
6/6 11:10:18.093  Servant of Allistarj hits you for 84.
6/6 11:10:18.406  You hit Servant of Allistarj for 106.

A pair of crits generates only 4 procs and then fades well before any further attacks.

I also believe that the various lag based theories are incorrect. Look at Disquette's two original examples. Even if we allow that lag may allow some extra flurried hits (which seems plausible), the flurry buff itself is fading far too late. The buff fades 3 full seconds after the last hit which "should" have been flurried. Furthmore, at least in the data we have, this only happens after flurry chains. Unless chain flurries are themselves creating some sort of processing lag, it seems unlikely that no one would have noticed regular 3 second lag in other contexts before.

I don't have any answer myself. My wild guess was going to be some sort of variable overwriting but where 2 swings only use one charge if they hit close enough together. But this doesn't mix with only chain flurries going all the way to six. I do have one observation that may be useful though:

6/6 11:09:42.312  You hit Servant of Allistarj for 281.
6/6 11:09:42.687  You hit Servant of Allistarj for 107.
6/6 11:09:43.109  You cast Healing Stream Totem.
6/6 11:09:43.531  Flurry fades from you.
6/6 11:09:43.921  Servant of Allistarj hits you for 103.
6/6 11:09:44.609  You crit Servant of Allistarj for 580.
6/6 11:09:45.437  You hit Servant of Allistarj for 103.
6/6 11:09:45.609  You gain 18 health from Healing Stream Totem VI's Healing Stream.
6/6 11:09:45.812  You gain Flurry.
6/6 11:09:45.828  Servant of Allistarj hits you for 96.
6/6 11:09:46.828  You hit Servant of Allistarj for 271.
6/6 11:09:47.531  You hit Servant of Allistarj for 111.
6/6 11:09:47.734  You gain 18 health from Healing Stream Totem VI's Healing Stream.
6/6 11:09:47.953  Servant of Allistarj misses you.
6/6 11:09:48.562  Flurry fades from you.

This is an example with only 2 flurried swings. Note that in this case there is actually another hit before the flurry buff appears. So it seems like whatever mechanism is used, this hit is charged to flurry even though it happened before flurry occurred! This would be consistent with some sort of queueing mechanism. If, for example, it queues a ProcCritEffects() for the crit followed by a ProcHitEffects() for the hit, then this outcome would make sense because by the time the hit effects are examined flurry will have procced, even though the hit occurred before flurry procced.