Rezarel

This is not a change in game mechanics. All tests done so far on both the PTR and Live have shown the same behavior. However, it appears some very old enchants do benefit from haste, which is probably why that behavior was accepted as common knowledge.

--------

I took a large quantity of data on the 2.2 PTR against blasted lands mobs -- approximately 150,000 swings -- and it appears that while Flurried (25% haste) the proc chance for various events drops.

I tested three different procs, with three different weapon sets, with both 0/5 Flurry and 5/5 Flurry in most cases.

Procs Tested:
5/5 Unbridled Wrath
[Mark of Conquest]
[Darkmoon Card: Heroism].

Weapons Tested:
MH: [Protector's Sword]
OH: [Protector's Sword]

2x[Ced's Carver]

2x[Blade of Unquenched Thirst]

Various people were beating on my mobs with the occasional kidney shot, etc. Also, the level of the mob varied. Neither of these should affect proc rates.


Analysis:
I parsed the logs by tracking how many weapon swings and procs occurred inside and outside of "You gain Flurry" and "Flurry fades from you" messages. I realize that the combat log positioning of the haste gain/loss events isn't perfectly accurate, but it's the simplest way to separate hasted from unhasted swings. I then calculated PPM rates assuming the proc chance was based on unhasted weapon speed:

Calculated PPM = proc chance for an individual swing * unhasted weapon swings per minute

As you can see in the graphs, this calculated PPM rate drops while hasted, implying that the proc chance drops while hasted.

Looking further at my data, the real PPM (total procs / time in combat) was nearly equal for both hasted and unhasted tests. This suggests haste effects will not get you more procs for items which use a PPM mechanic.


Thread Summary:
Warriors see decreased proc chances for Unbridled Wrath when flurried on both the PTR (post 1) and Live (post 78).
Rogues do not see increased uptime on an unspecified proc buff while using Slice and Dice on the PTR - Tyran, post 33. Short testing though - only 15 minutes per run.
Rogues see an unchanged proc chance on Fiery Weapon while using Slice and Dice on Live - Baldycenarius, post 37
The behavior for Shaman with Flurry and Mongoose is the same on Live and the PTR - Lujaar, post 42
Hunters see a reduced proc chance on Live for [Don Santos' Famous Hunting Rifle] when using a quiver and Improved Aspect of the Hawk procs - Kaber, post 72
Warriors see an unchanged proc chance on Lifestealing when flurried on the PTR (post 113)
Instants use unhasted weapon speed to determine proc chance - Sarlunas, post 120

Thread Summary Summary:
Fiery Weapon (tested on Live), Lifestealing, and Crusader (tested on the PTR) all see unchanged proc chances when hasted - i.e., they benefit from haste. All other procs tested so far on either Live or the PTR have a reduced proc chance while hasted to maintain a constant PPM.

Instant attacks use the unhasted weapon speed to determine proc chance.

Also, see post 121 for a more detailed thread summary.



Here's the data (error bars are one standard deviation):

 

• Intermission

So are you saying...

when attack speed increases due to haste procs, the calculation done to determine the %-on-hit from the PPM amount is recalculated to keep a constant true PPM, based on *current* weapon speed, not its original stat weapon speed? Or does the PPM actually go DOWN, over compensating, in an attempt to stop non-linear results from multiple haste procs/effects?

 

◊ Kalman

If an item has an internal cooldown, that's going to provide some difference in PPM between hasted and non-hasted results, since you'll have more swings during the cooldown (swings which are guaranteed not to proc).

http://mmorchive.net

The WoW forums, explained:

 

• Intermission

I think he got around that by tracking non-haste procs (Mark of Conquest) yet having a separate haste effect going on and off in the background (Flurry).

 

slant

To properly test without relying on flurry, use a rogue singlewielding a weapon with a high PPM enchant like fiery and the setup talent. Unhasted it should be 6ppm, with SnD up 7.8ppm. That's a big difference that should be definitively testable in just a couple of minutes.

Or if you have access to high-end gear, just equip some passive haste. That would allow for huge sample sizes.

 

Rezarel

I'm not sure if Mark of Conquest or Heroism have an internal cooldown, but I've gotten two Unbridled Wrath procs with the same timestamp, so I'm assuming that has no internal cooldown.

Edit: I found two simultaneous Mark of Conquest procs, and two Heroism procs separated by 0.219 seconds, so I doubt either of those items have an internal cooldown.

 

• Intermission

Can you post a conclusion?

From what I can see...


- You tested PPM of procs.
- Recorded PPM while non-hasted.
- Recorded PPM while hasted.
- We know (do we?) that PPM mechanics take the weapon speed of your weapon, and use that to apply a %-chance-on-hit to your attacks, white and yellow. EG a hunter with a 1 PPM trinket will get far more than 1 PPM due to specials. His weapon speed is 2.9, so it creates a % chance based on that. But in reality, he shoots specials between every auto, and each of this shots has a chance to proc too, based on the original speed.
- But, your results show that the PPM is lowered while hasted, when logic would think that it increases, using my above example. Same weapon, same %proc chance per hit, more hits = more procs in same timespan = higher PPM. But its not higher, its lower.

What is going on, Rezaral? Or are you asking us?

Write a conclusion!

 

Rezarel

Sorry, my conclusion was in the first line: Haste effects reduce the proc chance of items that operate on a proc per minute mechanic. This was contrary to my expectations, and I've seen several people on this board say haste effects don't affect proc chances, so I thought I should post my results here.

I guess my plots are a little confusing: basically, I created the graphs assumimg haste effects didn't affect the proc chance, observed a contradiction, and threw out my assumption.

The real procs per minute (total procs divided by minutes in combat) was very close to equal in all my tests, both hasted and unhasted. The PPM shown in my plots is assuming the proc chance is based on unhasted weapon speed, i.e., reported PPM = proc chance * unhasted weapon swings per minute. Since the proc chance goes down while hasted, this reported PPM will be lower.

To be entirely clear, if you're single wielding and have an item with a 2 PPM proc on it, you'll get 2 procs per minute regardless of any haste effects you have on you (assuming no misses/dodges/etc).


I should also note that I didn't test any specials, so I have no idea how the proc chance for those is calculated.

 

◊ Lactose

This could probably by tested quite easily by a Hunter using something without an internal cooldown. Unequip quiver (15% constant haste), say hello to Dr.Boom (or Blasted Lands mob), fire away. Repeat with quiver active.
For a greater haste difference, quiver & Serpent's Swiftness (20% constant haste) respec would be even greater.
That should take care of Auto Attacks only.
Adding in a steady/auto macro, specials should be accounted for as well.

Anything without an internal cooldown that's easy to come by?

SNAKEBALL™

 

Rannasha

[Don Santos' Famous Hunting Rifle] pops to mind. Whether it fits the "easy to come by" is less clear. I have it on my Hunter at least, might have some spare time somewhere to test it. The problem with it is that it's a proc that can potentially self-refresh, making it impossible to accurately track with a logparse.

 

• Intermission

Oh ok, that makes a lot more sense. For a while there I thought you meant it actually lessened the PPM. But keeping it equal is good. I'm guessing it was coded that way to help balance haste.

Good to know about this as a fact, when making calculations about gear choices regarding haste/procs.

 

◊ Lactose

This post seems to back up PPM remaining constant regardless of hastes active.

SNAKEBALL™

 

sp00n

Did you just make a back reference to yourself?

Item Ranking Rogue [horribly outdated]

 

◊ Lactose

I would never!

Right now, both threads seem to be relevant to each other, hence some linkage going on.

SNAKEBALL™

 

sp00n

Uh, I see.
Mod Rewrite and setting postings per page to anything other than the standard value of 25 doesn't seem to work well together.

For me, with 30 postings per page, your link just goes to the last page where the first posting is exactly yours and the one linking back to this thread.

Item Ranking Rogue [horribly outdated]

 

• Disquette

Since this flies in the face of conventional wisdom (and I understood it), and has good testing behind it - I like it! My question would be how instant attacks are factored in. I assume those still increase the real PPM rate compared to auto-attacking.

----- sig ---------------
Discoepfeand - rogue / Disquette - shaman
A stormstrike / shocks /watershield timer-bar addon: http://www.curse.com/downloads/details/9729/
"Moogle has mentioned this in passing a few times but never elaborated on it. Perhaps we can entice him to respond." - Malan

 

sp00n

I guess so as well. Calculating the PPM is rather simple, since you have the weapon speed already at hand.
But for instants, you'd have to count how many additional attacks a player has made to base the PPM on this new number. And that sounds rather unlikely to me.
A question could be though if the PPM for instants is based on the real weapon speed, the normalized weapon speed or the current actual weapon speed including all current haste effects.

Item Ranking Rogue [horribly outdated]

 

Tick

Why was it contrary to your expectations? Intuitively, one would expect "proc per minute" to be independent of factors like number of attacks made within a minute or number of special attacks used within a minute. On the other hand, that wouldn't hold true (allegedly :P) for flat x% procs.

Still, it's good to know someone spent the time to confirm it.

 

suicuique

Because PPM is just a concept and all procs in game are handled as flat % procs.
It was common belief (AFAIK) that "PPM procs" were converted in a flat % proc based on weapon speed (unclear if base, normalized or current ... though "current" was thought to be the least probable of the 3).
This post however implies that for PPM to remain a constant value (no matter if hasted or otherwise), a steady reevaluation of the "flat % chance" has to be applied. Read: the flat % chance seems to be based on current speed. This would cover hasted affects.

Something like:
current proc chance = (PPM/number of base attacks per minute) / current haste factor

Note that the term in paranthesis would need to be computed only once (when you enchant your weapon or equip your trinket e.g.). This computed proc chance would need to be staedily uptdated with the "current haste factor".
Would this explain the results?

ADDENDUM:
Special Attacks (like instants) will always improve the effective PPM value, because to effectively keep the PPM value constant regardless of the use of these instants would require to upkeep a heuristic. This is very very very unlikely. And can be easily disproven (sic?) by comparing "auto attack only" proc values to "normal combat" proc values (e.g. against mobs in blasted lands): there is a siginificant delta to be observed very easily.

 

Gryzemuis

Did anyone repeat these tests ?
Are these numbers large enough to make the conclusion that haste reduces per-swing proc-chances ?
Are we sure that before 2.2 (ptr) haste did not reduce per-swing proc-chances ?

I am leveling my mining at the moment (and after that my blacksmithing) to make a Dragonstrike mace. If the proc-chance under haste gets reduced, that in combination with the overall haste nerf, might be enough to make me change my plans.

If this behaviour is indeed a change from previous behaviour, then this is a pretty big nerf for rogues again ! Assuming a rogue always has s&d up, uses blade flurry, and has 2x mongoose, this changes causes a 35-40% reduction on several procs. Less mongoose procs, less trinket procs (dst, hourglass), less meta-gem procs. If it is implemented consistently, it would cause even less poison procs ! Less windfury procs.

What would be the overall effect of a nerf like this ? 10-20% reduction in dps ?

I find it hard to believe that Blizzard would sneak in such a huge nerf. If the proc-chance reduction would apply to some trinkets and some abilities, but not to others, what would be the selection criteria ? How do we know which abilities have their proc-rates impacted by haste, and which do not ? I'm really worried here.

 

Tick

Yes it would, and there are plenty of inexpensive and accurate enough ways to do it in a game (the AI people have already solved this one too :P). Of course, these are only approximations which have to be fast enough for a game engine, and therefore subject to player "exploitation" such as using special attacks which will impact the effective PPM rate (just like you observed).

So, maybe I'm talking with hindsight when I say it's intuitive, when I've seen similar concepts implemented. My apologies, I didn't by any means intend to degrade Rezarel's work, or yours.

 

suicuique

It certainly is more intuitive and does justice to the PPM concept as a whole.
Agree.

No need to.
And a second the wish for some further testing.

Sadly I do not have many passive haste items. But, equipping some of these and NOT speccing into flurry should give us a hint if haste begets procs in LIVE now.
And see if some delta in procs over time can be obeserved (haste items vs non haste items).