Proc Rate Determination

Machinator · 12/19/07, 3:49 AM

This thread is a product of the Execution/Mongoose thread to determine the proc rates of enchants and items and develop tools to do so. Currently we pretty much have assumed numbers, but the data doesnt always match and it would be good to nail these down.

Data Collection:
I think it is best if we set some standards for testing so we can combine the data for more accurate results. We want both fast and slow tests so we can tell the difference between PPM and % procs. I would suggest 1.4 and 2.6 speed, and maybe a 3.6 speed 2h, as they are most common. To start no passive haste or haste procs should be used as their effect on PPM enchants is in question.

Proc rate determination:
Im not sure what the best ways of doing this are yet without posting a whole bunch of stuff on statistics. Percent based procs are easy to understand. Proc Per Minute is the average procs you would expect per minute regardless of weapon speed. I have seen evidence, but not sure if it is conclusive, that for normal attacks(not special) the chance of a proc is adjusted by the current speed of the weapon under haste rather than the base speed.

Utilities:
It would be great if we could get a program to parse combat logs and output information like hits, procs, multiple procs, uptime, ect.

• Aldriana · 12/19/07, 4:15 AM

I'll work on assembling a program to do the parsing; I have most of the code in place, it's just a matter of grafting an appropriate UI onto it.

In terms of proc rate determination: for those of you that wish to take data, the general rules of thumb are as follows:
1) unequip all items that proc haste effects - Thundering Skyfire Diamond, Mongoose, Dragonspine Trophy, etc. - with the (possible) exception of the proc you're testing.
2) only equip a single weapon; while two-weapon testing is sometime useful, it is only circumstantially possible; in general, testing with one weapon is safer.
3) If you have haste rating items equipped, record how much and report this value with the data.

For information on the interaction between haste and PPM effects, I would refer you to the following thread: Proc Per Minute Mechanics & Haste
The information found there, while not 100% conclusive, is reasonably convincing that PPM effects on noninstant attacks is based on current weapon speed, and on instant attacks it is based on base weapon speed.

In terms of weapon speed: many rogue will find either 2.6 or 2.7 to be convenient to test, and either 1.4 or 1.5 as well; relatively fewer will find both to be convenient. Warriors/shaman, I can't speak for. However, we don't need to standardize across all tests; so long as data for any particular proc is collected for the same speeds, we can change between procs. So, for instance, Mongoose testing has started at 1.4 and 2.7; but we could still do, say, 2.6 and 1.5 for Executioner.

Additionally, some procs were pinned down in some detail in the Mechanics Testing Thread, which I can't find anymore but do remember a few results for:

1) Rod of the Sun King is ~4.5% proc rate, hence hypothesized as 1 PPM. Would need more testing to determine if it's actually 4.5% or if it's 1 PPM, as they perform differently with haste.
2) Dragonspine Trophy is 1 PPM with a 20 second cooldown
3) Tsunami Talisman is 10% on crits with a 45 second hidden cooldown.
4) Band of the Eternal Champion is 1 PPM with a 60 second hidden cooldown.
5) Madness of the Betrayer is 1 PPM with no internal cooldown.
6) Warp Spring Coil is 25% on specials, with a 30 second internal cooldown.

Stuff that I don't remember seeing detailed tests on, and/or forget what the answer was, that are of general interest to endgame melee DPS:

1) Mongoose
2) Executioner
3) Siphon
4) Blacksmithing Maces
5) Deathmantle 4/5 Set Bonus (was narrowed down but I don't remember seeing a conclusive proof)

There are probably others, but those are the ones that immediately spring to mind.

koaschten · 12/19/07, 5:52 AM

Might be of interest to give a short walktrough on what mobs to test on and how to reach them.

hannigaholic · 12/19/07, 6:29 AM

For a rogue (and anybody else who can get there quickly with either stealth pots or invisibility pots if possible) I think the best mobs to test on are the Gordok Spirits in Dire Maul North. Just stealth to the King's room and unload for hours on end behind a Spirit. They don't turn around to face you so you're attacking from behind, as you would in a bossfight, and they never drop below 100% so you can keep going forever.

Machinator · 12/19/07, 8:48 AM

The gordok spirits are probably best. Level shouldnt matter since we want all attacks to land anyway. I think blasted lands servants are probably as close as you can get to them, you will parry, but I dont think it will affect the data.

If you can solo a DM north boss then you can get to the spirits behind the king. Invis on its own wont help as you need to open the inner door somehow and there are mobs to kill. I did it in dps gear with a couple pots and a death to kromcrush's adds after I killed him. It can probably be improved upon for the better prepared people. Now that I think about it ogre suits might work too if you can get them.

The hard part is not going AFK while you test. I think you log after 15m of inactivity, so you need to press a button while you watch TV or something. I cant think of any way around that that would be legal. I load up the PTR and Live at the same time and run AV and test.

Also, so we can get it out of the way, does anyone have a whole lot of haste gear? I dont know how much you need to be statistically significant but somehow I would like to conclusively test if haste changes the proc rate of a PPM enchant or not.

koaschten · 12/19/07, 9:37 AM

Guess i can dig up crafted shoulders / bracers, 2 BT rings, Grips of Damnation in my bank, get shallow-grave trousers for badges and take a blade of infamy to Dire Maul. Should be 194 haste rating / 12.3% passive haste. I'll see if i can manage to get on the EU PTR over the holidays, last weekend when i tried i couldn't even get the EU PVE PTR to show up in the realm list, even after completly wiping and re-installing the wowtest folder.

Xoya · 12/19/07, 10:20 AM

I can guarantee that Syphon of the Nathrezim has no internal cooldown, at least none worth mentioning. I don't remember if I have seen it refresh itself but I know for a fact I've had it proc immediately after the buff ran out.

Malan · 12/19/07, 10:25 AM

Originally Posted by Aldriana

For information on the interaction between haste and PPM effects, I would refer you to the following thread: Proc Per Minute Mechanics & Haste
The information found there, while not 100% conclusive, is reasonably convincing that PPM effects on noninstant attacks is based on current weapon speed, and on instant attacks it is based on base weapon speed.

I could have sworn that the conclusion of that thread was that haste did not have an effect on PPM?

Astrylian · 12/19/07, 12:24 PM

One thing to add... Be careful about tests on those Gordok spirits, or really anything besides 70/71/72/73s. If you come up with a very different proc rate than expected, that could be why. At least in one case (Idol of Terror), it appears that the proc rate scales with target level. Most procs don't seem to work that way, but just thought I'd mention it, in case anyone's looking for an explination of why their proc rates seem high/low on the lowbie mobs.

• Aldriana · 12/19/07, 2:16 PM

Originally Posted by Malan

I could have sworn that the conclusion of that thread was that haste did not have an effect on PPM?

When people say things like "haste does not affect PPM", what they mean is that haste effects don't increase the number of observed procs; that is, absent other feedback mechanisms, if a rogue DPSs for an hour with no haste equipped he'll observe some number of procs; if he then equips 150 haste rating and DPSs for another hour, he will (statistically speaking) observe the same number of procs.

However, what this means for the proc rate on any given attack is that it's based on the current weapon speed rather than the base weapon speed; hence, the proc rate per attack changes with haste, whereas the proc rate per unit time does not.

I could be mistaken, but this is my understanding of what was determined.

Latito · 12/21/07, 3:57 AM

Went to DM, used a Searing Sunblade in my OH, no MH weapon. No haste so 1.3 speed attacks - 1.27 while mongoose'd. Only other proc was Hyjal Exalted ring, recheck proc rate on that if you want. After the sunblade broke I poped on Merc. Glad. Quickblade.. gear was the same otherwise. Since this was an offhand sword with swordspec, I was getting MH attacks occasionally - yes, I punched him a few times from SS procs. If you look at the dmg amounts, its always 185 w/o either proc up so its safe to say that its a MH punch, thus not able to get Mongoose. Just filter those attacks out, minor annoyance, sorry :P Arena partner logged on ~20% of the way into the Quickblade test, hence a much smaller data sample.

Oh.. this was all autoattack, no specials.

Also.. I saved the Grips of Damnation from sharding this week so I can put together 105 haste (belt, ring, gloves) if something needs testing with static haste. Not quite Daemona's 194.. but its something.

Free file hosting by Savefile.com
Free file hosting by Savefile.com

Borodin · 12/21/07, 7:29 AM

Just wondering if it might be worth waiting for the 2.4 (?) changes to the UI and specifically the Combat Log which might make things a lot easier to analyse.

If it's 3.0 that those changes arrive then obviously that's a long way off and I shall shuttup.

Malan · 12/21/07, 10:09 AM

No the log changes are scheduled for 2.4.

• Rezarel · 12/21/07, 10:36 AM

Originally Posted by Machinator

The hard part is not going AFK while you test. I think you log after 15m of inactivity, so you need to press a button while you watch TV or something. I cant think of any way around that that would be legal. I load up the PTR and Live at the same time and run AV and test.

When I took data from the Blasted Lands mobs, I wouldn't afk out until I left combat (when I died, or when someone else broke the crystal and killed my mob). I'm not sure if the Dire Maul mobs would actually put you in combat though.

Machinator · 12/28/07, 3:53 AM

I copied a rogue to the PTR to test in DM. The ogres do put you in combat and you dont seem to DC, just go afk if you dont do anything. Im not sure how the combat log reacts to disconnects though. I am gonna let one run overnight and see what I get.

The reason for finding someone to stack haste is to run one test without haste, and then again with lots of haste to confirm if you actually get more procs in the same time period. So if you can do 2 tests like that we can possibly settle that. I would assume Slice and Dice, Flurry, and hasted swings from parries work the same way, but thats much harder to keep up for an entire test.

Test Results:
I dug out my comp science class stuff and wrote a crappy program to look through each line and look for procs and determine how long they lasted. I assumed that any proc longer than 15s was 2 procs and longer than 30s was 3 procs. I learned that because of how the combat log is affected by lag you need to give some extra time for procs. The first time I ran through I had as much double procs as singles because the time between a gain and fade was something like 15.177 seconds. So I counted singles as less than 15.250s and double as less than 30.500s. Hopefully I remember to upload the log later.

Results before weapon broke:
Hits: 5135 (2.6 speed weapon, no haste)
Time: 222.517 minutes
0-15s: 132 gains
15-30s: 53 gains
30+s: 9 gains
Total Procs: 265
PPM: ~1.19 PPM

Since this is longer than my other tests I would assume more accurate. I will keep running these while I am at work/sleeping. Anyone good with stats that can give a number how many hits or minutes it would take to be say 99% or 99.99% accurate?

EDIT: My test was for Executioner.