EnhSim, DPS simulator - Page 7

Rhaegal · 09/15/08, 12:10 PM

It's not a question of the math. The statistics are obvious, but that's not the way it's being presented. If it was simple statistics, it never would have been a major topic of interest. Read these two sentences, Malan:

When you dual wield weapons, the chance on each landed attack outside the 3-second cooldown is approximatedly 36%, if and only if both weapons are imbued with Windfury.

When you remove the ineligible hits the observed proc rate from the eligible hits becomes 36%.

Those are directly from the TTT, and are in direct contradiction to what you're saying here today. The TTT says that each hit has a 36% chance to proc WF, and you're saying that every two hits have a 36% chance to proc WF. Those are very different statements.

Levva · 09/15/08, 1:14 PM

Malan is correct, and strangely you are correct too Rhaegal but both expressing it in different ways

It is very very simple from a programming point of view. When a weapon MH or OH lands a hit do the following psuedo code check

if weaponHasWF_embue AND not WFProcced then
   if rnd()<0.2 then 
      -- WF procced on weapon
      WFprocced = true
      StartWFCooldown()
      DoWFDamageCalcs()
   end
end

ie: its a FLAT 20% check to see if WF has procced but you MUST make sure the 3sec CD is not active. Doing an INDEPENDENT check for each weapon at 20% mathematically means that "outside the 3-second cooldown" (the key bit in the quoted text) the chance is 36% EXACTLY as Malan quoted.

The critical part is that you ONLY check a WF outside the CD. This is why off-hands can "steal" a WF proc, as its a single check per hit, if WF is available then there is a 20% chance of a proc for the OH.

If you run the sim with this calc you should then see WF proc rates coming out at 36% OF ELIGIBLE HITS as predicted. However only 20% of ALL HITS.

Pitbuller · 09/15/08, 1:20 PM

Levva: Statistic still say 36%. Go and check wws.

I think they have coded it "wrong" and when you have double wf game check wf proc twice per hit. And that way give 36% proc chance.

Levva · 09/15/08, 1:23 PM

It is incredibly confusing the various wording. I'm reasonably confident the code I wrote is how it works however I am very open to being proved wrong as this is a critical part of the sim and NEEDS to be correct.

Since the article refers to calcs originally done by Disquette and he has popped his head in this thread in the last couple of days could he give any indication of what he found (or repost original thread)?

Levva · 09/15/08, 1:32 PM

Originally Posted by tukez

(mwsAverage - 1.96*mwsSD) to (mwsAverage + 1.96*mwsSD)

be this

(mwsAverage - 0.96*mwsSD) to (mwsAverage + 0.96*mwsSD)

No its most definitely 1.96 * sample standard deviation to give a 95% confidence interval.

1 SD gives only about a 68.27% confidence interval (0.96 would be less)
2 SD gives 95.450% confidence

It is normal however to use 1.96 SDs to give a 95% confidence interval. See Standard deviation - Wikipedia, the free encyclopedia

The large variance you are seeing just shows how random the time taken is with low stacks.

However if we are seeing that taking less than 5/5 MW gives markedly less dps then the average time to achieve 5 stacks is not perhaps not a worthwhile statistic to display after all.

Disquette · 09/15/08, 1:51 PM

Hello all.

First, the best way to settle any sort of argument like this is by testing for one's self

That having been said, when I did my tests, my conclusions are:

1) Once you windfury, you can't windfury again for 3 seconds.

2) After those 3 seconds, any given attack has a 36% chance when dual-wielding (approx, the math behind it makes 36% look like a very attractive candidate) to proc another windfury.

If you're willing to spend a bit of time in blasted lands, anyone can verify or disprove this.

Malan · 09/15/08, 1:52 PM

Originally Posted by Pitbuller

Levva: Statistic still say 36%. Go and check wws.

I think they have coded it "wrong" and when you have double wf game check wf proc twice per hit. And that way give 36% proc chance.

I think you are calculating it wrong Pitbuller.

I have a BT parse from a weeks ago vs Brutallus.
267 Windfury "hits" (non crit), 105 Windfury critical hits, 4 windfuries parried
Total Windfury attacks: 376

However, WWS counts each windfury hit as unique, ie, the 2 instant attacks are each counted as a unique incident. Therefore, divide 376 by 2.
Total number of actual windfury procs is therefore 188.

On the same parse I had 631 "hits" and 308 crits for a total of 939 swings that connected.

188/939 = 0.20021 or ~20%

So my overall observed proc rate from my *total* number of hits is 20%. Which is what I said above. What you are saying is that if you make some approximations for number of hits outside the cooldown that you see 36%. Which is exactly what I said as well. Outside of the cooldown there is a 36% chance of a windfury. When taken into account for the total data set though its still a 20% effect.

Roywyn · 09/15/08, 2:38 PM

There has been a test recently that checked that when you perform a Stormstrike, your MH checks first for a WF proc, then your OH checks for a WF proc. (Hence you get more PH procs.)

The test proved "en passant" that the WF proc rate was 36% on each unless I'm completele mistaken.

The test was done by standing sideways to a Blasted Land servant so you could Stormstrike and get Windfury hit, but not get white hits.

That positioning still works on beta, so you can easily do the test there/on the PTR.
(If Windfury works that is.)

Xoya · 09/15/08, 3:07 PM

Originally Posted by Malan

The actual observed proc rate will be 20% if you examined a combat log, but while Dual Wielding each eligible hit from both hands (ie, those hits made outside the WF cooldown) have a combined 36% chance of one of the two weapons producing a proc. Each hand still has a 20% chance to proc when viewed individually though.

Originally Posted by Malan

Yah you're doing it wrong. Each hand still has a 20% chance to proc, its just that when you view the system as a whole the proc chance is 36% for any hand to proc. For your sim each hand should have a 20% chance to proc (but locks out the other hand obviously).

Like I said, its just a statistical thing. The 36% number is just saying "If you have windfury on both hands, you have a better chance of getting a proc", but you'll still only see an overall proc rate of 20% of your total eligible hits.

Originally Posted by Malan

So my overall observed proc rate from my *total* number of hits is 20%. Which is what I said above. What you are saying is that if you make some approximations for number of hits outside the cooldown that you see 36%. Which is exactly what I said as well. Outside of the cooldown there is a 36% chance of a windfury. When taken into account for the total data set though its still a 20% effect.

Underlines mine.

You really didn't say "exactly" those things. You said that each individual hand has a 20% chance to hit even when outside the windfury cooldown and that only as a combination of two hits do we see a 36% proc chance. Disquette has just told us that each individual hand has a 36% chance to hit when outside the windfury cooldown. Because of the 3-second cooldown, the observed proc rate when considering all hits is 20%.

Again, to reiterate, you said that "eligible hits" have a 20% chance to proc windfury. Disquette did not say that. If "eligible hits" had only a 20% chance to proc windfury, because of the 3-second internal cooldown we would see an observed proc rate on all hits of much less than 20%.

I know that you're a smart guy, Malan, but you contradicted yourself, and the vitriol and self-righteousness is getting really tiresome. You were wrong, other people were interpreting the TTT correctly and you were not, and If I get warned for this, whatever.

Rouncer · 09/15/08, 3:40 PM

Originally Posted by Roywyn

There has been a test recently that checked that when you perform a Stormstrike, your MH checks first for a WF proc, then your OH checks for a WF proc. (Hence you get more PH procs.)

The test proved "en passant" that the WF proc rate was 36% on each unless I'm completele mistaken.

The test was done by standing sideways to a Blasted Land servant so you could Stormstrike and get Windfury hit, but not get white hits.

That positioning still works on beta, so you can easily do the test there/on the PTR.
(If Windfury works that is.)

I've been spending a good deal of my free time this morning looking for that testing in the old enhancement thread (no luck so far) but that was what I remembered about it as well.

Basically Blizzard half-assed the coding for Windfury and any strike when dual wielding (wf/wf) would get the 36% chance to proc when off cooldown. Which integrates very well with Stormstrike since Stormstrike checks for a Mainhand proc before it checks for an offhand proc.

I'll keep looking but that's what I remember being the final outcome from ingame testing regarding it.

fauxpas · 09/15/08, 4:32 PM

Heh, the last two pages were extremely funny. We've had dualwielding + 3 Sec WF Cooldown for how long now? 1,5years?

And it's still not obvious how it works, apparently.

I'll try to find a wording that makes it crystal clear.

When you are outside the Windfury cooldown and attack with both weapons simultaneously, is it:

a) P(WFProc) = 0,36 + ((1-0,36)*0,36) = 0,5904

OR

b) P(WFProc) = 0,2 + ((1-0,2)*0,2) = 0,36

Malan · 09/15/08, 5:20 PM

Originally Posted by Xoya

other people were interpreting the TTT correctly and you were not

That's pretty funny considering.

tukez · 09/15/08, 6:26 PM

Hey questions about glyphs. Can we use same type of glyph in multiple slots and do they stack? Can we use minor glyphs in major slot? Also, if you use Glyph of Flametongue Weapon and dual FT, do you get 4% spell crit? Is the Glyph of Stormstrike bonus damage additive or multiplicative to the existing 20%.

tukez · 09/15/08, 7:13 PM

Now that they change the armor ignore to armor penetration rating, does Executioner buff change to rating also? I guess static bonuses on items change to rating at least, what about trinket procs like Madness of the Betrayer? AFAIK, debuffs like Sunder Armor or Faerie Fire won't be converted to rating.

Morelis · 09/15/08, 7:36 PM

Originally Posted by Roywyn

There has been a test recently that checked that when you perform a Stormstrike, your MH checks first for a WF proc, then your OH checks for a WF proc. (Hence you get more PH procs.)

The test proved "en passant" that the WF proc rate was 36% on each unless I'm completele mistaken.

The test was done by standing sideways to a Blasted Land servant so you could Stormstrike and get Windfury hit, but not get white hits.

That positioning still works on beta, so you can easily do the test there/on the PTR.
(If Windfury works that is.)

I was the one that did that testing, and it was posted in the think tank thread but it appears to have been purged at some point... I had done it to determine if the main hand is checked first on SS for WF procs(which it is) but the results also show the proc rate per weapon, not combined is indeed 36%.

Here's the WWS from the log I modified to show the separate procs, the only changes made to it were to descriptions and spell IDs to allow the WWS parser to display them separately:

Wow Web Stats

It's a bit annoying to see good testing thrown out.

Rouncer · 09/15/08, 10:50 PM

Originally Posted by Morelis

per weapon, not combined is indeed 36%..

THANK YOU!

I have been reading the old Enhancement thread page by page all day looking for that proof. I knew I wasn't losing it and that Windfury was 36% per weapon when off cooldown.

The numbers are basically perfect for what would be predicted by 36% chance per weapon.

726 Stormstrikes.

36% would be 261 mainhand windfuries (he had 249 or 34%)

That would leave 726-261 = 465 attacks that would get checked for offhand windfury (477 remaining in his test pool)

36% of that would be 167 offhand windfuries (he had 157 or 33%).

That basically fits although it does open the door that the percentage chance per weapon is slightly lower then the 36%, it actually may be 33-34% or it could just be that it is a small(ish) sample size and over a larger sample size it would show the 36% that was expected.

Very glad you saved that parse, and thanks again for posting it.

Morelis · 09/16/08, 12:12 AM

You left out crits.

546 SS, 194 MH WF = 35.5%
(546 - 194) SS, 120 OH WF = 34%

A little closer to 36% but what we would really need to be sure is a larger sample. It took over an hour and a half to get those 546 stormstrikes... Someone else can get more.

Chromean · 09/16/08, 3:51 AM

Originally Posted by Rounced

THANK YOU!

I have been reading the old Enhancement thread page by page all day looking for that proof. I knew I wasn't losing it and that Windfury was 36% per weapon when off cooldown.

The numbers are basically perfect for what would be predicted by 36% chance per weapon.

726 Stormstrikes.

36% would be 261 mainhand windfuries (he had 249 or 34%)

That would leave 726-261 = 465 attacks that would get checked for offhand windfury (477 remaining in his test pool)

36% of that would be 167 offhand windfuries (he had 157 or 33%).

That basically fits although it does open the door that the percentage chance per weapon is slightly lower then the 36%, it actually may be 33-34% or it could just be that it is a small(ish) sample size and over a larger sample size it would show the 36% that was expected.

Very glad you saved that parse, and thanks again for posting it.

Correct me if I'm wrong but doesn't WWS count windfury as two separate hits? So if you calculate the procs based on number of connected WF hits the results will show a higher percentage than it really is. I think it came up earlier in this thread already that you need to halve the number of connected WF hits and then calculate the percentage to get an accurate result . What my quick napkin math got out of that parse was ~17% for MH and ~13% for OH windfuries, it is a very small sample though you cannot expect absolute results as was pointed out already.

Morelis · 09/16/08, 4:53 AM

Originally Posted by Chromean

Correct me if I'm wrong but doesn't WWS count windfury as two separate hits? So if you calculate the procs based on number of connected WF hits the results will show a higher percentage than it really is. I think it came up earlier in this thread already that you need to halve the number of connected WF hits and then calculate the percentage to get an accurate result . What my quick napkin math got out of that parse was ~17% for MH and ~13% for OH windfuries, it is a very small sample though you cannot expect absolute results as was pointed out already.

That would probably be the proper way to do it, but since he didn't halve the stormstrikes(2 per hit as well) to begin with it ended up working out anyway. The numbers I listed were the halved values. As to the sample size I'm sure one of the statistics nuts around here can tell us exactly how accurate it is.

Raut · 09/16/08, 5:35 AM

Jesus fuck. This thread sure took a turn towards instant failboat all of a sudden.

1. The dataset for the "proof" that WF is so and so is low. Very low.

2. WWS is retarded and shows "hits" without crits so you have to add the crit nuber to hits to get the actual number of attacks.

3. WWS shows one WF proc as two WF attacks, like stated above.

Actual numbers:

      Hits  Crits  Total
SS   : 726 - 366 = 1092
MH WF: 249 - 139 = 388
OH WF: 157 -  83 = 240

So proc chance in this run was for MH: ((388/2)/1092)*100=17.77% and ((240/2)/1092)*100=10.99%.

The above numbers show you a few things. First you see why a very small dataset is bad(but you have to take my word for this until I can dig up my old statistics books to prove it, or someone else can), secondly you see how OH is locked out of WF by MH thus getting a very low proc chance.

Any chance you can put the actual log somewhere, Morelis so we don't have to put our faith in RETARDED WWS?

Edit: And just to let you know: Some of this is pieced together from guesswork on how WWS works so don't take my word for this as the end result. Need the log to be certain.

Pitbuller · 09/16/08, 5:51 AM

Raut: You divide with all SS attacks. You should use only mh SS attacks.
MH: ((388/2)/546)*100=35.53%
[(240/2)/ (1092/2 - (388/2) )]*100=34.09%

WWS parse with over 500 mh and oh SS attacks is big enough for checking 20% vs 36%. And it's closer to 36%

Levva · 09/16/08, 6:00 AM

Tukez could you settle the argument in the sim pls.

If you could record in the sim the number of hits (including crits), the number of WF procs, and the number of hits inside the WF cooldown. Then display these as percentages at the end of the run. We can then run the sim a lot and check.

If the output can show :
Total hits, Total WF procs, hits that were INSIDE the WF CD. We get WF procs/Total hits = Total % chance to proc AND WF Procs/(total hits - hits inside the WF cooldown) = total % chance to proc WF OUTSIDE the WF CD.

Now these figures will give us a 5000 hour sim of numbers of attacks at the current sim rate of 36% ON EACH WEAPON HIT. ie: I understand that every time you check a hit to see if it procs WF you use 36% chance.

You could then alter the sim and change the chance on each weapon hit proccing WF to 20% (which I believe it should be) and see what the rates are for inside and outside CD.

The sim can then show unbiased figures of what the proc rate ends up over a very very large sample, both inside and outside the CD.

Acutally it might be nice in the sim to have a "debug" tick box. So that normal runs didn't show these sorts of figures, and things like the average MW 5 stack times etc. Then if user turns on debug info they get the extra calcs. That way the sim runs full speed without debug info for most occasions and with the extra sometimes useful info when needed simply by ticking a box.

Edit: It might be nice to record MH hits, MH WF procs, MH hits inside WF CD separate from OH hits, OH WF procs, OH hits inside WF CD.

Levva · 09/16/08, 6:13 AM

Originally Posted by tukez

Hey questions about glyphs. Can we use same type of glyph in multiple slots and do they stack? Can we use minor glyphs in major slot? Also, if you use Glyph of Flametongue Weapon and dual FT, do you get 4% spell crit? Is the Glyph of Stormstrike bonus damage additive or multiplicative to the existing 20%.

The glyph interface has 6 round circles, 7 if you are an inscription profession guy. 3 of the circles (4 with inscription) are Major Glyphs, and 3 are minor. On using a glyph you fill up one of the slots. There is nothing on the Glyph (see Shaman Glyphs - Items - World of Warcraft) that says unique or unique equipped, so you might be able to get 3 of the same. I'll see if the beta is stable enough to try it this evening, although will need to find inscription person or level it :-(

The damage bonuses are almost certainly multiplicative as Blizzard changed a lot of effects, with patch 2.0 or 2.1 if memory serves, to be multiplicative and not additive as too many things were becoming possible to exceed or get close to 100% again would need proper testing. I can't see you having 3 Windfury Weapon glyphs and having +120% AP bonus or could it???

Although now the PTRs are actually up it should be possible to test it on there too if you can get through the queues.

Raut · 09/16/08, 6:16 AM

I knew I made a mistake... Sorry. At least it refreshes my hatred for WWS. (And my assumptions)

Why do you deduct MH WF procs from OH WF proc calculations? The argument stands as a whole and not only in cases where MH hasn't procced WF.
Something akin to (((388/2)+(240/2))/546)*100 = 57.51% of SS attacks procced WF.

We just have to disagree on the size of the dataset.

Philondra · 09/16/08, 6:21 AM

Originally Posted by Raut

I knew I made a mistake... Sorry. At least it refreshes my hatred for WWS. (And my assumptions)

Why do you deduct MH WF procs from OH WF proc calculations? The argument stands as a whole and not only in cases where MH hasn't procced WF.
Something akin to (((388/2)+(240/2))/546)*100 = 57.51% of SS attacks procced WF.

We just have to disagree on the size of the dataset.

As Rounced reminded us several pages back, SS will always attempt to proc from the MH first and, failing that, attempt to proc from the OH. Therefore, if SS procs an MH WF, the OH will never proc. We need to take this into account or else the OH proc rate will appear to be artificially low.

Edit to add: The purpose of this sim was NOT to find out what chance stormstrike has to proc WF. It was to find out what the chance is to proc WF while dual wielding if WF is currently not on cooldown. Stormstrike is merely the vehicle chosen to help ensure that WF is always off cooldown when attacking; if you just sit there and include autoattacking, latency and combat log lag will make it difficult to analyze results and determine the actual proc rate. Controlled tests can be done by turning your character model in such a way that specials will hit but your character will not autoattack.)

We then must compensate for the fact that SS attacks with both weapons, which we do by exluding the OH hit in cases where the MH procced WF, as the OH will have an observed proc rate of 0% in such a situation.