Theck's MATLAB TPS analysis - WotLK/3.x

[Iselian]

So much better than my pre-cal class. Fantastic as always, Theck!

[theckhd]

A few quick notes:

• Marvel pointed out that I made a slight error in the Mongoose uptime calculation.
  

  I believe you slipped in a division symbol in the formula. It should read: (0.9425)*(1/1.26+1/6)+(0.96)/9 = 1.012 successful attacks per second that can proc the buff.
  
  Thus the uptime is: 1 - (0.9733^(1.012*15)) = 33.689%

  
  I've gone back and updated the calculation to properly reflect this. I've also added the code to do this calculation to the enchant calculation, so that it will automatically recalculate this as gear changes.
  
  Also note that the uptime depends fairly heavily on hit and expertise - it decreases roughly linearly with miss/dodge/parry chance from 56% uptime at 0% miss/dodge/parry (i.e. hit-capped and exp hard-capped) to 40% at 28.5% miss/dodge/parry (i.e. 0 hit rating and 0 expertise).
  
  
• I noticed that in the default gear set, I've mistakenly made Crusader's Glory a 1.6-speed weapon (it's actually 1.5). I've decided to leave it this way, mostly because I think 1.6-speed weapons are going to be far more common, and keeping it that way gives me more universally useful values for things like parry-hasted swing speeds and proc-based uptimes.
  
  Note that since I store only the weapon dps and base swing speed, this only results in a slight systematic error in threat values due to our scaling with swing speed. It shouldn't change the qualitative results at all, nor really have a large effect on the quantitative results (the difference is about 21 TPS).
  
  Also note that I did not make the same mistake in the Weapon Speed Analysis; that calculation uses the proper 1.5 swing speed for Crusader's Glory. So the graphs in that analysis are going to be accurate.
  
  
• I'll be updating the Guides section of the site later this week, starting with the MATLAB entry. I intend for that to be the "TLDR" guide version of this thread, with just the most relevant plots and advice for newcomers. If you have suggestions for how to improve the guide version (once it's updated), you can post them here.
  
  This thread will remain as it is, in all of its number-crunching glory, and I'll keep posting the lengthy version of each analysis here. The guide should just end up being a much abbreviated version that's much easier to navigate, since it's not buried in a N-page thread (for large values of N). I plan on keeping the ToC in the first post updated as well as the guide though, and the guide will have links back to the relevant "unabridged" version of the calculation for people who want more information.
  
  In other words, the guide is where you can point the new players who ask "Is X better than Y?" From there, they can follow the links back to the calculations themselves if they want to see the gritty detail.

[theckhd]

On the topic of mongoose/berserk PPM testing, I spent an hour on a dummy today. 30 minutes with SoV, 30 minutes with SoW, using a Dalaran Sword with Berserking. The combat logs are attached.

WWS parses for each:
SoV - 47% uptime, 76 applications.
SoW - 20% uptime, 25 applications.

This likely settles the "do HV applications proc ppm-based enchants" question.

[Angella]

It must be somewhere in this thread, but what does DUH and non-DUH indicate? Way too much text and its pretty late.

[Dragondin]

DUH - Cutesy acronym that stands for Demon, Undead, or Humanoid. Uused to refer to mobs or bosses which give Crusade the extra 3% damage. Note that Elementals are also DUH mobs, but the acronym isn't as cute if you have to fit an E in somewhere. An NPC which only gets the base 3% from Crusade is often called a "generic" or "no-DUH" mob.

the first post declares this in the glossary.

[Iselian]

On the topic of mongoose/berserk PPM testing, I spent an hour on a dummy today. 30 minutes with SoV, 30 minutes with SoW, using a Dalaran Sword with Berserking. The combat logs are attached.

WWS parses for each:
SoV - 47% uptime, 76 applications.
SoW - 20% uptime, 25 applications.

This likely settles the "do HV applications proc ppm-based enchants" question.

I imagine that SoV also increases the uptime of Mongoose as well; does this significantly change its value in comparison to the other available enchants? With TotC's requirements on EH I think Blood Draining may still win out, but assuming ICC has some more avoidable damage... Just thinkin' ahead.

[theckhd]

I imagine that SoV also increases the uptime of Mongoose as well; does this significantly change its value in comparison to the other available enchants? With TotC's requirements on EH I think Blood Draining may still win out, but assuming ICC has some more avoidable damage... Just thinkin' ahead.

Mongoose is also 1PPM, so it should work identically to Berserking in terms of proc rate and uptime. This is already taken into account in the Enchant comparison (see the mongoose uptime calculation that immediately precedes it).

[Iselian]

Fantastic, thank you. I really ought to check the table of contents next time, or at least bother with the search feature. >.<

[DisRuptive1]

When off-tanking is it ok to use Exorcism to increase DPS/TPS or is it better to lose the global cooldown for the melee swings?

[tlitp]

mongoose uptime calculation

I've recalculated the uptime using exactly the same input variables, but this time not ignoring the first and second-order effects of mongoose (2% physical haste, its reflection into the parryhaste/reckoning factors) :
computed_uptime=52.98%
refined_uptime=53.26%

out of sheer curiosity, I've remade the computations for two additional cases, 3R/1V/3C and 2.6 base weapon speed (again, using the same input variables) :
(1V/3C,1.6)=53.26%
(3R/1V/3C,1.6)=54.45%
(1V/3C,2.6)=58.59%

TLDR : the uptime of mongoose scales fairly well with the base weapon speed; reckoning also boosts it further, but to a lesser extent

Theck, we'll have to rework the PHR module to somehow manage mongoose and blade ward; however, until we have solid data on blade ward...
@community : gief blade ward testing nao

[theckhd]

When off-tanking is it ok to use Exorcism to increase DPS/TPS or is it better to lose the global cooldown for the melee swings?

I've done the calculation somewhere on the boards in the past two weeks. It turned out that even against an undead mob, you'll lose more dps due to melee + SoV procs than you gain by casting Exorcism.

[trellian]

Since there's quite a debate going on in the gear forum, would you mind testing T9.25 headpiece versus T2-245?

I'm over softcap of expertise myself (some 10%-ish with SoV up) so I'm wondering how much of a threat loss would the loss of strength and BV be compared to gaining a lot of +hit.

I'm guessing it will be a loss, but not as much as most people seem to think.

[theckhd]

Threat with T9.245: 8007 TPS
Threat with T2.245: 7906 TPS

So the Judgement helm is a net threat loss of 100 TPS. That's at the expertise hard-cap (the default set has 57 expertise in it), but below hit cap (only at around 6% with the T2.245 helm).

I'm tempted to update the default gear set a bit to optimize it more. For example, going with 4-piece T9.245 might be a better estimate of what people are using for hard mode progression.

[Racolus]

Threat with T9.245: 8007 TPS
Threat with T2.245: 7906 TPS

So the Judgement helm is a net threat loss of 100 TPS. That's at the expertise hard-cap (the default set has 57 expertise in it), but below hit cap (only at around 6% with the T2.245 helm).

I'm tempted to update the default gear set a bit to optimize it more. For example, going with 4-piece T9.245 might be a better estimate of what people are using for hard mode progression.

?_? WTH, aren't hit supposed to be the better threat stat than BV? I put the T2.245 helm first on my wish list, grrrr...

[lythac]

?_? WTH, aren't hit supposed to be the better threat stat than BV? I put the T2.245 helm first on my wish list, grrrr...

You may have missed that T2.245 has 59 less strength than T9.245.

3.2.2 Stat Threat Analysis

And finally, here's the single-stat values pulled off of the first graph (STR scaling) at 1400 STR:

Code: Select all

         TPS per 10
Stat   ipoint    stat
STR     28.81   28.81
BV      25.36    8.24
Hit     23.34   23.34