WTB Awesome Theorycrafting from KeyboardNinja

[Azurestone]

WTB 3.0 gearz and stats theorycrafting from KeyboardNinja. What does the change in accuracy and alacrity mean for PVP and PVE? Should we keep going with main stat augs or power augs? How much crit is good? WTB KBN theorycrafting!

[Warlockanap]

I had a crunch on the numbers a few days ago for the augments, at least for Sorcerers it appears that main stat augments are the way to go -- personally I find losing 1.79% Crit Chance for a measly 29 Bonus Damage a terrible trade.

[Azurestone]

I had a crunch on the numbers a few days ago for the augments, at least for Sorcerers it appears that main stat augments are the way to go -- personally I find losing 1.79% Crit Chance for a measly 29 Bonus Damage a terrible trade.

 

If you were high power on everything but one piece of full crit rating, and use power augs for everything, you get massive boost in bonus dmg and that one piece of full crit definitely gives you that 1.79%.

[Shenyu]

Honestly for Operative Healer I'm just going to go with what I had prior to patch 3.0 and then I'll experiment from there on out.

 

I ran full Power/Surge with 3 Alacrity enhancements. Though I may try full power surge and 2 crit/alacrity enhancements and mess around with a high alacrity or high crit build. Frankly I just don't know and I can't wait to spend thousands and thousands of ranked coms trying out new builds.

 

It'd be nice to know the new DR points though just so we can base our theory off the cheat sheet.

Edited December 11, 2014 by Shenyu

[Azurestone]

It'd be nice to know the new DR points though just so we can base our theory off the cheat sheet.

 

Exactly. KBN plz. Stop rushing for world #1s! Bring some love to all us nubcakes who can't do theorycrafts! Many thanks ahead!

[Euclidi]

There is an excellent article for Vanguard DPS on dulfy.net that goes into stats FOR THAT CLASS, but there are some good points made that apply to all stats

[KeyboardNinja]

Well, I don't know of anyone who has completely worked out all of the optimal stat allocations for every class (I certainly haven't), but I do have some results that hint in certain directions. Hopefully they will be helpful…

 

• On the PTS, Tace calculated that power augments produced optimal mean DPS for every DPS spec, but only by a slight margin
  
• Oderken similarly calculated the same result for healers, though this was before a change which gave Merc healers a ton more surge (at the expense of some other things)
  
• Surge and crit bonuses are down across the board, with really only Merc healers still having substantial values of either
  
• Tace and I had a discussion a few months ago where he calculated that high crit (and by extension, trading power for main stat) is correlated with higher variance and lower means in DPS output for almost every class. At the time, Scrapper/Concealment was an outlier due to its massive crit talents, but those talents have been reduced significantly
  
• I've a whole bunch of parsing with various classes, and the results seemed to bear out Tace's math: the best average DPS and lowest variance (i.e. best raid DPS) comes from power augments, but the absolute best "god parses" come from main stat
  

 

So there's no hard, over-arching math here, but the results hint pretty strongly that power augments are optimal for almost everyone. Merc healers probably do want Aim augments (potentially also Arsenal mercs), but I doubt that many other classes want main stat augments instead of power. At least, if they care about averages. If you want to just optimize your peaks (e.g. for dummy parsing, or maybe for arenas), then main stat augments and a bit of crit are a good idea.

 

Regarding accuracy, no you don't want any for PvP. Unless you feel like you need help killing tanks, you just don't want it. MOAR SURGE.

 

With respect to alacrity in general, it's a fantastic stat, and while surge is point-for-point better than it until surge gets into the 300s, alacrity is far from a waste. The biggest problem is that alacrity magnifies downtime and latency as much as it magnifies DPS. So, if you're doing a boss that has a lot of downtime, or your connection has a lot of lag, or you're in PvP where there are a lot of rotational interrupts and such, alacrity is a LOT less attractive than it could be.

 

I know that doesn't definitively answer your questions, but hopefully it gives you a bit more information with which to make a decision. :-)

[Zoom_VI]

Regarding accuracy, no you don't want any for PvP. Unless you feel like you need help killing tanks, you just don't want it. MOAR SURGE.

 

Actually I wouldn't be so sure. Me and Grizz where figuring it would be good to have at least one likely two accuracy enhancements. Simply because the surge DR curve is a b*tch, and so is missing HiBs. This server's proclivity to have loads of sorcs in warzones tends to help accuracy too.

 

but definitly agree about the power augs, everyone knows how good you are at peeveepee is determined by how big yo crits are.

[KaiserTNT]

On the PTS, Tace calculated that power augments produced optimal mean DPS for every DPS spec, but only by a slight margin

 

If the margin is slight, then that raises the question: does the Unlimited Power raid buff (10% mainstat for 10 seconds) tilt the margin back into the favor of mainstat augments, or was that accounted for in the calculation?

Edited December 11, 2014 by KaiserTNT

[Luckygunslinger]

Tam delivered, send 1.5 mil to Sairotra or Sinais.... I'll make sure he gets it.

[Azurestone]

Well, I don't know of anyone who has completely worked out all of the optimal stat allocations for every class (I certainly haven't), but I do have some results that hint in certain directions. Hopefully they will be helpful…

 

• On the PTS, Tace calculated that power augments produced optimal mean DPS for every DPS spec, but only by a slight margin
  
• Oderken similarly calculated the same result for healers, though this was before a change which gave Merc healers a ton more surge (at the expense of some other things)
  
• Surge and crit bonuses are down across the board, with really only Merc healers still having substantial values of either
  
• Tace and I had a discussion a few months ago where he calculated that high crit (and by extension, trading power for main stat) is correlated with higher variance and lower means in DPS output for almost every class. At the time, Scrapper/Concealment was an outlier due to its massive crit talents, but those talents have been reduced significantly
  
• I've a whole bunch of parsing with various classes, and the results seemed to bear out Tace's math: the best average DPS and lowest variance (i.e. best raid DPS) comes from power augments, but the absolute best "god parses" come from main stat
  

 

So there's no hard, over-arching math here, but the results hint pretty strongly that power augments are optimal for almost everyone. Merc healers probably do want Aim augments (potentially also Arsenal mercs), but I doubt that many other classes want main stat augments instead of power. At least, if they care about averages. If you want to just optimize your peaks (e.g. for dummy parsing, or maybe for arenas), then main stat augments and a bit of crit are a good idea.

 

Regarding accuracy, no you don't want any for PvP. Unless you feel like you need help killing tanks, you just don't want it. MOAR SURGE.

 

With respect to alacrity in general, it's a fantastic stat, and while surge is point-for-point better than it until surge gets into the 300s, alacrity is far from a waste. The biggest problem is that alacrity magnifies downtime and latency as much as it magnifies DPS. So, if you're doing a boss that has a lot of downtime, or your connection has a lot of lag, or you're in PvP where there are a lot of rotational interrupts and such, alacrity is a LOT less attractive than it could be.

 

I know that doesn't definitively answer your questions, but hopefully it gives you a bit more information with which to make a decision. :-)

 

Dayum! KBN never fails to deliver. Good job man. I am gonna study your theory and Kwerty's surge/alacrity curve on dulfy. Thanks a lot, m8!

 

Bee-tea-dubljew, from what you said, I think it's ok to say "R.I.P. the 8 mil I spent on resolve augs..."

Edited December 11, 2014 by Azurestone

[wadecounty]

Crit and mainstat over power for higher lucky parses makes no sense, since technically, the lowest amount of crit possible would give you the highest potential lucky parse possible. Or am I confusing something here?

 

Also, are the differences more than negligible, or are we talking under 30 DPS or so here? Because if its that small I'm not going to worry about it and probably use mainstat augments still, on the off chance they change accuracy (they acknowledged it at least in the dev tracker today) which allows for more surge, which would then increase the value of crits.

[BoboHut]

Well, I don't know of anyone who has completely worked out all of the optimal stat allocations for every class (I certainly haven't), but I do have some results that hint in certain directions. Hopefully they will be helpful…

 

• On the PTS, Tace calculated that power augments produced optimal mean DPS for every DPS spec, but only by a slight margin
  
• Oderken similarly calculated the same result for healers, though this was before a change which gave Merc healers a ton more surge (at the expense of some other things)
  
• Surge and crit bonuses are down across the board, with really only Merc healers still having substantial values of either
  
• Tace and I had a discussion a few months ago where he calculated that high crit (and by extension, trading power for main stat) is correlated with higher variance and lower means in DPS output for almost every class. At the time, Scrapper/Concealment was an outlier due to its massive crit talents, but those talents have been reduced significantly
  
• I've a whole bunch of parsing with various classes, and the results seemed to bear out Tace's math: the best average DPS and lowest variance (i.e. best raid DPS) comes from power augments, but the absolute best "god parses" come from main stat
  

 

So there's no hard, over-arching math here, but the results hint pretty strongly that power augments are optimal for almost everyone. Merc healers probably do want Aim augments (potentially also Arsenal mercs), but I doubt that many other classes want main stat augments instead of power. At least, if they care about averages. If you want to just optimize your peaks (e.g. for dummy parsing, or maybe for arenas), then main stat augments and a bit of crit are a good idea.

 

Regarding accuracy, no you don't want any for PvP. Unless you feel like you need help killing tanks, you just don't want it. MOAR SURGE.

 

With respect to alacrity in general, it's a fantastic stat, and while surge is point-for-point better than it until surge gets into the 300s, alacrity is far from a waste. The biggest problem is that alacrity magnifies downtime and latency as much as it magnifies DPS. So, if you're doing a boss that has a lot of downtime, or your connection has a lot of lag, or you're in PvP where there are a lot of rotational interrupts and such, alacrity is a LOT less attractive than it could be.

 

I know that doesn't definitively answer your questions, but hopefully it gives you a bit more information with which to make a decision. :-)

 

Great work Keyboardninja. Best answer I think most of us have seen so far on the question.

 

Out of curiosity, how much main stat are you guys running with before swapping to power augs and should we have a certain amount of main stat before the swap. Should we maintain say 3500 or 4k main stat as we swap or just go all at once and be done?

[KeyboardNinja]

If the margin is slight, then that raises the question: does the Unlimited Power raid buff (10% mainstat for 10 seconds) tilt the margin back into the favor of mainstat augments, or was that accounted for in the calculation?

 

It was accounted for in both the calculation and the testing. I've done a ton of tests with my sorc in both specs, and the best average parses come with a full power build. The difference is significant enough that, when the main stat build out-performed the power build, it was a dramatic and extremely unusual parse (e.g. crazy crits). Usually, the power build noticeably spanked the main stat build by a significant margin.

 

Crit and mainstat over power for higher lucky parses makes no sense, since technically, the lowest amount of crit possible would give you the highest potential lucky parse possible. Or am I confusing something here?

 

This is incorrect, but it seems very intuitive. Or rather, it's correct, but it's not statistically significant. Uh, it's weird…

 

Ok, so when you stack more power, you get fewer crits but your crits are bigger. This means in turn that less of your DPS ceiling is tied up in whether or not you crit (since, for a given hit, the crit rating on your gear is worthless if that ability didn't crit). Thus, your "lucky crit" parses are both rarer and have less of an impact on your average. If you stack more crit rating (at the expense of power), then you're very literally gambling on critting often (i.e. that the stat will "pay off"). More of your average DPS is tied up in crits, which means that you do in fact have higher variance on your parses when you trade out power for crit rating. Intuitively, we could lower our variance arbitrarily far by trading all of our critical chance for more power (thus, never critting, but hitting like a truck all the time).

 

Of course, there's a threshold here. What is interesting in this sort of analysis is not the absolute gain in critical chance, but rather the relative gain. What percentage of your total critical chance did you gain from crit rating? If you're a class with enormous crit talents (pre-3.0 concealment, and maybe live marksmanship), you just might have a high enough native crit chance such that the relative gain from crit rating is low, meaning that the variance impact from crit rating is lower. Doesn't necessarily mean that it will raise the average (that depends a lot on surge talents, damage coefficients and such), but it means that the variance might be reduced by the crit rating rather than increased. When Tace was doing this math the first time, as I recall, the inflection point at which crit rating became a variance reduction was somewhere around 37% (or higher).

 

Note that there are definitely classes even at the current (192/198) gear tier where higher crit builds will yield slightly better DPS. Fury and Rage are probably examples of such, and I suspect Arsenal is similar. However, it will still yield more variance. So, while the averages may go up, the standard deviation on your parses will also go up significantly, making these builds theoretically superior but practically unreliable. This, paired with the main stat extrema, likely means that we're going to see a rise of players running gear that is optimized for dummy parsing and terrible for raids. Literally dummy parse heroes.

 

Also, are the differences more than negligible, or are we talking under 30 DPS or so here? Because if its that small I'm not going to worry about it and probably use mainstat augments still, on the off chance they change accuracy (they acknowledged it at least in the dev tracker today) which allows for more surge, which would then increase the value of crits.

 

It's a difference of about 100-150 DPS in full augmented 192s, at least for most classes.

 

Out of curiosity, how much main stat are you guys running with before swapping to power augs and should we have a certain amount of main stat before the swap. Should we maintain say 3500 or 4k main stat as we swap or just go all at once and be done?

 

Just go full power. :-)

Edited December 11, 2014 by KeyboardNinja

[HundredBears]

but it means that the variance might be reduced by the crit rating rather than increased. When Tace was doing this math the first time, as I recall, the inflection point at which crit rating became a variance reduction was somewhere around 37% (or higher).

 

Hmm. I get 50%. I think that we had a mathematician in one of the augment threads say the same thing (the argument's pretty much: there's no variance at 0% or 100% crit chance, therefore 50% by symmetry), but here's a more detailed version:

 

We want to express the variance in damage in terms of c, our total chance of critting. It happens to be quadratic, so we won't have any trouble working out how crit rating affects variance once we get to that point.

 

First off, normalize so that we do 0 damage on a non-crit and 1 on a crit. This is just to make the equations cleaner - the constants will cancel each other out if we leave them in.

 

Thus there's a c chance of critting for 1 damage and a 1 - c chance of not critting for 0 damage, giving an expected damage of c and a variance of:

 

c(1 - c)^2 + (1 - c)c^2

 

= c - 2c^2 + c^3 + c^2 - c^3

 

= c - c^2

 

Set the first derivative equal to zero to solve for the local maximum and we get:

 

0 =1 - 2c

 

c = .5

 

Of course, the difference between 37% and 50% doesn't change much, but it does save you two the effort of running the numbers for Marksmanship.

Edited December 11, 2014 by HundredBears

[TACeMossie]

When Tace was doing this math the first time, as I recall, the inflection point at which crit rating became a variance reduction was somewhere around 37% (or higher).

 

I do believe this maths was done before I showed up actually, cause I don't remember ever doing the maths on this subject.

 

On the subject of mainstat vs power though:

 

The gain in Main Stat is 0.21 Bonus damage and some crit rating per point (after force valor but before force might), whereas power is 0.23 bonus damage per point. As such:

 

If 23 Main Stat provides more crit than 2 crit rating, then Main Stat is better than Power when talking about augments, A-mods vs lower tier unlettered mods, etc..., assuming you're grabbing crit anyway

If 23 Main Stat provides less crit than 2 crit rating, then Main Stat is worse than Power when talking about augments, A-mods vs lower tier unlettered mods, etc..., assuming you're grabbing crit anyway

 

It is my understanding that in 192s, you'll have ~3600 main stat before augments.

 

This means at gear levels, 23 Main Stat is worth 0.04% crit chance.

Meanwhile, if you have 0 crit rating, 2 crit is worth 0.041% crit chance.

 

As you get higher in the crit, sure, Main Stat will become comparable, but only just, and also having more main stat will further devalue having main stat augments.

 

And thats why Power/Crit augments are doing better than mainstat ones.

Edited December 12, 2014 by TACeMossie

[BigDumbViking]

All this new information.. .much exciting! so wow!

[namesaretough]

I'm disappointed to see people so openly referencing things from pts.

[CaptMurphy]

Of course, there's a threshold here. What is interesting in this sort of analysis is not the absolute gain in critical chance, but rather the relative gain. What percentage of your total critical chance did you gain from crit rating? If you're a class with enormous crit talents (pre-3.0 concealment, and maybe live marksmanship), you just might have a high enough native crit chance such that the relative gain from crit rating is low, meaning that the variance impact from crit rating is lower. Doesn't necessarily mean that it will raise the average (that depends a lot on surge talents, damage coefficients and such), but it means that the variance might be reduced by the crit rating rather than increased. When Tace was doing this math the first time, as I recall, the inflection point at which crit rating became a variance reduction was somewhere around 37% (or higher).

 

It was suggested in the 3.0 Marksmanship Guide thread to include a couple of surge augments along with the rest being mainstats.

 

Based off your analysis it seems like power is still the way to go for us rather than mainstat, but would you agree to include the two surges since our class has a higher inherent crit chance? Or, is it still all power all day?

[BigDumbViking]

I'm disappointed to see people so openly referencing things from pts.

 

At least they aren't referencing the PMS...

[Luckygunslinger]

To be fair, the stuff he is discussing is in game. It's not like he's referencing an augment value that was on PTS only and comparing them to live, on top of that the stats of all gear was subsequently leaked following the first PTS week, so the values could have been tested in a mathematical vacuum if need be. That being said, if you want to be disappointed with him, fine, he's helping people, I think the good out weighs the bad. Edited December 12, 2014 by Luckygunslinger

[namesaretough]

I disagree with your implication that everything in this thread couldn't be said without referring to information that falls under a NDA. The relevant information IS on the live server and should be presented as such. I don't know, maybe I'm strange for wanting people who should know better to abide by things they've signed.

[CaptMurphy]

I don't know, maybe I'm strange for wanting people who should know better to abide by things they've signed.

 

Doesn't the NDA expire once the content goes live?

[namesaretough]

Doesn't the NDA expire once the content goes live?

 

No, it doesn't.

[KeyboardNinja]

I disagree with your implication that everything in this thread couldn't be said without referring to information that falls under a NDA. The relevant information IS on the live server and should be presented as such. I don't know, maybe I'm strange for wanting people who should know better to abide by things they've signed.

 

I'm very aware of what I signed; I do take my NDAs quite seriously. I work in a field that is dripping in intellectual property restrictions and legal minutia, so I'm very familiar with how the law works in this area. (if you want to kill some time, read up on contributor licensing for open source code bases, and then follow up by looking at how these rules change when a commercial company is contributing to open source)

 

The NDA does not cover player creations derived from public information. All of the theory crafting that happened "on the PTS" was done with respect to information that is now public. And in fact, most of that theory crafting has subsequently been made public, independent of the fact that it was discussed on the PTS. If discussion of said theory crafting were restricted by the NDA, then it would be impossible for people to (for example) publish guides on the 3 Cull rotation for snipers, update a certain tank statting post, or even talk about the value of various abilities.

 

I know Hillary's statements on this matter have been somewhat ambiguous. In fact, the NDA itself is somewhat ambiguous on this point. See sections 9.2 and 14.1, which in fact impose directly opposite restrictions. I choose to abide by an interpretation of this ambiguity which aligns with standard United States jurisprudence with respect to copyright law and information provenance, and I am confident that this interpretation is legally defensible.

 

This does not mean that the NDA is without force. There are many many things that it most certainly covers and it does remain in force over those things, but nothing I have said in this thread falls afoul of its restrictions.