Affliction Simcraft

[SaltLakeAtrocity]

After over 3 hours simcrafting a multi-tude of different builds and trying to find the one that yields the biggest outcome, I can't come up with anything definitive. I've tried studying the few top-end warlocks that actually log out in affliction builds. I've tried camping raiding test dummies. Nothing yields numbers I can say beyond a margin of error are effective stat builds. What little I was able to figure out:

Most top end aff locks are just barely above 4700 haste rating, but no spreadsheet that I've ever seen has shown any significance to this 4700 mark. Simcraft seems to show that a build with 4200 haste rating is slightly better than my build with 4700, but only by 36 dps (and there were other factors with that build).

Running with level 92 hit cap seems to sim better than any other specific-number hit cap goal. About 500 dps different between 5000+ hit rating and the 4080 hit rating mark.

Int gems have finally seemingly surpassed mastery gems, but by a margin so small it almost doesn't even matter.

Simming mastery as the highest valued stat yeilds less than 15 dps (yes, 15 dps, not 15%) less than simming haste as the highest valued stat. I've decided to chuck this off as margin of error.


And even these facts I find myself second guessing every moment I find out about them. Nothing seems consistent at all.

http://us.battle.net/wow/en/characte...rhead/advanced is my armory if that helps in any way

[Cyner]

Most top end aff locks are just barely above 4700 haste rating, but no spreadsheet that I've ever seen has shown any significance to this 4700 mark. Simcraft seems to show that a build with 4200 haste rating is slightly better than my build with 4700, but only by 36 dps (and there were other factors with that build).

Running with level 92 hit cap seems to sim better than any other specific-number hit cap goal. About 500 dps different between 5000+ hit rating and the 4080 hit rating mark.

These are basically just the 2 different haste breakpoints. Get to whatever one you can.

Run with whatever hit stat weight simcraft gives you. (Personally I go under hit cap)

Gemming: I go with mastery gems, but if Int gems were cheaper, I would go that way.
Basically everything you pointed out is such a small difference, there is no use in worrying about it.

[SaltLakeAtrocity]

Not having a clear-cut definitive 'best' drives me crazy. But because of work and other irl obligations I havn't had but only a couple hours to play around with stuff.

i understand the 4198 haste breakpoint, thats why i ran it. the 4700 breakpoint is completely beyond me.

And simcraft hasn't defined a perfect hit rating. Its weighted higher than haste for me, yet every time i run higher haste in lieu of hit it gives me higher dps sims (usually around 500 dps higher than a hit-over-haste build). That contradicts itself disgustingly. Once again proving that this class has been made quite ridiculous with stat weights.

[Cyner]

Not having a clear-cut definitive 'best' drives me crazy. But because of work and other irl obligations I havn't had but only a couple hours to play around with stuff.

i understand the 4198 haste breakpoint, thats why i ran it. the 4700 breakpoint is completely beyond me.

And simcraft hasn't defined a perfect hit rating. Its weighted higher than haste for me, yet every time i run higher haste in lieu of hit it gives me higher dps sims (usually around 500 dps higher than a hit-over-haste build). That contradicts itself disgustingly. Once again proving that this class has been made quite ridiculous with stat weights.

When you are running a higher haste build you are probably hitting the 2nd haste breakpoint, hence the higher dps you are seeing. But as far as "clear cut" there really isn't an answer, its more of a personal preference,
Don't mind seeing haunt miss? Then go under...It pisses you off? Then hit cap.

[SaltLakeAtrocity]

I've acknowledge the hit cap is a dps loss. Just how much hit cap yields precisely how much dps loss? Where's the breakpoint for when hit rating starts noticeably scaling down on dps? 4080 is a referenced number, but virtually no locks use that. Usually some arbitrary number around 13% hit. Where they got that number is beyond me. Just like I have yet to see any spreadsheet that says there really is a breakpoint at 4700 haste rating. I just have to assume it is, and guess what rating the actual break point falls on.

Its a debate between the: as-much-mastery-as-possible build

or the

as-much-mastery-as-absolutely-possible-after-hitting-4198/4700-haste-rating

or the

a-little-bit-less-mastery-for-little-more-hit-while-keeping-haste-breakpoint

or the

a-little-bit-less-haste-for-more-hit-so-you-dont-sacrifice-mastery



502 gloves with alot of mastery and higher spell power but almost 400 useless crit rating versus 496 tier gloves with hit and haste on them (both useful stats) but less spell power.




I've literally taken advanced engineering courses where the math didn't make me want to rip my hair out as much as this does. Quite overwhelmingly ridiculous.

[Woz]

It's not really a debate between all of those. It's a debate between:
- full haste (with or without hit cap)
- full mastery (with or without hit cap)
- haste breakpoint (4198 or 4717) then mastery then hit
- haste breakpoint (4198 or 4717) then hit cap then mastery

"all of the high end locks" understand the haste breakpoints and go for the 4717 one now that it's more easily attainable.

All that the 4717 breakpoint does is gets you another tick on one of your dots, which ultimately increases your DPS.

As far as your glove dilemma - if using tier gloves gets you 2 piece, then use them. If not - then don't. Don't even need to sim that to tell you that. The only time that wouldn't be the case is because the tier gloves would give you one of those two haste breakpoints.
You may want to review this chart for further knowledge of how the haste breakpoints work for us: http://www.totemspot.com/vb/entry.php?b=44

[Shauni]

The information above is perfectly correct.

You have to remember that going below hit cap actually rises your DPS, but also rises your necessary skillcap.

But, as for me, I choose the following minimal state :
- 4717 haste for the additional corruption tick (hmmmm moar shard regen !)
- 4080 hit which makes 12% hit (below that my poor skills makes my DPS drown)
- 5000 mastery. a good number made by my own opinion.

This is the minimal stat that I chose. You can try different builds like à 6k+ Haste for a godlike shard regen but you will have to sacrifice your hit/mastery. I've tried it but I was awfull.
There is no "perfect" state.

Then, when you have found your minimal values, use the simcraft stat weight to know what stat you should aim when upgrading your gear.

[Deleted]

I don't say I'm a top end warlock, but I raid in top50 guild as our only warlock, and I'm going with full haste, seems like I'm pretty different.
Yesterday I simmed my gear with 4717, 6637 and 9211 haste breakpoints, everything left reforged/gemmed to mastery.
They all did inside 50dps of each other which can be ruled as RNG even with my 50k iterations.

I go which way I like to play, and it has always been haste so I go with it.
My DPS got ~20k better in heroic will of emperor after going from mastery to haste, altho I got LFR Elegon trinket before that(and played a bit better too I think)

My armory if you like to see it for proof
http://eu.battle.net/wow/en/characte...pappa/advanced

edit: Don't be afraid to be different, what suits you may not be good for others, I've seen top warlocks going for pretty pure int and int+mastery gemming and they are doing great in WoL

[Mayhem]

nice transmog einomies

and thanks for the info, i´ll give it a try with more haste and test it myself

[Shauni]

Mayhem, remember that Einomies has 491 ilevel, as he say, don't try to imitate. Test, simulate and draw your conclusion.

Buuuuuut personnaly, I wouldn't try a full haste build at entry raid level gear....

[Deleted]

I was at 485 when I changed to full haste

[Deleted]

when you use simcraft you also need to take into account the limits it has bcoz no matter how good it is programmed it can never simulate YOU, it will always be a program that simulates static events in a fictitous encounters, not unlike the real testing on an actual encounter but ofc in the end you choose how you use simcraft.

i have 13,76% hit, 4717 haste and 58,6ish% mastery and i was told i would be better off to reduce my hit to 12%, so i ran some simcrafts with 12,04% hit, 4717 haste and 60,8ish% mastery and what this change got me was a 0,2% dps upgrade or 185ish dps, it hardly seemed worth it, so i went back to my old stats since more stable dps is better for me than unstable dps and that is what i prefer and that is all that matters, ppl tend to forget that simcraft is a tool, not gospel and shouldnt be taken too seriously.

[Shauni]

Which is hardly entry-raid gear level, but good information nevertheless

[Mayhem]

yeah no i am far from bis or any of his stats, as of now it would be impossible to switch to fullhaste ^^ but i hope in a week or two my luck will actually find me again, stupid mages

[Rustjive]

Not having a clear-cut definitive 'best' drives me crazy. But because of work and other irl obligations I havn't had but only a couple hours to play around with stuff.

i understand the 4198 haste breakpoint, thats why i ran it. the 4700 breakpoint is completely beyond me.

And simcraft hasn't defined a perfect hit rating. Its weighted higher than haste for me, yet every time i run higher haste in lieu of hit it gives me higher dps sims (usually around 500 dps higher than a hit-over-haste build). That contradicts itself disgustingly. Once again proving that this class has been made quite ridiculous with stat weights.

As has been said over and over again, simming stat weights and reforging to them and simming again gets you nowhere. Use the reforge plots.

The breakpoints are 4198 (UA/Agony lust/DS breakpoint), 4717 (Corruption w/o Lust/BS breakpoint), and after that probably 6637 (Agony-15 at 6400, UA-9 at 6637). After that 9261 is probably the last relevant breakpoint (theoretically).

There's no such thing as perfect hit rating. There are a variety of situations in actual raids that preclude that. You might be killing level 90 adds, you might be killing level 92 adds. Find out what situations you're raiding in, and reduce hit to fit your comfort level in this situations and keep it at that. There is absolutely no point in reforging to x% hit and realizing that you hate it for a 1% DPS increase.

[Whydrood]

After over 3 hours simcrafting a multi-tude of different builds and trying to find the one that yields the biggest outcome, I can't come up with anything definitive. I've tried studying the few top-end warlocks that actually log out in affliction builds. I've tried camping raiding test dummies. Nothing yields numbers I can say beyond a margin of error are effective stat builds. What little I was able to figure out:

Most top end aff locks are just barely above 4700 haste rating, but no spreadsheet that I've ever seen has shown any significance to this 4700 mark. Simcraft seems to show that a build with 4200 haste rating is slightly better than my build with 4700, but only by 36 dps (and there were other factors with that build).

Running with level 92 hit cap seems to sim better than any other specific-number hit cap goal. About 500 dps different between 5000+ hit rating and the 4080 hit rating mark.

Int gems have finally seemingly surpassed mastery gems, but by a margin so small it almost doesn't even matter.

Simming mastery as the highest valued stat yeilds less than 15 dps (yes, 15 dps, not 15%) less than simming haste as the highest valued stat. I've decided to chuck this off as margin of error.


And even these facts I find myself second guessing every moment I find out about them. Nothing seems consistent at all.

http://us.battle.net/wow/en/characte...rhead/advanced is my armory if that helps in any way

If simcraft is giving you a number that is as low as 15 dps, it is statistically insignificant. You can run the same toon through a sim over and over and your chances of getting the number you got when you first simmed is nearly 0. It is a margin of error.

As far as spreadsheets saying why 4717 is better: https://docs.google.com/spreadsheet/...ZmFYa0E#gid=12

As far as stat weights go, 500 dps really isn't much in the long run. Let me put that into a bit more perspective for you:
In a 5 minute fight that equates to be 150,000 damage (500dps*300sec).
In your current gear you will pull on average about 89.7k DPS, which over 5 minutes is 26.7 million damage (89,700dps*300sec).
150,000 damage (or 500 DPS) accounts for 0.006% (yes that number is right, I even rounded up, then checked on google) of your total damage done (150,000dps/26,700,000 damage done).

I mean it would be great if we could run the numbers and get the same thing every time, but with the amount of RNG that goes into calculating affliction dot damage, this just isn't possible.

[Caltiom]

If simcraft is giving you a number that is as low as 15 dps, it is statistically insignificant. You can run the same toon through a sim over and over and your chances of getting the number you got when you first simmed is nearly 0. It is a margin of error.

As far as spreadsheets saying why 4717 is better: https://docs.google.com/spreadsheet/...ZmFYa0E#gid=12

As far as stat weights go, 500 dps really isn't much in the long run. Let me put that into a bit more perspective for you:
In a 5 minute fight that equates to be 150,000 damage (500dps*300sec).
In your current gear you will pull on average about 89.7k DPS, which over 5 minutes is 26.7 million damage (89,700dps*300sec).
150,000 damage (or 500 DPS) accounts for 0.006% (yes that number is right, I even rounded up, then checked on google) of your total damage done (150,000dps/26,700,000 damage done).

I mean it would be great if we could run the numbers and get the same thing every time, but with the amount of RNG that goes into calculating affliction dot damage, this just isn't possible.

Statistical error is no problem at all. The central limit theorem allows you to calculate a exact error (with certain confidence level, 95% by default in SimC). You can decrease that error as much as you like, you just need more iterations.

The problem are other inaccuracies: mostly potential modelling inaccuracies/errors. So you're right in saying not to care too much about those small differences. I'd say everything that goes below 0.1% falls under that.