SimCraft Stat Weights sim lower DPS than Mr. Robots

[Blaizze]

So I have been using Simcraft a lot lately and also have begun using Mr. Robot a bit more for reforges.

When I run Simcraft with my gear and analyze my stat weights and input them in Mr. Robot, run the reforge and sim the toon again, my DPS Always comes out lower with the Simcraft stat weights than if I just optimize it using the Mr. Robot default stat weights.

It isn't much of a difference, about 500 or so maximum at times, but it is still always lower... no matter how many iterations etc.

Am I obtaining my stat weights incorrectly via simcraft (just selecting the analyze boxes of all important stats) or is Mr. Robot just optimal overall.

[Deleted]

Simcraft stat weights are right, you just don't know how to interpret them. Stat weights measure how much dps you win when you add a point of a certain stat, for example, Int's weight is 4.0 means that each point of int provides a boost of 4 dps, but when you reforge you aren't just adding a certain stat, you are removing from one to add to another, which might provide to be a loss of dps overall.

[Deleted]

Are you using simcrafts reforge plots or just simming, finding stat weights, then shoving those in askmrrobot?

If you're just ramming the simcraft stat weights in and reforging /gemming all of your gear around them, that's not what simcraft is telling you - it's saying gaining a few more of stat x will be better than stat y, not telling you to shift all of your stats into stat x.

Secondary stats all scale in synergy with each other to some degree, ANY change in your stats will change the values of the others.

[Blaizze]

Are you using simcrafts reforge plots or just simming, finding stat weights, then shoving those in askmrrobot?

If you're just ramming the simcraft stat weights in and reforging /gemming all of your gear around them, that's not what simcraft is telling you - it's saying gaining a few more of stat x will be better than stat y, not telling you to shift all of your stats into stat x.

Secondary stats all scale in synergy with each other to some degree, ANY change in your stats will change the values of the others.

I was just finding the stat weights and ramming them into the stat weights portion on Mr. Robot. I assumed that is what people also did with Reforge Lite.

[Swol]

Blaizze,

Like Nagassh mentioned, SimC is finding a very localized solution when it analyzes the stat weights. Those numbers it calculates are not meant to be used with a tool like Ask Mr. Robot. AMR is examining stats over a very broad range, and our default weights are tailored for this expectation. We use SimC to help us create the default weights, but we don't use SimC's stat weight analysis - we use it to run simulations and then do our own analysis of the results. That is why you are always seeing better results with our default weights - they were designed for use with a gear optimizer!

In general, the value of stats varies much less than some folks would lead you to believe. You can use the same set of weights for a very large range of stats and get optimal (or very near optimal) theoretical results. We offer multiple stat weight "presets" for some specs where there can be a major shift due to gear or talents, but there really aren't that many. We definitely don't discourage people from making their own weights - but we do caution people about what you found for yourself: you can't use the weights SimC spits out by default with a gear optimizer - they weren't intended for that. Our defaults are there to save people a lot of work that it takes to find the actual weights which would get you the best result in SimC.

Also, if you are using any soft caps in Mr. Robot and providing a weight before and after the cap -- SimC does not calculate those values for you. You have to run extra simulations and manually find those values.

[Deleted]

Unless I see wrong, AMR only has soft caps for haste, and every other stat is unrestricted. Therefore it is just as sophisticated with crit-mastery balancing, as are people who pick scale factors from simc results of a single reforge profile.

Even if AMR changed the priority of crit and mastery at some conditions, depending possibly on max amount of secondary stats given, or intellect and so forth - it would still only go from one extreme to the other. How realistic optimization do you think that is? And if by chance there are no changes in the priority then, most likely, both simc and AMR are approving the same stat, and when people use it with reforge tools they get the exact same gear as with your defaults.

Also, if you are using any soft caps in Mr. Robot and providing a weight before and after the cap -- SimC does not calculate those values for you. You have to run extra simulations and manually find those values.

Never make me go to AMR again.
...

Besides there is only one haste scale factor per cap, since you are close to a cap and range of the stat is therefore small. By scale factor I mean the partial derivative of dps by haste, so that this variable has infinitesimal increases. On a dot haste cap the entire game does not break down, usually and the partial derivative bellow and above the cap has the same approximate constant value. At the cap itself dps just isn't continuous.

If you had a scale factor calculated with large enough increases, then all scale factors before cap would be bad information. Such a value does not tell whether the cap itself is more dps than investing in other stats (which is probably the reason why it is in the AMR). For example you can make huge scale factors show up same way at some 20k haste caps just as the normal ones. But that does not mean that the actual partial derivative has not declined alot by then.

[Deleted]

I was just finding the stat weights and ramming them into the stat weights portion on Mr. Robot. I assumed that is what people also did with Reforge Lite.

If you want to find more accurate results in terms of reforging via simcraft, you'll need to enable reforge plots and set those up properly, it'll take simcraft longer to sim, but it'll give you a graph that shows increases / decreases in dps over a reforge range. Usually only worth doing with your two best secondary stats (eg; mastery and haste for affliction).

Evrelia posted a tutorial video for simcraft a while ago, that can probably explain it more clearly than I can - you should check for that in the affliction sticky or his youtube channel.

Alternatively, you can still get very good results by reading the stickies / theorycrafting threads / checking the armories of well performing warlocks and seeing how they're gearing (though take into account any differences in ilevel) and using addons like reforgelite to reforge into exactly what you want.

[Woz]

MrRobot weights are based DIRECTLY from the Simcraft profile for the respective spec. Literally. All they do there is import the Simcraft file, they are never going to be reflective of YOUR weights unless you input them yourself.

[Yuyuli]

So i did some Simulcraft, not sure if i did it right or not but im getting Int > SP > Hit > Haste ~= Crit > Mastery, can this be right? Whats ~= actually supposed to mean, how much Haste shall i get according to this, the more haste the better, or can i exchange Haste with Crit? I really don't know much about this.
My ilvl is about 490 if that matters.

[Brusalk]

So i did some Simulcraft, not sure if i did it right or not but im getting Int > SP > Hit > Haste ~= Crit > Mastery, can this be right? Whats ~= actually supposed to mean, how much Haste shall i get according to this, the more haste the better, or can i exchange Haste with Crit? I really don't know much about this.
My ilvl is about 490 if that matters.

Don't look at those, look at the actual value of the statweights.

IIRC that is showing the logical negation of what the actual results are:

IE Haste ~= Crit > Mastery should be Mastery > Crit==Haste.

[Yuyuli]

Don't look at those, look at the actual value of the statweights.

IIRC that is showing the logical negation of what the actual results are:

IE Haste ~= Crit > Mastery should be Mastery > Crit==Haste.

Thanks but yeah i did look at the actual stat weights and it's still Haste > Crit > Mastery in these, but i don't actually know how it really works
Haste(1.41) > Crit(1.40) > Masterry(1.30)

I am a bit confused, Mastery looks pretty low for me, but then again im using Patchwerk as setting, what Settings should i actually use? Is there some thread on here where i can actually check whats the best overalls ettings for simulcraft destro warlock or do i have to check this on per fight basis? Cleave, no cleave etc?

[Zoopercat]

Woz, we actually do a little more than what you are suggesting when we get our stat weights. Here's more information on how we generate them.

Also, revulva responded in another post regarding SimC's weights, and why they sometimes are not better. Sometimes they are, but you need to take the extra step of simming them to be sure.

The stat weights SimC calculates for you are a very localized solution, and aren't necessarily meant as a general stat priority for your character. They may not yield the most optimal results when used with a gear optimizer that finds the highest score for a set of stat weights. Ask Mr. Robot does a lot of extra work to find stat weights that will work over large ranges of stats to give people near-optimal theoretical DPS.

You can test this out, using the OP's character as an example, I calculated weights using SimC based on his gear as it currently is in the Armory. The results were:
Int: 3.93, SP: 3.18, Hit: 4.12, Crit: 1.42, Haste: 1.44, Mastery: 1.62.

So, one comment right off the bat: saying Hit is worth more than Intellect might be theoretically true... but it is not what you want. No one wants pure hit gems in their gear, because we already know that you can get better DPS by finding hit elsewhere. You also don't want to load up on gear with tons of Hit, since you can easily reach the hit cap without doing so. Your stat weights need to reflect what you actually want to happen on your gear - we make these adjustments in the weights we create. Mr. Robot is usually smart enough to deal with a high hit weight, but it causes your ranked lists of gear to look pretty funky with hit weighted so high.

If you optimize Oikodomeo with the Mr. Robot default weights, I get 110,446 DPS. If I use the weights from SimC, I get 108,810 DPS. The SimC weights favor mastery, but that is just a localized bump in the value, you don't really want to go switching all your stats to mastery. SimC is saying that if you added a small amount of stats to your current gear (like 500 rating), you should add it to mastery. But... you can't just add one stat to your gear - you have to take it out of something else. The reforge plots try to help with this... but I find those plots pretty confusing.

[Woz]

Woz, we actually do a little more than what you are suggesting when we get our stat weights. Here's more information on how we generate them.

Also, revulva responded in another post regarding SimC's weights, and why they sometimes are not better. Sometimes they are, but you need to take the extra step of simming them to be sure.

Hmmmm...Interesting. My mistake then!

[Berthold]

Don't look at those, look at the actual value of the statweights.

IIRC that is showing the logical negation of what the actual results are:

IE Haste ~= Crit > Mastery should be Mastery > Crit==Haste.

nope. ~= just means that it is round about the same. != means not equal.

[Deleted]

Woz, we actually do a little more than what you are suggesting when we get our stat weights. Here's more information on how we generate them.

1.We wrote a program that runs hundreds of thousands of SimC trials with hundreds of gear sets.

Already a wrong thing to do from individual's point of view. Individual simc at that person's current gear is always superior.

2. Here’s how that works: we make an actual gear set for a particular spec. We then run this through SimC and get stat weights. We put those stat weights back into Mr. Robot to see what new gear he suggests. We take that new gear and run it through SimC. We get new stat weights. We put those stat weights into Mr. Robot and get new gear. You can see how this is cycling. We continue to do this until the stat weights converge to a stable set that isn’t changing gear and optimizations.

That’s why our stat weights work well across all gear levels. They are NOT based off the BiS set like others do, they are instead based off real gear sets that 95% of players are going to have.

When you say your final step is the one, where the current scale factors no longer change reforging or the items used, this means that it is maximizing dps, correct? If the final step is at optimum dps, where gear, or items, do not change, it means that you have formed a BiS list of items. Even if you restrict this gear to be not so BiS-in-game, so it only contains T15N items for example, it is not so different from using a default gear profile, found in ready simc examples. You just compete for the correct BiS, and people have to be lucky to get themselves the exact same items thay you used.

No other kind of "converging" exists, unless the stat priority is constant with all gear, and reforging problem is as simple as minimizing crit in affliction, for example. Expression I put in italics is never correct in other cases, and this reality is not affected by your program. What you could do instead of looking for optimum, is to average the scale factors since you went through so many. Then your site would also have little to do with individuals, but atleast you had something to warn them about, and with statistics you could show what the weirdest stats are.

Problem with your method of using only scale factors is probably rising at the moment, if your max dps peak has, for example, a stat like crit maximized on the gear unrestrictedly, but the best scale factor is actually mastery, when simulated in that profile. Changing to pure mastery makes things vice versa and you just have to claim that the higher of theese two profiles is the best possible option. Then you can say you present us with special scale factors that are not tied to that reforging, but only leads to it. If you are unlucky the program might give you a gear that has minimal amount of "better stat" as base, and it only has to fill it through reforges. In the background the highest simulated dps can result from this. But you probably look at the item first and judge it with a scale factor.

Next thing is somewhat contrary with the supposition that your method makes huge decisions between investing to stats, if allowed. But generally optimization that takes steps from items and stats to anothers, also won't prove that it has reached the highest score. There can be several sort of peaks for dps, defined in stat space, and usually the first steps in the cycle, start to lead towards one of them. Once at a given peak, you would have to make drastic changes to end up with variables that are closer to other peaks. Most likely then your method can end up with its result being at a maximum that is only local, and ignore the other solutions.

I think your program only has tiny number of steps per given item set. If stats compared are haste, mastery and crit, then you only go through profile such as
"max haste, all that is left to mastery, and minimize crit"
"max haste, all that is left to crit, and minimize mastery"
...

So as explained previously, besides being wrong your site is useless in times like theese, when several stats are close to each other.

3. You may have noticed we work closely with Icy-Veins. Those guys are great (and super smart too). They run their own stat weight calculations with SimC and independently come up with their own weights. We compare our stat weights together and when they differ, we geek out determine the reason. Then both sides update weights accordingly. We also run them past theorycrafters to get feedback.

There are no published studies on your side or anyone else's. People simply on this forum have a simc, and might come up with different stat weights, but you show no reason with them? Why would we belive there is any such a talk going, and that it has any benefit for people's gear decisons?

Notice that the main reference of icy-veins is just that they work with AMR:

http://www.icy-veins.com/destruction...rity-reforging

These statistics have been obtained by combining common sense, in-game testing, and simulations using Simulation Craft.

If you are unsure what the optimal reforging strategy is for the above priority, you can Ask Mr. Robot for a personalised optimisation of your character's gear, by filling the form below.

Together you would be conquered?

(Yet icy-veins 4.8.2013 gives "Critical Strike Rating, Haste Rating, and Mastery Rating. - These statistics have roughly the same value, and depending on your gear, one will be slightly better than the others." So they admit they don't have any answer, but still they seem to restrict the answers, into type of answers, where one stat is being superior and should be maximized completely.)

If icy-veins is not a warlock theorycrafter, can you name someone that is? If there is such a thing, what can they usually do to you?

Also, revulva responded in another post regarding SimC's weights, and why they sometimes are not better. Sometimes they are, but you need to take the extra step of simming them to be sure.

If you optimize Oikodomeo with the Mr. Robot default weights, I get 110,446 DPS. If I use the weights from SimC, I get 108,810 DPS. The SimC weights favor mastery, but that is just a localized bump in the value, you don't really want to go switching all your stats to mastery. SimC is saying that if you added a small amount of stats to your current gear (like 500 rating), you should add it to mastery. But... you can't just add one stat to your gear - you have to take it out of something else. The reforge plots try to help with this... but I find those plots pretty confusing.

Doesn't the reforge plot converge into one stat that can autoreforge your gear? That is sad.

I can not reproduce anything said here without the actual input information. 1.6k dps from reforging would break boundaries of common warlock sense, but this was a shaman after all.

So I will just make another example

warlock="my current"
level=90
race=undead
role=spell
position=back
professions=jewelcrafting=600/enchanting=600
talents=http://eu.battle.net/wow/en/tool/talent-calculator#Vb!110111
glyphs=healthstone/demonic_circle/siphon_life/nightmares/unending_breath/eye_of_kilrogg
spec=destruction

actions.precombat=flask,type=warm_sun
actions.precombat+=/food,type=mogu_fish_stew
actions.precombat+=/dark_intent,if=!aura.spell_power_multiplier.up
actions.precombat+=/summon_pet,if=!talent.grimoire_of_sacrifice.enabled|buff.grimoire_of_sacrifice.down
actions.precombat+=/snapshot_stats
actions.precombat+=/grimoire_of_sacrifice,if=talent.grimoire_of_sacrifice.enabled
actions.precombat+=/service_pet,if=talent.grimoire_of_service.enabled
actions.precombat+=/jade_serpent_potion

actions=curse_of_the_elements,if=debuff.magic_vulnerability.down
actions+=/jade_serpent_potion,if=buff.bloodlust.react|target.health.pct<=20
actions+=/dark_soul
actions+=/service_pet,if=talent.grimoire_of_service.enabled
actions+=/run_action_list,name=aoe,if=active_enemies>3
actions+=/summon_doomguard
actions+=/rain_of_fire,if=!ticking&!in_flight&active_enemies>1
actions+=/havoc,target=2,if=active_enemies>1
actions+=/shadowburn,if=ember_react&active_enemies=1&(burning_ember>3.5|buff.dark_soul.up|target.time_to_die<10)
actions+=/shadowburn,if=ember_react&active_enemies>1&(burning_ember>3.5|buff.havoc.up)
actions+=/immolate,cycle_targets=1,if=ticks_remain<add_ticks%2&target.time_to_die>=5&miss_react
actions+=/conflagrate,if=charges=2
actions+=/rain_of_fire,if=!ticking&!in_flight
actions+=/chaos_bolt,if=ember_react&target.health.pct>20&(buff.backdraft.stack<3|level<86)&(burning_ember>3.5|buff.dark_soul.remains>cast_time|buff.skull_banner.remains>cast_time)
actions+=/conflagrate
actions+=/incinerate

head=hood_of_the_crimson_wake,id=94959,gems=burning_primal_80int_160hit_180int,reforge=mastery_hit
neck=megaeras_shining_eye,id=94803,reforge=hit_haste
shoulders=shaskin_mantle,id=85373,gems=80int_160hit_60int,enchant=200int_100crit,reforge=mastery_haste
back=shadowspike_cloak,id=95117,enchant=180int,reforge=hit_crit
chest=robes_of_the_thousandfold_hells,id=95328,gems=320int_80int_160crit_80int_160hit_180int,enchant=80all,reforge=hit_haste
shirt=sawbones_shirt,id=14617
tabard=consortium_tabard,id=31776
wrists=frostborn_wristwraps,id=94804,suffix=-336,gems=80int_160crit_60int,enchant=180int
hands=gloves_of_the_thousandfold_hells,id=95325,gems=160int_60int,enchant=170mastery,reforge=mastery_haste
waist=rubylinked_girdle,id=89930,gems=80int_160crit_160int_60hit,reforge=mastery_crit
legs=shaskin_leggings,id=87189,gems=320int_60int,enchant=285int_165crit,reforge=mastery_crit
feet=homewarding_slippers,id=95224,enchant=140mastery,reforge=haste_crit
finger1=fragment_of_fear_made_flesh,id=86156,enchant=160int,reforge=crit_haste
finger2=signet_of_the_shadopan_assault,id=95138,enchant=160int,reforge=haste_hit
trinket1=volatile_talisman_of_the_shadopan_assault,id=94510
trinket2=wushoolays_final_choice,id=94513,reforge=hit_crit
main_hand=suenwo_spire_of_the_falling_sun,id=96167,gems=160int_160int_60int,enchant=jade_spirit,reforge=haste_hit

# Gear Summary
# gear_strength=80
# gear_agility=80
# gear_stamina=21805
# gear_intellect=18872
# gear_spirit=80
# gear_spell_power=9440
# gear_hit_rating=5131
# gear_crit_rating=5493
# gear_haste_rating=4538
# gear_mastery_rating=5825
# gear_armor=15147
# meta_gem=burning_primal
# tier14_2pc_caster=1
# tier15_2pc_caster=1
# main_hand=suenwo_spire_of_the_falling_sun,weapon=staff_3.30speed_7946min_11920max,enchant=jade_spirit

450 s +/- 20 %, 50k sims

Scale factors put to AMR for reforge: Int-Hit-Crit-Haste-Mastery

My current: (does not apply)
AMR: 4.10-2.80-1.75-1.70-1.65
SimC: 4.10-2.80-1.70-1.75-1.65 (theese are just a pun)

450 s +/- 20 %, 50k sims

My current (that has 0.08 % more excess hit than others)
gear_intellect=18872
gear_spell_power=9440
gear_hit_rating=5131
gear_crit_rating=5493
gear_haste_rating=4538
gear_mastery_rating=5825

144189, (259 nore points of a 1.75 stat would put me highest)

AMR
gear_intellect=18872
gear_spell_power=9440
gear_hit_rating=4872
gear_crit_rating=6202
gear_haste_rating=4722
gear_mastery_rating=4960

144384

SimC
gear_intellect=18872
gear_spell_power=9440
gear_hit_rating=4872
gear_crit_rating=4106
gear_haste_rating=6818
gear_mastery_rating=4960

144541