Pawn waaay off

[Rizloff]

So I found what seems to be my highest siming setup and imported my pawn string
( Pawn: v1: "FURY": Class=Warrior, Spec=Fury, Strength=11.48, CritRating=11.19, HasteRating=11.81, MasteryRating=13.16, Versatility=13.86 )
and Pawn showed that a ring I had was a 26% upgrade so I equipped it and re simmed

results

Initial
683871 dps
After supposed 26% upgrade ring
675373 dps


How can it be so wrong only swapping one piece?

[HockeyVG]

Your weights seem off for starters, though it is gear dependent and could be what works with your gear.
Another thing to keep in mind, Pawn does not take into account special effects such as the Melandrus (Auto-Attack Ring) or Collapsing Futures.

Also be aware that once you equipped that ring, your stat weights probably changed again. Pawn is a useful tool to get a general idea of what pieces might be better, but in the end your sim is the determining factor.

[Tehterokkar]

Do not trust pawn with rings/necks as they are highly volatile when it comes to changing stat weights for some specs(such as Fury).

Armor slots(pieces with Agi/Str/Int) are much more stable with pawn.

[Clozer]

So I found what seems to be my highest siming setup and imported my pawn string
( Pawn: v1: "FURY": Class=Warrior, Spec=Fury, Strength=11.48, CritRating=11.19, HasteRating=11.81, MasteryRating=13.16, Versatility=13.86 )
and Pawn showed that a ring I had was a 26% upgrade so I equipped it and re simmed

results

Initial
683871 dps
After supposed 26% upgrade ring
675373 dps


How can it be so wrong only swapping one piece?

So it wasnt Pawn's but simcraft's fault.... mostly it's about haste and some breakpoints, aka you could do another special ability during your cd's and stuff.

[Rizloff]

Your weights seem off for starters, though it is gear dependent and could be what works with your gear.
Another thing to keep in mind, Pawn does not take into account special effects such as the Melandrus (Auto-Attack Ring) or Collapsing Futures.

Also be aware that once you equipped that ring, your stat weights probably changed again. Pawn is a useful tool to get a general idea of what pieces might be better, but in the end your sim is the determining factor.

After finding a couple other upgrades siming I now get the following weights

( Pawn: v1: "FURY": Class=Warrior, Spec=Fury, Strength=11.97, CritRating=11.58, HasteRating=15.85, MasteryRating=14.08, Versatility=13.90 )

Do these weights still seem way off? Why could this be?
Currently at
31.7% Haste
37.96% Mastery
15.21% Crit
4.5% Vers

Do not trust pawn with rings/necks as they are highly volatile when it comes to changing stat weights for some specs(such as Fury).

Armor slots(pieces with Agi/Str/Int) are much more stable with pawn.

After finding the couple upgrades, the supposed gains from some of the ring and neck pieces are now lower (under 10% gains) but still not actual gains.

[Deleted]

Post your character profile pls.

[Rizloff]

Post your character profile pls.

worldofwarcraft.com/en-us/character/sargeras/gilaria

had to drop the front so its not a link because I dont have enough post to put links or something

[Deleted]

After finding a couple other upgrades siming I now get the following weights

( Pawn: v1: "FURY": Class=Warrior, Spec=Fury, Strength=11.97, CritRating=11.58, HasteRating=15.85, MasteryRating=14.08, Versatility=13.90 )

Do these weights still seem way off? Why could this be?

These weights don't seem off at all.
But I'd agree with you, I've no idea what they actually mean, I've often found that swapping gear round according to the weights actually produces worse results when re-simmed.
I've no idea what the algorithm is that gives these weightings and how it could be interpreted for any practical purpose.
But what I do believe is that these weights can not be used in the way pawn suggest they can.

[HockeyVG]

After finding a couple other upgrades siming I now get the following weights

( Pawn: v1: "FURY": Class=Warrior, Spec=Fury, Strength=11.97, CritRating=11.58, HasteRating=15.85, MasteryRating=14.08, Versatility=13.90 )

Do these weights still seem way off? Why could this be?

These weights fall more in line with what "normal" Fury weights could be. Again, Pawn is only really useful for a general overview of "Hey, this might be an upgrade." You shouldn't follow it to the letter.

[Sting]

I'm no expert on the matter, but this could be it:

If you were over a certain haste breakpoint in your first sim, simcraft could recommend going for mastery now instead of haste, as is shown in the stat weights. You import the pawn weights, and all of a sudden it's telling you that the ring you have in your bags with a large amount of mastery but little to no haste is a great upgrade. What pawn doesn't tell you is that losing the haste you have on your current ring will actually drop you below the breakpoint, screwing your dps.

Always trust sims over pawn weights. If you're unsure, try the target dummy.

[Rhyzoh]

These weights don't seem off at all.
But I'd agree with you, I've no idea what they actually mean, I've often found that swapping gear round according to the weights actually produces worse results when re-simmed.
I've no idea what the algorithm is that gives these weightings and how it could be interpreted for any practical purpose.
But what I do believe is that these weights can not be used in the way pawn suggest they can.

No idea of the algorithm?
Don't over think it
It runs a sim of your character with its current stats, for example it could say you did 500k DPS.
Then it runs a sim of your character with its current stats + 1333 haste, and it say you did 530k DPS.
So according to the sim adding 1333 haste increased your DPS by 30k.
Therefore you stat weight for haste is 30000/1333 = 22.5.

Its all documented here and there are multiple different setting that could be set up to change how the calculation is made
https://github.com/simulationcraft/s...i/StatsScaling

[Deleted]

No idea of the algorithm?
Don't over think it
It runs a sim of your character with its current stats, for example it could say you did 500k DPS.
Then it runs a sim of your character with its current stats + 1333 haste, and it say you did 530k DPS.
So according to the sim adding 1333 haste increased your DPS by 30k.
Therefore you stat weight for haste is 30000/1333 = 22.5.

Its all documented here and there are multiple different setting that could be set up to change how the calculation is made
https://github.com/simulationcraft/s...i/StatsScaling

[Edited, thought II was replying to OP]
Thanks for the link, very interesting.
I'm still not sure the purpose of the stat weights, if I paste them into the pawn addon then make a gear change then re-sim it seems as likely as not that simcraft believes it is a dps loss.
So I have taken to resimming each gear change as opposed to using pawn recommendations.
I'm fairly sure this isn't just due to a bug in the pawn addon applying the multipliers, as occasionally I did double check then myself.

[Deleted]

As Tehterokkar said pawn doesn't seem to work well with jewelry cause they don't have main stats. Thus it sort of overvalues the potential upgrades and definately doesn't account gems in those 3 items, it does calculate them in the rest thought. Thats what i have experienced at least.

[Rhyzoh]

[Edited, thought II was replying to OP]
Thanks for the link, very interesting.
I'm still not sure the purpose of the stat weights, if I paste them into the pawn addon then make a gear change then re-sim it seems as likely as not that simcraft believes it is a dps loss.
So I have taken to resimming each gear change as opposed to using pawn recommendations.
I'm fairly sure this isn't just due to a bug in the pawn addon applying the multipliers, as occasionally I did double check then myself.

In my experience mast/verse/crit have been fairly linear, its haste that has some volatile behavior mostly due to the various different breakpoints.
Some of the breakpoints for example are,
Fury Haste Breakpoints
(NOTE: With 0 delay, requires more haste in practice)
9.09% = 5 GCDs into Enrage with 4PC
9.09% = RP + 5 GCDs into BC
12.5% = 4 GCDs into Enrage without 4PC
12.5% = RP + 4 GCDs into Enrage with 4PC
20.0% = RP + 3 GCDs into Enrage without 4PC
25.0% = 6 GCDs into FB
28.6% = 7 GCDs into BC
33.3% = RP + 5 GCDs into FB
35.1% = RP + DoS + 3 GCDs into BC (this one assumes 0.3 sec delay)
35.7% = RP + 6 GCDs into BC
36.4% = 6 GCDs into Enrage with 4PC

As far as the usefulness of these non constant stat weights, well you're right since they are constantly changing, the result is not that great. This was actually a big motivation for Newton when he created calculus. A velocity is distance divided by time, but this is only true when velocity is constant. The whole point of derivatives is to find the instantaneous rate of change. DPS however is not a well defined function, it would contain many different variables and be nearly impossible to come up with, which is why simc was even created in the first place.

In any case, I do not use pawn nor would I recommend it. All it does is addition and multiplication and I don't understand the communities obsession with it. But I wouldn't go as far as to say that say simc stat weights are useless. They are just another tool at your disposal to help you give an idea of what stats are worth relative to each other.

[Archimtiros]

In my experience mast/verse/crit have been fairly linear, its haste that has some volatile behavior mostly due to the various different breakpoints.
Some of the breakpoints for example are,
Fury Haste Breakpoints
(NOTE: With 0 delay, requires more haste in practice)
9.09% = 5 GCDs into Enrage with 4PC
9.09% = RP + 5 GCDs into BC
12.5% = 4 GCDs into Enrage without 4PC
12.5% = RP + 4 GCDs into Enrage with 4PC
20.0% = RP + 3 GCDs into Enrage without 4PC
25.0% = 6 GCDs into FB
28.6% = 7 GCDs into BC
33.3% = RP + 5 GCDs into FB
35.1% = RP + DoS + 3 GCDs into BC (this one assumes 0.3 sec delay)
35.7% = RP + 6 GCDs into BC
36.4% = 6 GCDs into Enrage with 4PC

As far as the usefulness of these non constant stat weights, well you're right since they are constantly changing, the result is not that great. This was actually a big motivation for Newton when he created calculus. A velocity is distance divided by time, but this is only true when velocity is constant. The whole point of derivatives is to find the instantaneous rate of change. DPS however is not a well defined function, it would contain many different variables and be nearly impossible to come up with, which is why simc was even created in the first place.

In any case, I do not use pawn nor would I recommend it. All it does is addition and multiplication and I don't understand the communities obsession with it. But I wouldn't go as far as to say that say simc stat weights are useless. They are just another tool at your disposal to help you give an idea of what stats are worth relative to each other.

Damn Ryzoh going Philosophiæ Naturalis on yo ass!

FWIW, Mastery is a fairly linear stat, until the introduction of Haste. Funny thing about Haste is that affects all the other stats - the more haste you have, the more often you're rolling for a crit, the more often you have the chance to be enraged, the more attacks you fit inside enrage, etc.

This is an enhanced view on a smaller scale, but the trend is fairly similar in long term projections as well (a little closer together, but haste tends to be slightly higher than mastery):


If you were to look a little further out, you'll see they intertwine, one will be on top, then the other, and back again. That's due to their synergistic relationship, and the advent of those breakpoints Ryzoh mentioned (there are even more than those listed, but they are the more important ones):

*Note: these two images are simply examples I pulled, reflective of the relationships, not necessarily one another.

This is why you can't simply run simc and trust the stat weights you're given. Every time someone says "you need to sim your own gear to find out if haste or mastery is better for you", I cringe. That's not a good answer, because it only gives you a very short term look at the stats. Change one gem or piece of gear, and they could flip right back. Ultimately, they're never going to be completely linear, but they do follow a trend, making it fairly easy to tell what you should be doing with them.

TLDR: Stat weights lie.

[TheBull300]

My problem with gear is it is to mathy. Not once can you look at something and say it is better. You have to sim and re sim untill the cows come home to know what gear set up works best untill your next item drops to do it all over again