askmrrobot a good site?

[Darkfriend]

Strength is worth *some* dps, so just imagine crit is worth three times as much as strength.

240+180=420
320*3/2=480

In this case crit wins clearly... but now let's look at Darkfriend's suggestion:


Stat weights of AMR= Strength 3.45 + Crit 1.8

1.8*2 = 3.6
3.6/3.45 = 1.04

It's already in a fair SEP range of .9-1.1 ?!

Know let's use simcraft with my daily quest focused warrior hero ilvl 500ish, 10.000 iterations later i've got the number 117.195 dps i should reach in a patchwerk fight raid scenario.

Now let's add
a) 3000 crit rating = 122.582 dps
b) 1000 strength = 119.969 dps
c) 1260 strength + 1440 crit= 123.405 dps

Interesting part is a) vs c), A is slightly more than three times your gem suggestion (960 orginal, +4%) and C is exactly three times your gem suggestion from AMR. That's a worthwhile and totally noteworthy *sarcasm* 0.6% dps difference after all, but in favor of AMR. Now what? All good warriors use AMR and bad warrior's don't? NO! ->



As some guy stated before, AMR is after all only a calculator.

Your math is wrong. SEP assumes a value of strength=1 where the secondary stat is relative to 1, greater meaning its worth more than one strength. So if simcraft shows strength being 3.51 and crit being 3.11 the SEP of crit is .89. Askmrobot shows a SEP for crit of .73.

I'm not sure where you get the 1.8x2 value either. If you're trying to do the whole "gem if the SEP of the secondary stat is greater than .5" NO DUH. That isn't the point. The point is if you have a bad SEP for crit, you're going to be meeting socket bonuses when you shouldn't be. Example. Mastery has a SEP of around .63 If by gemming hit/crit you can get at least 160 mastery from another piece of gear via reforging (since you no longer need that hit to cap) than you can substitute hit's value for masteries when deciding if a socket bonus is worth meeting. So (160x.63)+(160x.89)+(socket bonus-normally 60 str)=100.8+142.4+60=303.2 SEP. Gemming pure crit and ignoring socket is 320x.89=291.2. In this case, gemming crit isn't as good on the surface.

There are two caveats though, and a note. Personally, when I tested to see if I could use hit/crit gems, I ended up gaining 120 strength and 200 mastery, at the cost of 320 crit. Not worth it. First, at higher gear levels, the chance of being able to exchange 160 hit for at least 160 mastery and still arrive at hit cap decrease greatly, especially with Feather. So any point you don't get in mastery either has a lower SEP because it went somewhere else (haste) or because it's going towards overcapping hit, which has a SEP of around .42. Secondly, if the socket is red, and the only other option is 80 strength 160 crit, than AMRs undervalued crit will show str/crit being the better gem. Again, the problem is 160 mastery is better than 80 strength. So 80+60+(160x.89)=282.4.

Lastly, let's just check what other warriors are doing- and not just any warriors. You can see where they are meeting sockets and where they are ignoring. These are all warriors I know personally or else are known for being incredible theorycrafters, or are at the very least good players.

http://us.battle.net/wow/en/characte...alopy/advanced
http://us.battle.net/wow/en/characte...ision/advanced
http://us.battle.net/wow/en/characte...dsoul/advanced
http://us.battle.net/wow/en/characte.../Tact/advanced
http://us.battle.net/wow/en/characte...Varri/advanced
http://us.battle.net/wow/en/characte...ighty/advanced


Landsoul has been doing that for a while now. It's not better (generally speaking), but I haven't asked him yet why he still has it. I'll try and find out. Also, his high amounts of haste are simply incidental- there is a lot of haste on his gear, he isn't gemming/reforging into it.

[Deleted]

Strength is worth *some* dps, so just imagine crit is worth three times as much as strength.

240+180=420
320*3/2=480

In this case crit wins clearly... but now let's look at Darkfriend's suggestion:


Stat weights of AMR= Strength 3.45 + Crit 1.8

1.8*2 = 3.6
3.6/3.45 = 1.04

It's already in a fair SEP range of .9-1.1 ?!

Know let's use simcraft with my daily quest focused warrior hero ilvl 500ish, 10.000 iterations later i've got the number 117.195 dps i should reach in a patchwerk fight raid scenario.

Now let's add
a) 3000 crit rating = 122.582 dps
b) 1000 strength = 119.969 dps
c) 1260 strength + 1440 crit= 123.405 dps

Interesting part is a) vs c), A is slightly more than three times your gem suggestion (960 orginal, +4%) and C is exactly three times your gem suggestion from AMR. That's a worthwhile and totally noteworthy *sarcasm* 0.6% dps difference after all, but in favor of AMR. Now what? All good warriors use AMR and bad warrior's don't? NO! ->



As some guy stated before, AMR is after all only a calculator.

In your stat weight comparison you appear to be comparing 2 crit with 1 strength. No idea why you're doing that, some confusion about gems I guess, but we're comparing 1 point of crit with 1 point of strength, not 2 with 1.

1 point of crit rating is worth about the same dps as 1 point of strength for reasonable gear. Which means a 320 crit gem gives you roughly twice the dps benefit of a 160 strength gem. For my own gear, which is nothing extraordinary (SMF 516 equipped, 2 t15/2t14) adding 1 point of strength increases my dps by 3.86 while 1 point of crit rating increases my dps 3.41. Hence 1 crit = 0.88 strength. For TG, and at higher gear levels, 1 crit approaches and even surpasses 1 strength.

I have no idea what AMR is using as the basis of their stat weights but running any smf fury warrior above ilevel 500 should show crit as worth around 0.85 strength or more, while they're still using 0.78 or something. Edit: even pulling some random 491 SMF warrior out of the logs thread gave crit = 0.84 strength.

I also don't have a clue what your examples are about. 3000 crit is equivalent to 1260 strength + 480 crit in terms of gems, why are you comparing it with 1260 strength + 1440 crit and trying to use an example with nearly 1k crit worth of itembudget more as evidence of anything? Socket bonuses certainly don't account for that much.

Edit: right, figured out you were replying to Xanthan again (first you reply to Xanthan, then Darkfriend, then go back to Xanthan without making it clear). Why you made it randomly complicated by multiplying by 3 I don't understand, and was the cause of my confusion, as socket bonuses couldn't explain the massive disparity in item budget before I realised you were multiplying by 3. The comparison is 420 strength and 480 crit vs 960 crit, or 420 strength vs 480 crit. If crit is worth less than 0.875 strength to you then matching socket bonuses is worthwhile in this case. If it isn't, then not matching socket bonuses is a gain. At the gear level where one is seriously considering carapace of the core, crit is worth more than 0.875 strength, though Xanthan somewhat overstated his case. There is also the option of gemming expertise/crit in red sockets, which is superior as long as one is able to convert that expertise to mastery elsewhere through reforging, which is not possible in a lot of gear sets.

[hotness]

AMR is giving me stat weights of STR 3.15 and crit 2.8. So the sep will be .889 correct? Is that close to what it should be for a TG warrior around 503 ilvl? I've also run a simcraft, should I trust those numbers? I don't see where it shows a value for exp/hit after soft caps

[Deleted]

AMR is giving me stat weights of STR 3.15 and crit 2.8. So the sep will be .889 correct? Is that close to what it should be for a TG warrior around 503 ilvl? I've also run a simcraft, should I trust those numbers? I don't see where it shows a value for exp/hit after soft caps

Yep that sounds in the right ball-park at least, perhaps a little low. For TG crit is generally around 0.9-1.1 strength, for SMF about 0.85-1.0.

[Pariah]

Lastly, let's just check what other warriors are doing- and not just any warriors. You can see where they are meeting sockets and where they are ignoring. These are all warriors I know personally or else are known for being incredible theorycrafters, or are at the very least good players.

http://us.battle.net/wow/en/characte...ighty/advanced

Freakin' Sinnermighty.

You made the list, sir! You made it!

[Zoopercat]

I loaded all of the characters Darkfriend linked into AMR and SimC. I ran each of them in SimC at 50,000 iterations 3 times: with their own gear, with Mr. Robot's default optimizations, and with Mr. Robot's optimizations with the Crit:Strength ratio suggested by Darkfriend. The DPS margin of error for all of them is between +/- 40 to 50 DPS.

Jalopy:

• His own setup: 176,526 DPS
• Mr. Robot's default setup: 176,910
• Mr. Robot's custom weight optimizations: 177,092
• Result: Mr. Robot's optimizations beat Jalopy, but the custom weights do 0.1% better.

Collision:

• His own setup: 161,598 DPS
• Mr. Robot's default optimizations: 161,695 DPS
• Mr. Robot's custom weight optimizations: Mr. Robot matches Collision's optimizations, no changes.
• Result: Mr. Robot's default optimizations win by 0.06%

Landsoul (In TG Spec, so I didn't run custom weights since the STR:Crit ratio is already 0.88)

• His own setup: 164,600 DPS
• Mr. Robot's default optimizations: 165,459
• Result: Mr. Robot's optimizations win by 0.5%

Tact:

• His own setup: 157,283 DPS
• Mr. Robot's default optimizations: 156,473 DPS
• Mr. Robot's custom weights: MATCH Tact's optimizations, no changes!
• Result: Custom weights win by 0.5%

Varri:

• His own setup: 162,387 DPS
• Mr. Robot's default optimizations: 161,808 DPS
• Mr. Robot's custom weights: MATCH varri's optimizations, no changes!
• Result: Custom weights win by 0.36%

Sinnermighty (TG Spec, so I didn't run custom weights since STR:Crit ratio is already 0.88)

• His own setup: 166,418 DPS
• Mr. Robot's default optimizations: 166,686
• Result: Mr. Robot's optimizations wins by 0.16%

Overall results:

• Edit: I originally said that Mr. Robot's default weights did better in 33% of the cases listed, but I realized 2 of those cases were for A TG warrior where the STR:Crit ratio for the default weights was already 0.88. So I'm editing this to say that of the 4 Fury profiles, Mr. Robot's weights did better in 1 of the 4 cases.
• 50% of the warriors listed match Mr. Robot's optimizations when using Darkfriend's STR:Crit ratio. So either they are using the tool, or Mr. Robot independently matches their own optimizations. So I'd say that's some good evidence that Mr. Robot is doing things right
• When the optimizations differ, we're talking about 0.1% to 0.5%. So it's an exaggeration to say that Mr. Robot is WAY off. It's fair to say that we might need to update weights, which we will definitely look into.
• I will have to test the weights on 480, 490, and 500 iLevel warriors to see which ones end up winning. Right now, 4 Fury profiles isn't enough evidence to change it, but it is leaning towards Darkfriend's weights. We'll run more tests to determine that.

[Darkfriend]

Another issue is that simcraft isn't as accurate for judging warrior DPS, due to how it handles execute phase, but this won't influence stat weight much (if at all). So I'm not sure how accurate using Sim-C to test effective DPS really is. Id have to go line by line on Sim-C to see, since if it's majorly off on the current weight it's going to devalue haste slightly while increasing the value of mastery/crit. (Haste is worth less in execute phase than mastery/crit I believe but by how much, I'm not sure.)

[hotness]

Landsoul has been doing that for a while now. It's not better (generally speaking), but I haven't asked him yet why he still has it. I'll try and find out. Also, his high amounts of haste are simply incidental- there is a lot of haste on his gear, he isn't gemming/reforging into it.

Yeah, but he isn't always reforging out of it either. It's interesting looking at what AMR suggests for that particular case. Going slightly under hit/exp cap (.6/.8) in order to get rid of 3.75 haste in favor of some crit and mastery. Seems possible that could be better eh?

As for wowreforge, I went there and put in my character. There is no stat weight for strength so how do you know what the SEP is they are using?

It's all a little confusing, I suppose the proof is in the pudding though, is your dps higher than a similarly geared warrior? I should probably be studying rotations and making macros or something. And of course killing the boss with good mechanics is more important than 1% dps. It is just too bad I can't optimize what drops! Damn no weapon dropping bosses.

[Darkfriend]

Yeah, but he isn't always reforging out of it either. It's interesting looking at what AMR suggests for that particular case. Going slightly under hit/exp cap (.6/.8) in order to get rid of 3.75 haste in favor of some crit and mastery. Seems possible that could be better eh?

As for wowreforge, I went there and put in my character. There is no stat weight for strength so how do you know what the SEP is they are using?

It's all a little confusing, I suppose the proof is in the pudding though, is your dps higher than a similarly geared warrior? I should probably be studying rotations and making macros or something. And of course killing the boss with good mechanics is more important than 1% dps. It is just too bad I can't optimize what drops! Damn no weapon dropping bosses.

You don't always reforge out of haste because of the massive amounts of hit/exp on a few pieces of gear, meaning it's very often more valuable to get rid of those over reforging out of haste and being over hit or exp cap. Example, the helmet.

[Pariah]

Zoopercat,

Impressive investigation!

[meteo]

Darkfriend I'd suggest that you think a little bit more before you post. It's obvious that you don't like ARM but you can't use other tools when you think that they support your point and if that very tool is being used against you, it's suddenly bad and the information you can get out of it should be ignored...



B2t: I'll stay out of this discussion because it doesn't seem to lead anywhere although the people from ARM tried really hard which earned them a lot of bonus points from me (best tool for quick item comparisons). My only complaint is that ARM calculates after every single change which makes it quite annoying when you want to change multiple things.

[bigbad]

The problem with default weights is that they can't be accurate for everyone. Looked decent when I plugged my char in but it's easy to see cases where it would give inaccurate results. The problem is people don't understand how it works and assume the results are accurate, I much prefer SimC. And if you don't like the actionlist from SimC you can just change it but it's pretty good now.

[Simca]

The problem with default weights is that they can't be accurate for everyone. Looked decent when I plugged my char in but it's easy to see cases where it would give inaccurate results. The problem is people don't understand how it works and assume the results are accurate, I much prefer SimC. And if you don't like the actionlist from SimC you can just change it but it's pretty good now.

The thing is that the site doesn't need to replace SimulationCraft; AMR is a gear profile builder, not a DPS simulator. It works well alongside SimulationCraft. SimulationCraft is really, really bad for comparing small changes, so that's a niche AMR fills great, even for hardcore raiders. There's really no reason to exclusively use one tool.

Also, most of the complaints about SimulationCraft are just bugs, not always actionlist issues. SimulationCraft has great development, but nothing is perfect or sacred and people should always be wary of any results that don't make sense on first glance. On more than one occasion, certain SimulationCraft specs have modeled something accurately but there was an in-game bug which caused it to actually be wrong (and therefore give bad advice). The opposite scenario has also occurred, of course.

[jason1975]

I loaded all of the characters Darkfriend linked into AMR and SimC. I ran each of them in SimC at 50,000 iterations 3 times: with their own gear, with Mr. Robot's default optimizations, and with Mr. Robot's optimizations with the Crit:Strength ratio suggested by Darkfriend.

This, so much this. Even after all the extra effort, we come up with a DPS difference of at best 0.5%. Anyone who understands how these weights are put together should realize this is so far underneath the margin of error that it's not worth discussing. (This includes latency, SimC limitations on rotations, actual sim margin of error, etc).

AMR rocks, it's a great tool, and they do an amazing job on the stat weights (just look at how they've replied to this thread). And if someone has better math than they do, they are incredibly responsive.

I think Darkfriend owes them an apology, frankly.

[Darkfriend]

This, so much this. Even after all the extra effort, we come up with a DPS difference of at best 0.5%. Anyone who understands how these weights are put together should realize this is so far underneath the margin of error that it's not worth discussing. (This includes latency, SimC limitations on rotations, actual sim margin of error, etc).

AMR rocks, it's a great tool, and they do an amazing job on the stat weights (just look at how they've replied to this thread). And if someone has better math than they do, they are incredibly responsive.

I think Darkfriend owes them an apology, frankly.

Nope, not until it's actually a good tool. They are using the inaccurate parts of other tools (simcraft's DPS numbers aren't accurate due to the way it handles the rotation) to try and prove that their own tool is good. Their SEPs are still inaccurate, and until it's better than doing it any other way, I'm going to advise against it's use if a person really wants to min/max. As well, as you will continue to note, skilled warriors are saying it's a flawed tool with a bad rep, while we haven't seen one say that AMR is good. I think that alone speaks volumes.

Honestly, if we were talking about the problems with Sim-C, wowreforge, wowreforgelite, or any of the other numerous tools I'd say to use them where they are accurate, and adjust for their inaccuracies when they are wrong-such as the way sim-c accounts for the execution phase.

The problem with AMR is that it's entirely superfluous. In it's current state it's too inaccurate to use, from the optimizer to the BiS list. If you go through all the hoops of simming your character, and plugging the numbers into AMR, you're just adding an extra step, you're better off using wowreforge.

Essentially, there is no reason to use AMR. I don't hate the devs, anymore than I hate the ones at Icy-Veins (Who I've also been in contact with in regards to mistakes on their site) But I can't honestly say it's a good site, when for the many reasons shown it isn't. The day I can test it out and get equal or better results consistently, for any class, I'll be glad to endorse it. But until it stops telling me to gem full haste on a DK with base stat weights (really?) or use a bad SEP for crit (among other things) I will not say its good.

[Zoopercat]

Darkfriend, your only disagreement is the ratio of Crit:Strength, correct? I used SimC to evaluate your claim that the ratio should be 0.88. I concluded that in 3 of the 4 cases, your suggested values performed BETTER! I didn't use it to prove we were right, I used it to evaluate the stat weights. I said we're going to evaluate this over a larger range of gear cases before we update the values, but it's leaning in the direction you suggested!!

I'd also like to point out that many of these top warriors you keep referring to - Mr. Robot agrees with their optimizations (every single optimization, nothing needs to change), when we plug in the 0.88 Crit:Strength ratio. So even if they say it's bad... Mr. Robot is in agreement with their min/maxing optimizations. So I don't know why they would say Mr. Robot is wrong - that would mean THEY are also wrong.

[Doctor Funkenstein]

Yes, it's a very good and helpful site.

If you're relatively new to optimizing your gear, its default values are all perfectly acceptable. If you're not, you have the freedom to customize your stat weights and adjust them to whatever outcome you personally prefer. The only thing lacking is an inability to simply set what values you want (X% hit, Y% expertise, Z% crit) rather than weight values, but that's a minor quibble more than anything.

The people who are screaming NO!!! are, by and large, simply elitists who hate to admit to themselves that a website is doing all the easy math for them. Instead, they prefer to think that basic arithmetic is an arcane and eldritch science that only they and they alone have the power to understand. How do I know this to be true? Because that's really all askmrrobot does, thus there's no reason to decree that it is the epitome of evil and incompetence. If they received bad results because of it, it was because of their own ineptitude at setting their custom weights, not the simple math the site is doing for them.

Don't let anyone else tell you otherwise. The only possible room for disagreement is the stat weights, and as previously mentioned, you are completely and utterly capable of changing them as you see fit.

[OnyxShadow]

Mr. Robot will be beneficial to about 95% of players. The 5% that can do better already have better utilities or make their own.

[Darkfriend]

Darkfriend, your only disagreement is the ratio of Crit:Strength, correct? I used SimC to evaluate your claim that the ratio should be 0.88. I concluded that in 3 of the 4 cases, your suggested values performed BETTER! I didn't use it to prove we were right, I used it to evaluate the stat weights. I said we're going to evaluate this over a larger range of gear cases before we update the values, but it's leaning in the direction you suggested!!

I'd also like to point out that many of these top warriors you keep referring to - Mr. Robot agrees with their optimizations (every single optimization, nothing needs to change), when we plug in the 0.88 Crit:Strength ratio. So even if they say it's bad... Mr. Robot is in agreement with their min/maxing optimizations. So I don't know why they would say Mr. Robot is wrong - that would mean THEY are also wrong.

Yes, for the optimizer that was my main issue.

[Noctus78]

I used SimC to evaluate your claim that the ratio should be 0.88. I concluded that in 3 of the 4 cases, your suggested values performed BETTER! I didn't use it to prove we were right, I used it to evaluate the stat weights. I said we're going to evaluate this over a larger range of gear cases before we update the values, but it's leaning in the direction you suggested!!

Pretty cool that this thread will lead to perhaps an update, even initiated by Darkfriend and how Zoopercat picks it up
Thanks both for that matter, well done.

Darkfriend's opinion is brought a bit (too) strong and perhaps not looking at the whole picture, but if you read between his lines and skip some others, it aint that bad. like when he says "I'm going to advise against it's use if a person really wants to min/max."
For a large group (at least my warrior p.o.v.) AMR will do the job fine enough to be not that far of. The real top top (and i envy their knowledge) might do slightly better and is perhaps also needed in their guilds to beat HC modes asap.