Im confused on Results as Mage (TBC)

Character specific
1

Edit: I was able to find export/snapshot ID: 571fe1b9a104489fa159efd2bfd58fc1

So for an Arcane Mage with Totem of Wrath, you only need 3% spell hit from gear, so I set my hit threshold to 3%. However, the recommended set is still showing 6.49% hit. At first, I assumed this was simply because my best available pieces already have hit baked in and I couldn’t realistically get any lower.

But when I started looking more closely at the recommendations, things stopped making sense. For example, it’s recommending Mantle of Three Terrors, and listing the next two best options as ZG class shoulders and Arcanist (T1) shoulders — even though I have significantly stronger DPS shoulders available with no hit, such as Mana-Etched Spaulders, Evoker’s Silk Amice, Destroyer’s Mantle, etc.

Because of that, I’m pretty confused about how Mr. Robot is arriving at these recommendations.

I would paste an export link, but I can’t find a way to export my setup. I don’t see the “Send to Addon” option that Google suggests. Instead, I’ll explain my process and include a screen recording:
https://imgur.com/a/6fHsfBH

Like I mentioned, I initially set my spell hit threshold to 3%, yet the optimizer returned a setup with 6.49% hit. My first thought was to figure out where I could safely remove excess hit.

I also didn’t have Glyph of Power selected because I wanted to use the ZG head enchant instead. So I excluded Glyph of Power and re-optimized my gear.

After doing that, the optimizer changed my wand from Rod of Dire Shadow to Nether Core Control Rod, presumably because I lost some hit from switching to the weaker head enchant. However, I don’t understand why this matters when my hit threshold is only 3% and I’m already well above it.

Next, I excluded Nether Core Control Rod and optimized again, which caused Rod of Dire Shadow to be selected once more.

I then looked at my shoulders to see where I could trade hit for stronger DPS stats. The optimizer still recommended Mantle of Three Terrors, which is a hit piece. What confused me even more was that it ranked Arcanist Mantle and ZG shoulders above clearly stronger DPS options like Mana-Etched, Destroyer’s Mantle, and Evoker’s Silk Amice.

At this point, I’m struggling to understand how Mr. Robot is calculating these decisions, because the results don’t seem logical.

Another thing that confused me is gemming. I’ve always been told that Arcane Mages generally gem Intellect or Spell Power, yet Mr. Robot is recommending Spell Power gems in red sockets and Spell Power/Crit gems in yellow sockets, with no pure Intellect gems at all.

Another confusion is why would it recommend for shattrath jumpers to use one 9sp gem and then one 5sp + 6stam gem.. those gems dont even give a socket bonus? Why randomly suggest a stam gem?

Also shoulder inscriptions dont seem to be an option on shoulder enchants

2

The optimizer handles all of this for you already. The optimizer knows what the hit cap is, and will work to get you as close to it as makes sense, without going over (within the constraints of the gear you have available). That custom setting is there for the following case: sometimes it is actually optimal to remain under the hit cap, but, a lot of classic players really do not like that. This setting allows you to force a minimum amount of hit you want on your gear, so you can poke the optimizer to get more even if the calculated score would be slightly less. In your case, you do not need this setting, so I recommend disabling it, as the optimizer is already getting you to the hit cap.

As far as trying to napkin math each decision the optimizer makes… it’s good to do a sanity check, but you aren’t going to be able to figure out exactly why it makes each specific decision. You have approximately 1 trillion possible combinations of gear in your bag. If you change one item, that could potentially affect all the other items you want. The whole point of our website is to generate a custom solution for the gear you have - because it is extremely difficult to do so manually.

The optimizer is very good at finding the most efficient way to get you as close to the hit cap as possible without wasting stats.

Also, the per-slot rankings of items should be viewed as an estimate. To give a “true” ranking to each item in the list, we would have to calculate a full best in bag solution with each item locked in. That would be very expensive and slow. It is common, as you have pointed out, for an item in that list to be ranked higher than the item picked in the best in bag solution. But, if you lock that item in, you will almost always see the total score go down compared to what the optimizer picks. (There are rare cases where you can manually lock in an item and increase the score slightly. We do have to allow a little bit of error margin to keep the website fast).

For the inscriptions, you need to pick your faction in your character settings. I’ve seen another question about this, which probably means we need to make this option less hidden on the UI. Select the “hamburger” menu next to your character name and pick your faction and professions. This will make sure you can see all the options you want to see in the optimizer.

Could you post a snapshot where the optimizer puts a sp/stam gem in the shattrath jumpers? I was not able to recreate that result using the steps you described here.

I appreciate the feedback. I hope my response here sheds some light on how the optimizer is working. I believe in this case that the optimizer is picking the correct gear for your character. Once you know that the optimizer automatically handles spell hit optimally, I think that clears up a lot of the questions. The hit cap and swapping items around that is one of the main difficulties/constraints the optimizer is particularly good at and tuned for. The hit you see in the Char Sheet on the right includes any source of hit that applies to all spells, like totem of wrath. If you mouse over it, you see that you have 41 hit rating on the suggested solution, which is just a tad over 3% on your gear.

3

Okay thanks for clearing up the hit and threshold issue.. but im still confused as to why theyre recommending certain things above others. Also why if why i have totem of wrath my recommended set has 8.5+ hit.. and certain items they suggest have hit. You would think items with more dps stats would be better when ur over hit cap by so much

Anyway heres a SS of gem https://imgur.com/a/lUQASpg

here is “send to addon” export if that helps too

bib
$50;US;Dreamscythe;Zèrø;Bloodlust;10;2;70;4:375,10:375;1;.s1;2500050310130150330125000000000000000000000000533003013020010000000;.q1;21709s11t74e2928;5799s10t49e9;288s3t51x24030y0e-216;185s15t78e-57;153s2t79;40s9t57x0e-2295;5s8t80x0y26e2280;50s5t60x3y0z-29e12;157s18t82;32s13t63;147s6t64;21s1t83e-73;17s17t84;182s14t85;370s16t86e81;130s12t89e259;1247s7t68x0y0z0e-180$@MageArcane1@Arcane@e\2928\0\27924\12 SpellPower\22449=2,22446=2@e\2937\0\33997\20 SpellPower\22446=6,22449=6,22457=6@e\2721\0\29467\14 CriticalStrikeSpell, 15 SpellPower@ee\2664\0\27962\7 ArcaneResistance, 7 FireResistance, 7 FrostResistance, 7 HolyResistance, 7 NatureResistance, 7 ShadowResistance\22446=4,21884=4,22451=4,22452=4,21885=4@e\369\0\34001\12 Intellect\22447=3@e\2649\0\27950\12 Stamina\22445=12@e\2661\0\27960\6 Strength, 6 Agility, 6 Intellect, 6 Stamina, 6 Spirit\22449=4,22445=4,22446=4@e\2588\0\24164\7 HitSpell, 18 SpellPowe@e@e\2669\0\27975\40 SpellPower\22449=8,22446=8@e\2748\0\31372\20 Stamina, 35 SpellPow@er@e\2935\0\33994\15 HitSpell\22446=8,22445=2,22449=2
$50;US;Dreamscythe;Zèrø;Bloodlust;10;2;70;4:375,10:375;1;.s1;2300050300000000000000000000000000000000000000535023313035013041051;.q1;21709s11t74e2928;5799s10t49e9;288s3t51x24030y0e-216;185s15t78e-57;153s2t79;40s9t57x0e-2295;5s8t80x0y26e2280;53s5t61x7811y0z0;154s18t82;32s14t63;168s1t83e-61;17s17t84;182s13t85;370s16t86e81;86s6t88;44s12t89e259;1247s7t68x-7837y0z0e-180$@MageFrost1@Frost@e\2928\0\27924\12 SpellPower\22449=2,22446=2@e\2937\0\33997\20 SpellPower\22446=6,22449=6,22457=6@e\2721\0\29467\14 CriticalStrikeSpell, 15 SpellPo@eer@e\2664\0\27962\7 ArcaneResistance, 7 FireResistance, 7 FrostResistance, 7 HolyResistance, 7 NatureResistance, 7 ShadowResistance\22446=4,21884=4,22451=4,22452=4,21885=4@e\369\0\34001\12 Intellect\22447=3@e\2649\0\27950\12 Stamina\22445=12@e\2588\0\24164\7 HitSpell, 18 SpellP@ewer@e\2669\0\27975\40 SpellPower\22449=8,22446=8@e\2748\0\31372\20 Stamina, 35 Spell@eower@e\929\0\13945\7 Stamina\11176=5@e\2841\0\44968\10 Stamina\

it happens when i optimize - then go to shoulder enchant and exlcude scourge naxx enchnt and reoptimize. The gem now says 5sp + 6stam on boots, when originally it was 8int gem i believe.

4

See, here’s another example of something I can’t quite figure out why or how it’s happening. I’m sorry for all the questions, but when something looks off like this it just makes me not trust any of the results.

For example, I blocked the Vitality enchant because I don’t want to pay for it. So I re-optimized to see what the new set would be, and it tells me to put the Fortitude enchant on my boots instead of Vitality, and then change my necklace from Torc to Brooch.

So how did I lose about 13 spell power? Nothing else seems to have changed except those two things, and neither enchant has spell power. In fact, the new neck it tells me to equip actually has 3 more spell power than the previous one.

So where or how did I lose ~13 spell power? https://imgur.com/a/XA4Kgsd

146fe77f342b41b393a5292a84e02dd7

5

Screenshots/videos are difficult to use for debugging.

The best way for me to comment on why the stats of two setups are different would be to make a snapshot and copy the ID here of each case, then I can look at the two directly and compare them.

6

Okay, so here is a fresh example that I think represents my confusion/frustrations pretty well.

Step 1:
I did my initial optimization.
7098b30e24b3456c9365a43d92a128ea

Step 2:
I then made some slight edits afterwards, like removing all shoulder enchants I don’t have access to so it would recommend the Scryer Honored – Inscription of the Orb instead. I then re-optimized. There were no changes except switching to that enchant.
e8b50966ef214986bd4a5cc195630f1d

Step 3:
Next, I excluded Glyph of Power because I didn’t want to buy it and I already had a ZG head enchant available in the bank.
c0d4c49094e44f0c9836e5a2c1d32aeb

At this point, no major gear changes had been made.

Step 4:
My next edit was to exclude the + 15 spellpower bracer enchant because it’s stupidly expensive right now (like 400g in mats). After doing that, the results told me to change my shoulders from Mantle of Three Terrors to Mana-Etched. 2c1b8c863d8c4ef4853247c1bef672fa

This one kind of makes sense to me because with the bracer enchant change I’m losing 15 SP and gaining 12 Int instead. With the new shoulders that have gem slots, I’m effectively gaining that spell power back.

Now my last edit is the one that really doesn’t make sense to me.

The Incanter’s Robe chest they had recommended this whole time said to use:

  • 1 × +7 SP gem

  • 2 × +4 Int / +3 Crit gems

But my robe in-game already has:

  • 1 × +7 SP gem

  • 2 × +6 Int gems

I didn’t want to spend gold to re-gem them, so I manually edited the yellow sockets to use +6 Int gems instead.

0cd02d4473984fd8a50d1a0110f7250d
(snapshot after changing gems)

After re-optimizing, the results now tell me to: 95f04f75ce4642a8b106ba99332a2fde

  • Change my weapon enchant from +40 SP → +30 Intellect

  • Change my boot enchant from +12 Stamina (Fortitude) → Vitality (MP5)

On top of that, it is now weirdly telling me to change the gems in my Mana-Etched shoulders from:

  • 2 × +7 SP gems

to:

  • 1 × +6 Int

  • 1 × +5 SP ???

It is also telling me to change my Baba’s Cloak of Arcanistry to Sethekk Oracle Cloak.

All of these changes make no sense to me.

All I effectively did was replace two gems worth 8 Int + 6 Crit with two gems worth 12 Int, and it suddenly tells me to do all of that?

I added extra Int and only lost a bit of Crit by changing gems, so why would it suddenly want me to:

  • swap a 40 SP enchant → 30 Int

  • change one of my shoulder SP gems → Int

  • and the weirdest part of it all — change one of my 7 SP gems to a 5 SP gem?

It also tells me to change my cloak from 22 SP + 14 Crit to 22 SP + 12 Hit, when I’m already way over hit cap anyway (8%+).

So it just doesn’t make sense.

And on top of that, if you compare that final snapshot to my first one before I made any edits, it actually shows a higher % DPS increase.

So why wouldn’t it have just recommended that setup to begin with?

I’ve noticed that happening a bunch when making certain edits — that my modified versions show higher % DPS increases, and it makes me wonder why the website didn’t recommend them itself in the first place.

7

First, it is worth zooming out here and noting that all of these changes make very small changes to the total score. Everything happening here is less than a 1% change, with some of the changes being even less than that.

As I mentioned before, the optimizer does not do a brute-force exhaustive search, as that would take a long time and cost a lot of compute for negligible gain. There is a small margin of error as the optimization algorithm does its magic. Gems and enchants are by far the most difficult and complex optimization problem to solve. Each time you specifically exclude something or lock something in, you change the starting conditions and constraints, so you will get a slightly different result.

If you lock in specific gems, you effectively reduce the problem, so sometimes the optimizer can find a slightly better score overall.

The one gem being down-graded to 5 SP is probably a small flaw with the algorithm. You have your gem+enchant threshold set to 0.2%. That setting can complicate the optimization, and it appears it is having a little trouble in this case. If I remove the threshold, that particular issue goes away. We will see if we can tweak the code to stop that from happening when the threshold is set. With the level of scrutiny you are approaching this, you may want to remove the threshold as you tinker, especially if you are going to manually adjust gems anyway.

As for the cloak, keep in mind that hit has some value above 6%. You filler is frost bolt, which does not benefit from Arcane Focus. So, hit is less valuable above 6%, but not worthless. You are optimizing for a 180s fight in fairly low level gear. If you execute your rotation for 3 minutes straight in your gear, you will have to cast a lot of frost bolts. I highly recommend that arcane mages reduce the fight length to something much shorter if you want to optimize for your non-filler damage. As it is right now, the optimizer is probably finding that you are spending a LOT of time casting frost bolt.

To me, this looks like your character is right on some threshold where it can almost cast more spells or replace less arcane blast with filler frost bolts. Once you lock in a little more intellect, it’s seeing that if it got just a little more mana, it could find a tad more score. Increase the fight length by 10 seconds and you will see the suggestions chang on the boot enchant.

This is an unavoidable side effect of trying to do gear optimization down to a very small margin of error. Will you actually see a difference in game with these changes? Probably not, as our calculations assume a completely idealized fight where you never have to stop casting spells.

If the only thing we had to do was optimize gems and enchants (which is what the website started as way back in the day), we could get the “perfect” solution every time. But we have to optimize gems/enchants AND all the pieces of gear. Something has to give, and the total amount you can swing your score with gem/enchant optimization is small compared to gear, so we allow a little fluff there. In this case, locking some pure intellect gems into the incanter’s robe is enough of a “hint” to help the optimizer find that extra 0.5% dps or so.

8

I understand that every exclusion and change basically resets the search and determines new parameters, but some of the decisions don’t quite make sense to me.

Like I mentioned, the first 2–3 changes I made didn’t change much or tell me to swap any gear. When I excluded the +15 spell power enchant, it finally told me to change my shoulders, where the new ones had two gem slots that it recommended gemming for spell power. This makes sense to me because the whole time you can tell the optimizer is favoring spell power. I’m losing 15 spell power from the enchant, so it suggests changing shoulders to get that spell power back — that makes sense.

But then when I lock in two gems to give me more intellect and lose just a small amount of crit rating, it suddenly tells me to basically abandon all my spell power enchants. That’s the part that doesn’t really make sense to me.

The whole time, step by step, it seems to be favoring spell power. I lose some from an enchant and it wants me to get some back from a shoulder swap — that tracks. But then as soon as I change two gems — where the change only gives me +4 total intellect and -6 total crit rating — now all of a sudden it wants me to remove a bunch of spell power enchants from other places?

I would understand it if it were the other way around. For example, if the gem changes I made caused me to lose intellect, then it recommending changes to gain that intellect back would make sense.

Also, the gems it’ was recommending — the +4 Intellect / +3 Crit Rating ones — I can’t even seem to find anywhere in game or online.

9

Is the gem this one:
https://www.wowhead.com/tbc/item=23101/potent-flame-spessarite#taught-by-item

If it’s not actually obtainable in-game we can remove it from the list.

10

We will look into whether those gems are real or not - there is often data in the game client that is unused. It is difficult to determine sometimes if they are actually in the game.

Regarding the choices - I will refer you a bit to my previous post. We are doing a straight calculation over the specified fight length. Depending on your gear, it can be in a state where you are right in between two different types of setup that score very close.

In reality, there are at least hundreds, usually many thousands of solutions within 1%, even 0.5% of the theoretically brute-force optimal score. Those solutions can have different stats. Manually tinkering can sometimes flip you between different ones. It seems like it is doing something “weird” - but really it is not. The total score is changing very little.

Set the gem/enchant threshold to 1% and all of these little things go away. It picks the intellect gems and sticks with what you want, I think (without even needing to lock in the intellect gems). I find that setting to work best disabled or set to 1%.

It is difficult to explain, logically and mathematically. This is all a side effect of there being SO MANY sets of gear well within 1% of the theoretical most optimal solution. We simply cannot prevent all seemingly weird swapping. What we can tell you for sure is this: if you use the gear we suggest, you will be good to go. Realistically, no source of information for WoW can tell you which of two sets of gear that are 1% different in a simulator or mathematical model are “actually” better. If we had it our way, we’d show a “cloud” of gear choices that all fit within that 1% range and let you pick one. But, over the years, we have found that people prefer a deterministic solution. Same input = same output every time.

11

Okay thanks.. i was able to find the gem.. i couldnt find the name on the sim after it suggested different ones and when searching google for gems with those stats and looking at gem lists… none was coming up. It is in the game though.

Is there anyway to include unique gems, but only suggest ones you have in your bags/bank? I have a bunch that I saved and won, and wondering if they might be better then non unique gems they recommend. But when i turn include unique gems on, it just recommends ones i dont have yet.

12

We don’t have customization that granular. You would be surprised at how difficult the unique gems are to deal with in the code. You could lock the unique gems you have into a couple of items that will likely be picked, and have the optimizer work around them.

13

I’m going to jump in here with some of my own odd findings in the mage optimizer.

Snapshot: 658e4d1579b14fc3984bbf8dbd0fa4b7

Issue 1: It’s suggesting “Runed Blood Garnet”, but when I lock in Brilliant Golden Draenite into every gem slot instead, the score goes from +4.16% to +5.82% (and doesn’t make any other item swaps)

Issue 2: It’s ignoring my vanilla trinkets. I have Mind Quickening Gem and Hazza’rah’s Charm of Magic (spell haste for 20 seconds and 200 arcane spell power for 20 seconds, respectively) but is instead recommending two trinkets that only have 28 spell crit rating. It doesn’t seem to matter how long the fight is, even if I set the fight length to 20 or 30 seconds it ignores those trinkets. 200 spell power should be pretty clearly better than the crit trinket.

14

We don’t include a lot of the trinket estimate for old trinkets, they become quickly irrelevant. Even Kara gear will make those not worth it.

This is another case of a highly constrained optimization. Locking in one intellect gem hints the optimizer to change your boots. I’m not sure why it is getting stuck in this case. The gem code is enormously complex at this point. As more end game gear becomes available to you tend to see less difference.

Lowering the fight length, to, say 45 seconds. You will see there is almost no difference between the int/sp gems. Your settings here are right on some sort of breakpoint. If I ever have some extra time, putting in some fight length smoothing could be worth it to avoid these cases.

That being said - when you play the game, you have no idea what the fight length will be, really. Maybe you would know within… 20 seconds? 10 maybe with the same group every week? As you change the fight length, you can see things move around. That’s how the game actually works. Stressing about a potential 1% theoretical difference when the fight ending 5-10 seconds earlier/longer would have a similar effect isn’t good for the soul.

15

I have a question.. say i wanted to find my best max SP/crit set for dungeons that dont really require hit.. how can i do that? I set hit threshold at 0 but doesnt seem to really do what im looking for

16

Those thresholds force a minimum of a stat. We don’t have an option right now to force a different cap/max of a stat or lower the level of the enemies you are fighting.

We didn’t go there for classic because optimizing gear for dungeons seemed pretty low value. Having that extra hit on your gear won’t make you lose that much damage in a dungeon setting.

17

Also, another thing i noticed is it seems the optimizer really looks to complete socket bonus’s that arent so important. Like on spellfire gear.. bonus is stamina.. it will put a 5sp/6stam gem in there instead of intellect gems. It seems the sim is always looking to complete a socket bonus, even though its not important. Because ill then take the optimized set it gives me… go plug it in to a “real sim” and just change the gems to all int or w.e the case is an get greatly higher results. Not sure if im missing a “ignore socket bonus” button somewhere or what

18

I would have to see your snapshot on that - the optimizer gives no value to stamina for mages, so it is not actively trying to get stamina socket bonuses unless doing so would make almost no change to your final score.

When you say the “real sim” gives you “greatly higher results” with intellect gems, what is the magnitude? Changing gems from what we suggest to intellect gems could not change your dps by more than 1%, maybe 2% on the outside.

As we have discussed, this arcane mage case is falling on some sort of edge case where the optimizer can’t “get to” a solution with some specific items, resulting in a slightly higher margin of error than usual. We will see if we can throw a hint at the algorithm to try these intellect gems.

All that being said, we have always been very up front about what we are doing. We intentionally sacrifice a little bit of accuracy for speed and convenience. That is part of what sets our site apart from using an open source simulator. We are examining ALL the possible sets of gear you could create an picking one for you. It has to be fast. Often, our technique will find good sets of gear that you would not have though to try. We think this is great, but often leads to people thinking the website is “wrong”.

Any theoretical model has error. People do not play the game like a simulator simulates it or like a mathematical model assumes they play. You have to move sometimes. Sometimes you wait a few seconds to do damage. The fight might end a little sooner or later than you expected. You might get lucky procs. We could keep tweaking the optimization algorithm forever to try to get closer to that one set our model things is BiS without getting too slow. But… why? We already know that we simply cannot predict actual performance within 1% (with a model OR a “real simulator”), why would we then worry about a 1% margin of error in the optimization algorithm?

The classic version of the site is “jumpier” because it only looks at one exact fight length. When you do a simulation with an open source monte carlo simulator, they will typically vary the fight length by +/- 20% on each iteration and then average the results together. That helps with the inherent noise in simulation and averages out some of the changes in dps due to fight length. We do something similar to this with the simulator we use for the retail version of the website. But, for classic, we don’t have the time/resources to develop a full simulator. We also can’t use the open source simulator because it is too slow. So, we settled on a very good mathematical model that is very fast.

19

Yeah I know and i appreciate the feedback. Its really my self diagnosed OCD getting the best of me haha. I use sims but it takes so long to really manually do all the different combinations possible and quite frankly not realistic.. so I use it in combination with your website. Which like you said, kind of gives you that starting foundation quickly. It just really throws my head for a spin when I plug AMR into sims and it sims lower than another set. Then ill make some tweaks to AMR’s recommendations that show a higher %DPS on AMR, edit it on the normal sim.. and its a DPS loss. So just alot of back and forth conflictions that drive me crazy sometimes.

Have you guys ever thought of showing DPS results in addtion to just DPS%? Also, im not seeing improved judgement of crusader on the debuff list?

20

It’s under the raid buffs, Improved Seal of the Crusader