Fringe Frost Theory: Crit, PBoI, & You

[Akraen 230]

A couple days ago I made a big post about the value of crit as PBoI impacts our spells and crit damage. This is absolutely complicated. It's also exploratory and meant to be taken as a mental activity.

 

A huge disclaimer: do not just wildly follow what this post says. Until this is verified and conclusive, I still recommend the "safe" 14242 haste > mastery strategy of gearing. However, if you'd like to test this out and have a PBoI, please reply to this thread with your logs so I can take a look.

 

Frost mage analysis has been a huge undertaking for me this expansion. I generally do it without much help (though big thanks to Lhivera for providing me with all the spell calculations so that I didn't mess that up!). As a result, I can only work on things as I have time and I'm subject to tons of human/education error. So if you notice errors, please nicely point them out and help me make this as accurate as possible. I'm not betting on a horse in this race, it's only the truth I seek!

 

Okay, so now on to the details, which I'll explain much more indepth this time because there's a lot here.

 

I will begin by explaining why I thought of this: I noticed a lot of the top-parsing frost mages were formerly fire or arcane, but having an offspec week. Therefore, their gear wasn't very well-itemized for frost-- or so I thought. Their numbers were very icicle-heavy and even if they didn't play very well (low uptimes, etc), they still did well.

 

I know what the community must be saying, "Akraen, you told us haste!" - Yeah, but I am not going to try to push something that is suboptimal, if it turns out PBoI boosts mastery/crit's relationship so much then of course I need to revise my advice. Everything in my Advanced guide holds true if you do not have a PBoI!

 

Okay, so here is where I started, 7 possible builds all based off AMR profiles using gear available to me (I have no HWF BBoY, so I used HWF KTT in each build to be consistent). After that, I ran calculations using Lhivera's TC-Lite to find the DPET adjusted by PBoI in Single, 2, 3, 5, and 10 target scenarios. This is just DPET though, not the whole story. Haste clearly adds the most DPET, but is that the whole story?

 

 

Let's look at each ability more visually:

 

Okay, now that the above info is digested for 580 PBoI, what about 567 and 574? Below is an extension of the original calculations that will detail the difference among PBoI item level is not that significant:

 

Celestalon and Muphrid very rightly pointed out that even though it is useful to look at DPET, it doesn't tell the whole story because you need to factor in ET%, which is basically the time you spend using each ability. The thinking here is that even though you may get a ridiculous amount of DPET out of Living Bomb with haste, you can only use the ability once every 12 sec, or possibly snapshotting it within 6 sec-- for the below ET calculations I attempted to provide a reasonable scenario for how often one might snapshot-- I double checked my numbers against T16H patchwerk sim (don't yell at me, it made sense) - just to see how much time was spent on each ability. Turns out my ET%s are very similar to the simcraft build, meaning they at least aren't unreasonable:

 

Now that we have ET%, we can take the DPET values and multiply them out to find out which build appears to be the strongest.

Finally, this is the visual representation of the above combination of DPET & ET%, resulting in numbers that resemble DPS:

 

 

End-Disclaimer: This is a work in progress. If someone out there wants to help me with this project, please let me know. I'll gladly email you the excel workbook.

[Blackd 8]

i commented yesterday on your other tread and was looking for it to see any reply haha

Anyway i tried to stack crit with 561 PBoI and 566 ilvl overrall. Conclusion: dont do it. Its gonna save u alot of cash from reforging trust me =)

[Akraen 230]

i commented yesterday on your other tread and was looking for it to see any reply haha

Anyway i tried to stack crit with 561 PBoI and 566 ilvl overrall. Conclusion: dont do it. Its gonna save u alot of cash from reforging trust me =)

Hi - no offense to you directly, but this isn't how to really contribute to the thread.

 

Nothing about this thread is "do this" or "don't do this" - it's theorycrafting. Think Elitist Jerks. I'll be moderating this thread heavily as opinions have no use to the OP.

 

When testing this - do not forget that crit implies RNG. Theoretical numbers might not be seen in a short fight, or even a long fight!

[Kver 65]

And that is exactly my primary concern with this: crit's practical behaviour is incomparable to crit's theoretical behaviour. In your calculations you(naturally) assume that crit%*crit dmg% is the increase. In practise, this will never be the case; especially at lower gear levels people will notice a large DPS range which fluctuates regularly.

 

What are your thoughts on mage armours? It would occur that in a BiS M>C>H build frost armour would be a bit weak. Do you reckon burst haste and its multiplicative nature still make it end on top?

[Akraen 230]

And that is exactly my primary concern with this: crit's practical behaviour is incomparable to crit's theoretical behaviour. In your calculations you(naturally) assume that crit%*crit dmg% is the increase. In practise, this will never be the case; especially at lower gear levels people will notice a large DPS range which fluctuates regularly.

 

What are your thoughts on mage armours? It would occur that in a BiS M>C>H build frost armour would be a bit weak. Do you reckon burst haste and its multiplicative nature still make it end on top?

 

I'm inclined to still recommend Frost Armor because you need it in order to get to the 9522 LB breakpoint. However, if it ends up that the 9522 LB breakpoint isn't worth it, then Mage Armor would take the lead.

[Narsha 12]

I have reforged/regemmed and doing Flex or something tonight to test it out and post logs, hope more do the same just cause interesting!

[Akraen 230]

Scratch that - Mage Armor wins, here's my new profile which you can look at the stat weights:

 

http://www.askmrrobot.com/wow/gear/75b3a04b-6079-4fab-80ee-8c327f8b8cc9

[Methusula 17]

This seems like it will have the added side benefit of being even better for on demand burst. I'm progressing on H Siegecrafter atm and that could be quite valuable.

I'm running standard mix build right now but I have a 580 PBoI and 577 item level with slightly different gear than you Akraen, so I will log our farm kills tomorrow with the M>C>H setup and post results for analysis.

[Kver 65]

Scratch that - Mage Armor wins, here's my new profile which you can look at the stat weights:

 

http://www.askmrrobot.com/wow/gear/75b3a04b-6079-4fab-80ee-8c327f8b8cc9

Thought as much. People tend to forget that Mage Armor actually profits from the PBoI effect as well. With lower levels of haste in particular, Frost Armor is pretty meh, especially when you value mastery higher than haste.

[Frostedmages 3]

I included this in a PM to Akraen, but I'll add it here to gather some opinions.

 

If the goal here is to begin collecting a good number of logs on different fights to start trying to draw conclusions, I feel like we should have some controls on the types of logs that are consolidated. The better we can control for various in game effects, the better we can analyze the results without confounding variables mucking up the interpretation.

Some initial things that come to mind:

 

Flasks? Pots (1 or 2)? Food? All Raid Buffs? Mastery / crit / haste rating ranges for each spec. Active Racial usage? Frost/Mage Armor? 10/25man? ilvl?

Controlling for these ahead of time will make the analysis downstream much more rigorous and allow for stronger conclusions; especially if we're dealing with small (but real) changes in DPS.

 

Alternatively, people can simply post the conditions under which their log was gathered and then like conditions can be consolidated.

Edited March 17, 2014 by Frostedmages

[Kver 65]

I included this in a PM to Akraen, but I'll add it here to gather some opinions.

 

If the goal here is to begin collecting a good number of logs on different fights to start trying to draw conclusions, I feel like we should have some controls on the types of logs that are consolidated. The better we can control for various in game effects, the better we can analyze the results without confounding variables mucking up the interpretation.

Some initial things that come to mind:

 

Flasks? Pots (1 or 2)? Food? All Raid Buffs? Mastery / crit / haste rating ranges for each spec. Active Racial usage? Frost/Mage Armor? 10/25man? ilvl?

Controlling for these ahead of time will make the analysis downstream much more rigorous and allow for stronger conclusions; especially if we're dealing with small (but real) changes in DPS.

 

Alternatively, people can simply post the conditions under which their log was gathered and then like conditions can be consolidated.

Do you normally theorycraft in unbuffed situations? I also believe Akraen has clearly specified which build uses what, though it would be nice if he could include it in this topic, since this seems to be a chapter of a series. These calculations as far as I know use Frost Armour, since I was the one who suggested Mage Armour.

 

Flasks and potions don't really matter in a theoretical setting where you use base values in order to figure out initial values. What you are supposed to do is read those calculations and values, and then draw your own conclusion. Think about the effect a trinket proc has on each individual build.

 

It's not statistical research, it's theory.

 

Edit: Not trying to fight you for suggesting proving this with logs. I just feel that you can't prove it with logs, just like barely anything has been proven with logs when it comes to spec/build efficiency. This should stay in the realm of theory for the advanced, at least until it's finished.

[Frostedmages 3]

I could have sworn somewhere in the post he was saying it needed to be verified and for people to post their logs to go through?

 

Theory is the starting place, yeah? Application to raiding scenarios is where the theory is tested and verified.

 

If that's the case, then taking every variable into account is the best way to approach the situation, as trivial as some of those things may be.

Edited March 18, 2014 by Frostedmages

[Frostedmages 3]

http://www.warcraftlogs.com/reports/RZHkCNcjxG3LgTqf

 

M > C > H in Flex 1, no Immerseus.

 

No flask/potions used. 275 food. all buffs, no active racials. Mage Armor. ilvl 577. Mastery: 22787, Crit: 10971, Haste: 4049. 574 PBoI.

 

The dude doing the pulls here sucked and just pulled randomly; so I never go a proper opener. Oh well. (I also derped and use NT on Sha, don't tell anybody!)

 

I've gotten the OK from my guild to start testing this stuff in our runs, so I'll have more organized stuff soon.

 

Some notes from the experience:

 

The FB cast time you calculated/expected was much faster than mine when using this build. I ended up having a 1.7s fb cast time when using mage armor. I tested it with Frost Armor and it only went down to 1.6.

 

Im curious if changing your fb cast times changes this analysis much. When thinking about it, ET% of frostbolt is going to decrease quite a bit as you weight crit/mastery higher due to the loss of haste. While increasing crit/mastery means every Frostbolt is worth more, it also means there is some point where you are losing enough frostbolts to make it so the gains in damage per frostbolt are not worth this loss in overall frostbolt numbers.

 

Could you re-do your analysis, using a set time period and altering the ratio of spells used in them to something that reflects changes which occur when you are jumping from extremely fast frostbolt (haste capped or 14242 haste) to very slow frostbolt (M>C>H) and see how the altering of spell ratios changes the overall damage outcome?

 

Once it's not midnight I may spend some time on a test dummy with full raid buffs and see what the real change in the number of frostbolts over a given period of time is with M>C>H compared to standard frost.

Edited March 18, 2014 by Frostedmages

[Kver 65]

I could have sworn somewhere in the post he was saying it needed to be verified and for people to post their logs to go through?

 

Theory is the starting place, yeah? Application to raiding scenarios is where the theory is tested and verified.

 

If that's the case, then taking every variable into account is the best way to approach the situation, as trivial as some of those things may be.

Except you don't want that before the theory is even finished, especially not when theory is on a level like this. It completely changes the way we play Frost Mage. There is even a level of crit/mastery where you simply ignore ice lance altogether. I understand you want this proven, but I also know that people read these topics because they've got the name 'Akraen' stuck to them, just like they believe anything Vykina says. You've already seen people ask "crit gud now?". 

 

I swear, the moment you actually start to take those things into account and start "proving" things this place will explode.

[Oltier 471]

Except you don't want that before the theory is even finished, especially not when theory is on a level like this. It completely changes the way we play Frost Mage. There is even a level of crit/mastery where you simply ignore ice lance altogether. I understand you want this proven, but I also know that people read these topics because they've got the name 'Akraen' stuck to them, just like they believe anything Vykina says. You've already seen people ask "crit gud now?". 

 

I swear, the moment you actually start to take those things into account and start "proving" things this place will explode.

I think Akraen should create a shadow profile and post his theories with that :P

[Akraen 230]

http://www.warcraftlogs.com/reports/RZHkCNcjxG3LgTqf

 

M > C > H in Flex 1, no Immerseus.

 

No flask/potions used. 275 food. all buffs, no active racials. Mage Armor. ilvl 577. Mastery: 22787, Crit: 10971, Haste: 4049. 574 PBoI.

 

The dude doing the pulls here sucked and just pulled randomly; so I never go a proper opener. Oh well. (I also derped and use NT on Sha, don't tell anybody!)

 

I've gotten the OK from my guild to start testing this stuff in our runs, so I'll have more organized stuff soon.

 

Some notes from the experience:

 

The FB cast time you calculated/expected was much faster than mine when using this build. I ended up having a 1.7s fb cast time when using mage armor. I tested it with Frost Armor and it only went down to 1.6.

 

Im curious if changing your fb cast times changes this analysis much. When thinking about it, ET% of frostbolt is going to decrease quite a bit as you weight crit/mastery higher due to the loss of haste. While increasing crit/mastery means every Frostbolt is worth more, it also means there is some point where you are losing enough frostbolts to make it so the gains in damage per frostbolt are not worth this loss in overall frostbolt numbers.

 

Could you re-do your analysis, using a set time period and altering the ratio of spells used in them to something that reflects changes which occur when you are jumping from extremely fast frostbolt (haste capped or 14242 haste) to very slow frostbolt (M>C>H) and see how the altering of spell ratios changes the overall damage outcome?

 

Once it's not midnight I may spend some time on a test dummy with full raid buffs and see what the real change in the number of frostbolts over a given period of time is with M>C>H compared to standard frost.

ET% remains relatively unchanged based off cast times. You only have one filler spell, and it is a filler spell. Increasing its cast time simply reduces the amount of likely FoF procs similar to how reducing haste reduces the amount of BF procs from reduced LB/NT ticks. Just a quick adjustment to my ET% table of a 1s to 3s cast time (just for science), only throws off the ET% by 1.5%, which doesn't disturb any trending at all.

 

You'll find the DPET is where cast time has the biggest influence and those are indeed separated out among the 7 possible builds.

[Akraen 230]

@Frosty - your icicle damage wasn't near what I expected. WTB Malkorok parse!

[Frostedmages 3]

I'll have DS/Malk/IJ/TFP up for you tonight. We should be cleared through that point by 8 or 830PM tonight.

 

We have another Frost mage in the guild who uses the standard haste cap > mastery build. I'll try to wrestle him into raids along side me tonight and we can try to use him as a gague for how spell breakdowns change. If you think that's too variable let me know and I can toss you links to some of my parses when I was using haste cap > mastery for stuff like easier icicle comparisons.

 

My first thought on why Icicle damage may not be where expected is because my haste is looooooooooww. 4k is seriously low, meaning I'm missing out on a ton of Frostbolts and WElly bolts in comparison to the normal haste build. I don't have anything to support that idea though, just a quick thought.

 

On the ET% thing: I did a similar thing with your sheet and messed around with the cast times after I did that run and saw that there was little difference - as you note. This actually surprises me quite a bit. I don't know if this means that there is something off with the sheet, or if the bonus damage from heavily inlfated mastery truley does make up for the loss in number of icicles generated due to less FrB/WEbolts. I'm inclined to believe the later if forced to choose though. Hope we get more logs soon after tonight and can start piecing things together

Edited March 18, 2014 by Frostedmages

[mastamage 2]

@ 104.41% Mastery, 5500 Haste and 35% Crit my Icicles are doing anywhere from 30-36% of my overall damage, through various logs. Not one Malkorok pull was like another (tossed up multiple times 1 attempt, hardly any on another). Those attempts were with Mage Armor. On attempts with Frost Armor, Icicles were consistently doing 24-26% of my overall damage.

 

edit: fixed mastery % used

Edited March 19, 2014 by mastamage

[Soffish 1]

Reeeeally sloppy clear tonight so far, but toying with mastery > crit frost. Using Mage Armor all night. Bomb damage high, Icicles damage high, Welly damage comically high. Lazy farm deaths early on. Maybe more later.

 

http://www.warcraftlogs.com/reports/cbRACw7m4q9yJLFP

[Akraen 230]

I ran some heroic 25 last night and have come to the conclusion that our stats don't matter and only having 6k haste is the least pleasant thing in life. I mean it was seriously hard to play-- evocate took so long to channel I was rushing to avoid things. I wanted to die.

 

So I don't care if it's better - I'm using my haste build... but I do want to find out, for science!

[Soffish 1]

I'm in the same boat. Huge QOL penalty gimping haste.

 

It seems like it would be pretty decent and "worth" it for any of the encounters where you're cleaving 75%+ of the time, as you really feel the extra juice on the icicles, but I'm in the same boat. We cleared through Paragons last night and afterward I reforged/gemmed back to 14242 in preparation for Garrosh tonight. 

 

I was surprised my Spoils numbers weren't higher, even though I was pretty lazy about using Freeze. Siegecrafter felt pretty good. Good for incidental padding on Paragons.

[Frostedmages 3]

I would have logs for you but last night started out bad, so I didn't feel comfortable playing a potentially suboptimal spec/gearing in the face of angry RL. lol. >.<

 

I think there is something wrong with the spreadsheet though. I'm not sure %ET is a good way to figure out how many times you are casting each spell during an encounter. I tried really messing with the frostbolt cast time (like making them completely ridiculous on both ends) and it had no real effect, when in reality this should be completely changing things.

 

I did mess around with it in flex, and I agree it feels terrible.

Edited April 10, 2014 by Oltier
Use nicer words to express yourself, please :)

[mastamage 2]

 

I did mess around with it in flex, and I agree it feels terrible.

Going from 9.7k haste Arcane build it was much easier to test and transition for me but idk if it's even worth it at this point. Play what you want and leave it at that.