[TC] SimC Frost APL Discussion

[lynx]

Thanks for the crit correction.

I checked in game, simc and TCOM and you're right, while the data on Wowhead/Wowdb is wrong.
Correct coefficients should be (from in game):
419 + 0.402 * SP

TCOM is off by 1, but it doesn't matter.
Data in simc is hot fixed, by applying 20% on the coefficient. Was committed by Navv, 9 months ago just with the comment "Mage Hotfixes".

Still, frostbolt on meta gem sims higher than FoF-IL. Just imported T16H with the default APL, Patchwerk, 300s, 10k. The difference for me was ~1.5k dps. On "HeavyMovement" there was a 4k dps increase (though it uses RoP so the difference might be even bigger).
Why it's higher on "HeavyMovement" is a no-brainer, but explaining Patchwerk is more difficult.
I think that the difference in dpet is not that big and because of that if you retain your FoFs for after meta runs out you still have the high damaging ice lances, while not loosing a lot during meta up time. While if you spend FoFs during meta you're left with much weaker Frostbolts (in terms of dpet) afterwards.

[Komma]

Thanks for the crit correction.

I checked in game, simc and TCOM and you're right, while the data on Wowhead/Wowdb is wrong.
Correct coefficients should be (from in game):
419 + 0.402 * SP

TCOM is off by 1, but it doesn't matter.
Data in simc is hot fixed, by applying 20% on the coefficient. Was committed by Navv, 9 months ago just with the comment "Mage Hotfixes".

Still, frostbolt on meta gem sims higher than FoF-IL. Just imported T16H with the default APL, Patchwerk, 300s, 10k. The difference for me was ~1.5k dps. On "HeavyMovement" there was a 4k dps increase (though it uses RoP so the difference might be even bigger).
Why it's higher on "HeavyMovement" is a no-brainer, but explaining Patchwerk is more difficult.
I think that the difference in dpet is not that big and because of that if you retain your FoFs for after meta runs out you still have the high damaging ice lances, while not loosing a lot during meta up time. While if you spend FoFs during meta you're left with much weaker Frostbolts (in terms of dpet) afterwards.

"Cast frostbolt if above 66.7% haste, no FO, FoF stacks < 2, and not during first 30 secs of bloodlust". Insert after "ice lance during AT":

Code: Select all

actions+=/frostbolt,if=!action.frozen_orb.in_flight&buff.bloodlust.remains<10&buff.fingers_of_frost.stack<2&spell_haste<0.6

502498 --> 503897 (100K iterations, Error bar 112.5. Credits to soggs for thinking of BL < 10 secs!)

1500 DPS gain. Meta camping is apparently a real thing!

P.S. Your APL is outdated, Lynx. It's still doing 2T16 FoF camping. I think it should be updated in latest 548-6 release.

[lynx]

Thanks for verifying and refining it further!

I did use 548-6 default APL (it's not the one I posted before) in my latest experiments.

[Frosted]

Thanks for the crit correction.

I checked in game, simc and TCOM and you're right, while the data on Wowhead/Wowdb is wrong.
Correct coefficients should be (from in game):
419 + 0.402 * SP

TCOM is off by 1, but it doesn't matter.
Data in simc is hot fixed, by applying 20% on the coefficient. Was committed by Navv, 9 months ago just with the comment "Mage Hotfixes".

Still, frostbolt on meta gem sims higher than FoF-IL. Just imported T16H with the default APL, Patchwerk, 300s, 10k. The difference for me was ~1.5k dps. On "HeavyMovement" there was a 4k dps increase (though it uses RoP so the difference might be even bigger).
Why it's higher on "HeavyMovement" is a no-brainer, but explaining Patchwerk is more difficult.
I think that the difference in dpet is not that big and because of that if you retain your FoFs for after meta runs out you still have the high damaging ice lances, while not loosing a lot during meta up time. While if you spend FoFs during meta you're left with much weaker Frostbolts (in terms of dpet) afterwards.

"Cast frostbolt if above 66.7% haste, no FO, FoF stacks < 2, and not during first 30 secs of bloodlust". Insert after "ice lance during AT":

Code: Select all

actions+=/frostbolt,if=!action.frozen_orb.in_flight&buff.bloodlust.remains<10&buff.fingers_of_frost.stack<2&spell_haste<0.6

502498 --> 503897 (100K iterations, Error bar 112.5. Credits to soggs for thinking of BL < 10 secs!)

1500 DPS gain. Meta camping is apparently a real thing!

P.S. Your APL is outdated, Lynx. It's still doing 2T16 FoF camping. I think it should be updated in latest 548-6 release.

Where does this come in terms of the entire APL? I'm assuming it's one of the lowest, below FoF-IL?

[Komma]

Below AT FoF IL, above regular FoF IL.

[Komma]

Greetings, friends. Some APL updates, along with some other related SimC news.

There's been a lot of suspicion that Frost is simming way above what it should be, so I spent a lot of time this past week trying to find potential reasons, with the help of experienced Frost mages like Kuni. The good news is that we found and fixed some bugs on the way. The bad news is that all the basic audits checked out, and we haven't found any reasons that might explain the Frost sim's high numbers. The bugfixes have also added a few thousand DPS extra to the margin we're trying to account for.

Regarding Lynx's suggestions, I've implemented prototypes for some of them in my private branch. The first one is the mana gem for invocation users, which turned out to be more complicated than expected. The basic and obvious "use mana gem at low mana" clause turned out to lose DPS instead. After much hunting, I found that, running out of mana and evocating early was actually allowing frost mages to avoid using evocation (every minute) during PBoI's procs (every 2 minutes). To fix that, I had to remind the mage to evocation slightly early, if PBoI was about to come up and Invoker's Energy had less than 20 seconds left.

The second suggestion of using frostbolts during meta gem turned out to be a good thing to do, although not for the suggested reasons. As it turned out, camping FoF procs and using Frostbolts during meta gem (and last few seconds of bloodlust) is a DPS gain simply because Frostbolt can better take advantage of the haste gain beyond 50%, and FoF generation is low enough that you can bank procs for after the haste buff expires. Iterating on this idea and finding the optimal caluse turned out to be much more difficult, but here's what I settled on:

Code: Select all

actions+=/frostbolt,if=!action.frozen_orb.in_flight&spell_haste<0.55&buff.bloodlust.remains<(2.5-buff.fingers_of_frost.stack)*8*execute_time&buff.tempus_repit.remains<(2.5-buff.fingers_of_frost.stack)*8*execute_time

Logic: On average, you gain ~1 FoFs per 8 Frostbolts. By using the remaining time on the haste buff, we can calculate the expected number of FoF procs that will be gained, and compare them against remaining space in the FoF proc bank. As long as the remaining time has less than a 50% chance to proc a 3rd wasted FoF, ignore procs and spam Frostbolt.

Personally I feel like this action's condition is a bit voodoo-ish, but it seems to work best from the options I've tested. As a safeguard, I also had to add an extra action before this to use FoF procs before they expire.

I'm going to commit these to the current development codebase, but if you feel like some improvements can be made, feel free to suggest them!

[lynx]

Thanks for checking further into this and refining the ideas!

Maybe I'm nitpicking here, but on average 1 FoF should be generated per 6.67 Frostbolts (with median being 5). Or conversely in 8 Frostbolts you should get on average 1.2 FoFs.
This does reduce dps though, will take another look a little later.

[Komma]

I guess I should explain the trains of thought behind the magical "8" and "2.5" numbers.

The starting point was "how many frostbolts before we exceed 0.5 change to create FoF (ie. >50% chance of munching an FoF). The number was 4. So I had "last 4 frostbolts should ignore FoF stacks". Doubling that gives 8, which is roughly the generation of 1 FoF. So, (2.5 - current FoF stack) provides the 50% munch chance threshold, and multiplying it by 8 scales it by roughly the predicted amount of FoF stacks that will be generated.

So yeah, 8 was selected only because it's double "4", which comes from requiring a whole number for the final threshold, since you can't have half a frostbolt cast.