[TC] Bomb Snapshotting for better DPS: Fact or fiction?

[Komma]

If you ask for "How to increase my dps" advice in forums, one suggestion that often pops up is "refresh your bombs if you get a trinket/meta proc". The common term used for this is "snapshotting". This is so popularly recognized as an improvement technique, that even reputable guides recommend doing it as part of the frost rotation.

Personally I've always been a skeptic, because I haven't seen any convincing evidence of snapshotting giving benefits. Unlike Moonkin or Warlocks, our mage bombs only last for 12 seconds, can't have their strength copied and spread to other targets, and don't offer us supercharged attacks as powerful as Starsurge. Napkin math also creates questions: on single target, bombs are only 10-20% of a frost mage's damage. How much stronger would snapshotting need to make bombs, to be worth spending more than 1 GCD each 12 seconds?

As a SimC APL writer, the obvious thing to do is write commands and make robo-mage do my bidding. But first we'll have to define the question more precisely:

Does snapshotting Living Bomb, on a 450 second Single Target Patchwerk encounter, increase the DPS of a ilvl 588 Frost Mage?


Part 1: How do we make SimC Snapshot Living Bomb?

Let's start with figuring out how exactly we can make robo-mage snapshot.

Case 1: "Refresh Living Bomb only when you have a meta gem proc/bloodlust AND trinket proc EXTRA when compared to current bomb, and 2 or less ticks remain on it."

Code: Select all

actions+=/living_bomb,if=talent.living_bomb.enabled&80%spell_haste>dot.living_bomb.haste_pct&stat.spell_power*0.8>dot.living_bomb.spell_power&ticks_remain<=2&target.time_to_die>tick_time*3

This is a very conservative way of snapshotting - it mostly happens only if you suddenly have trinkets and meta gems line up at the same time in the middle of the fight.

Case 2: "Refresh Living Bomb only when you have EITHER a meta gem proc/bloodlust OR trinket proc EXTRA when compared to the current bomb, and 2 or less ticks remain on it."

Code: Select all

actions+=/living_bomb,if=talent.living_bomb.enabled&100%spell_haste>=dot.living_bomb.haste_pct&stat.spell_power*0.8>dot.living_bomb.spell_power&ticks_remain<=2&target.time_to_die>tick_time*3 actions+=/living_bomb,if=talent.living_bomb.enabled&80%spell_haste>dot.living_bomb.haste_pct&stat.spell_power>=dot.living_bomb.spell_power&ticks_remain<=2&target.time_to_die>tick_time*3

This is basically a looser case of Case 1. It happens a lot more often, and usually means you snapshot whenever you get a meta proc or trinket proc mid fight.

Case 3: "Refresh Living Bomb only when EITHER meta gem proc OR bloodlust is about to expire, AND 2 or less ticks remain on the current bomb."

Code: Select all

actions+=/living_bomb,cycle_targets=1,if=talent.living_bomb.enabled&ticks_remain<=2&((buff.tempus_repit.up&buff.tempus_repit.remains<action.frostbolt.execute_time)|(buff.bloodlust.up&buff.bloodlust.remains<action.frostbolt.execute_time))

This is to test the common "snapshot meta or bloodlust before they run out because extra ticks give you more FFBs".

Case 4: "Refresh Living Bomb only if you have HWF BBoY, the proc is at 9-10 stacks, and your current bomb has 2 or less ticks remaining."

Code: Select all

actions+=/living_bomb,if=talent.living_bomb.enabled&ticks_remain<=2&trinket.stacking_stat.intellect.up&trinket.stacking_stat.intellect.remains<action.frostbolt.execute_time

This is an extreme case for reference. If snapshotting works at all, surely the 32K intellect buff would do the job?


Part 2: When should we Snapshot Living Bomb?
This question isn't the same as part 1. The question "When" isn't referring to the what buffs you have or their duration. It's a question of where you fit it into your rotation - do you do it when you have a BF proc available? What about FoF, 1 or 2 stack? What about during AT?

While I didn't test this extensively, the existing profile puts BF ahead of LB refreshes. This produces 99% LB uptime, and allows minimal proc munching. Obviously, snapshotting is of lower priority than the regular bomb refresh, and therefore also BF. The only question is whether it should be ahead of Ice Lance. Sample tests said it didn't matter, and I wanted to avoid proc munching, so I opted to place it at just above Frostbolt. This means that snapshotting LB casts will only be done if there's no proc to use, and the alternative was to cast frostbolt.


Part 3: Results
A picture is worth a thousand words, so let me just show it to you:

450 seconds single target Patchwerk, Error bar ~67.4 DPS, 250K iterations

Conclusions:
1. Trinkets and meta gems alone are not worth early refreshes.
2. Trinket and meta gems together still aren't worth early refreshes.
3. Snapshotting the very last second of a meta/lust proc barely breaks even.
4. Snapshotting 10 stack BBoYs is probably worth it.
5. Even if you're a god at snapshotting and do it in the best case (HWF BBoY 10 stacks), don't expect more than a ~0.2% DPS increase on average.

In realistic situations, snapshotting requires concentration. If even the best snapshotting in the world can't get you more than a microscopic gain, it's going to cost you elsewhere. Overall, I conclude that:

Snapshotting on single target fights is not a DPS increase.


Part 4: What next?
Well, the sample BiS frost profile is a 14242 gcd-cap haste build, and some folks have suggested that it's a suboptimal haste BP for snapshotting. The 18960/19347 haste BP builds with more crit should have higher ratios of Living Bomb damage, compared to 10-12% LB damage on the 14242 gcd-cap haste build. Will it have different results? I'm not hopeful, but we'll see.

EDIT: Link for Part 5 19347 haste build: http://altered-time.com/forum/viewtopic ... =212#p2109" target="_blank

[nathyiel]

To summarise it, the question is the gain from early refresh superior to the loss from having more GCD ?

If you're simulation are accurate, then the gain is about 0.7%.

[Komma]

If you're simulation are accurate, then the gain is about 0.7%.

...? Where are you getting this 0.7% number?

[nathyiel]

If you're simulation are accurate, then the gain is about 0.7%.

...? Where are you getting this 0.7% number?

wrong number ... it's 0.159%. In short, it's inside the error.

On snapshooting, it's valuable on haste proc only if 1 ticks or more. but LB is bad for this, NT is better for this.
I think it can be viable in fight when bomb is a big part of our dps (spread boss).

[Akraen]

I would still snapshot on TFP for extra cleave from additional NT ticks.

It's good to know how much each haste buff and combination of multiple haste buffs gives your NT.

[Komma]

Part 5: 19347 haste
This is known as the "bomb snapshotting" build, because of the extra LB breakpoint achieved with it. I made a generic 19347 build with AMR, and ran the above setup on it.

Gear: (No, I didn't keep the AMR url, sadly)

Code: Select all

legs=leggings_of_furious_flame,id=105516,upgrade=4,gems=160hit_160mastery_160hit_160mastery_120int,enchant=285int_165crit shoulders=chronomancer_mantle,id=99401,upgrade=4,gems=80int_160mastery_80int_160mastery_120int,enchant=200int_100crit,reforge=crit_mastery wrists=bracers_of_sonic_projection,id=105626,upgrade=4,enchant=180int,reforge=crit_mastery off_hand=revelations_of_yshaarj,id=105650,upgrade=4,gems=80int_160mastery_60int,enchant=165int chest=robes_of_the_tendered_heart,id=105440,upgrade=4,gems=80int_160haste_320haste_160haste_160hit_180int,enchant=80all,reforge=hit_mastery feet=shaseared_sandals,id=105439,upgrade=4,gems=320mastery,enchant=140mastery,reforge=hit_mastery back=xingho_breath_of_yulon,id=102246,upgrade=4,gems=80int_160mastery_60int,enchant=lightweave_embroidery_3 main_hand=arcweaver_spell_sword,id=105550,upgrade=4,gems=80int_160mastery_60int,enchant=jade_spirit,reforge=crit_mastery,weapon=sword_2.2speed_6013min_11168max waist=belt_of_ominous_trembles,id=105515,upgrade=4,gems=80int_160mastery_160hit_160mastery_320mastery_120int,reforge=hit_haste hands=chronomancer_gloves,id=99397,upgrade=4,gems=160exp_160mastery_80int_160mastery_120int,enchant=170mastery,reforge=crit_mastery,addon=synapse_springs_mark_ii head=hood_of_swirling_senses,id=105420,upgrade=4,gems=sinister_primal_160haste_160hit_180int,reforge=crit_haste trinket2=black_blood_of_yshaarj,id=105648,upgrade=4,reforge=haste_mastery finger1=extinguished_ember_of_galakras,id=105498,upgrade=4,reforge=hit_mastery neck=untainted_guardians_chain,id=105465,upgrade=4,reforge=hit_haste finger2=iyyokuks_hereditary_seal,id=105628,upgrade=4,reforge=hit_haste trinket1=purified_bindings_of_immerseus,id=105422,upgrade=4

450 second single target, 25K iterations, error bar ~200DPS. All haste/mastery values preamp.

The conclusion is clear: 19347 haste builds do not favor snapshotting any more than 14242 haste builds.

Here's a haste/mastery reforge plot for the non-snapshotting profile. All of the profile essentially had the same plot, regardlesss of snapshotting strategy used:


It is worth mentioning that 19347 is the peak of a small local maxima. This is because the previous point is just above the 18960 haste BP, and 19347 clears another BP under meta/bloodlust. As expected, 14242 is slightly stronger on single target fights.

So what's the verdict? Personally I feel like this is pretty good evidence against snapshotting. Early bomb refreshes on single target fights are risky, costly in terms of player concentration, and provide no meaningful gain, regardless of your gear setup.

[Kuni]

So what's the verdict?

I'd call it dead, and that's what I'll recommend to people going forward.

[Akraen]

I interpret it as definitely dead. And 19347 > 14242 on any fight with more than one target for pretty much any duration. But it's so close it shouldn't matter.

[Komma]

I interpret it as definitely dead. And 19347 > 14242 on any fight with more than one target for pretty much any duration. But it's so close it shouldn't matter.

Don't quote me on this, but if I remember correctly, Frosted's results were "three targets or more" rather than "more than one target". This would probably apply to fights like TFP, Galakras, Spoils, maybe Norushen etc. Mind you, we also use NT instead of LB on many of those fights, and the results might be different. The "3 target threshold" is probably due to mastery/splitting ice cleaving putting 14242 ahead on two targets. Maybe he can do some clarification.

[Frosted]

I interpret it as definitely dead. And 19347 > 14242 on any fight with more than one target for pretty much any duration. But it's so close it shouldn't matter.

Don't quote me on this, but if I remember correctly, Frosted's results were "three targets or more" rather than "more than one target". This would probably apply to fights like TFP, Galakras, Spoils, maybe Norushen etc. Mind you, we also use NT instead of LB on many of those fights, and the results might be different. The "3 target threshold" is probably due to mastery/splitting ice cleaving putting 14242 ahead on two targets. Maybe he can do some clarification.

It was three targets, which are in range of each other (Which really is not that many fights, and none of the hard ones). I didn't post anything about 2 target (I think I've simmed that before and 14242 was king, due to splitting icicles being so much damage and loving mastery)

EDIT: Here it is, same stats as the above 2 scenarios. 2 target.

[Akraen]

six or one half dozen then - pick whichever you like-- with our levels and BBoY though I think it's reasonable to enjoy the haste if you enjoy haste-- other factors cause larger margins than this quite small one.

[Komma]

Wanted to add a quick note that a bug in SimC has been fixed, where the bomb explosion damage was being undercalculated (https://code.google.com/p/simulationcra ... ae988d975c" target="_blank). This does not bode well for snapshotting, because snapshotting with early refreshes comes at the cost of explosion damage. This puts another damper on potential gains, and only makes things look worse.