Please upgrade your browser for the best possible experience.

Chrome Firefox Internet Explorer
×

[Theorycrafting] DPS Scalars Moving from Dummy to Boss

STAR WARS: The Old Republic > English > Classes > Roles > Damage-dealing
[Theorycrafting] DPS Scalars Moving from Dummy to Boss

Hockaday's Avatar


Hockaday
09.15.2013 , 06:39 AM | #51
This is interesting, and will give it a read. Would you be able to figure a scaler for Hybrid sorc? (2/16/28) I'm guessing it leans more towards madness.

This is my highest parse for the spec, I'm trying to find a spec from a better sorc than I. http://www.torparse.com/a/418426/tim...012/0/Overview I have parsed for 2.3k but gcd hit and dropped my parse down...(thanks swtor!) If you need the actual spec, it's around or I can give it to ya.

Edit: I will supply stats later today. I'm in 69s, sorry...not my raiding toon (sin tank for the!...sadness...)
The Empire is not evil, and the Republic is not good...it is all a matter of opinion.
Nerf Hungering! It's too OP.=Bartle Test Results:= CE VIP
{Reward-Driven 85%} {Exploratory -60%} {Physical 32%} {Thoughtful -26%} {Social -60%} {Competitive 100%}

Mathemagica's Avatar


Mathemagica
09.15.2013 , 01:11 PM | #52
@THoK-Zeus: Yes please, take it apart. If I find something myself I will update my model.

KeyboardNinja's Avatar


KeyboardNinja
09.15.2013 , 03:50 PM | #53
Assault Commando, parse courtesy of Pizza... Note that this is a pre-2.4 parse (i.e. live servers).

74.71% of the damage dealt was K/E. Thus, the armor debuff increases DPS by 0.7471 * 0.1076 = 8.04%.

The execute talent is a 30% increase to burning effects below 30%. I'm *guessing* this affects Plasma Grenade / Fusion Missile, but I don't actually know. Periodic Elemental effects make up 25.29% of the total damage dealt. Thus, the damage increase from the execute is 0.2529 * 0.3 * 0.3 = 2.28%.

Altogether we have, (1 + 0.0804) * (1 + 0.0228) - 1 = 10.5%.
Computer Programmer. Theory Crafter. Dragonslayer on The Ebon Hawk.
Tam (shadow tank) Tov-ren (commando healer) Aveo (combat sentinel) Nimri (df scoundrel)
Averith (hybrid sniper) Alish (lightning sorcerer) Aresham (jugg tank) Effek (ap powertech)

oaceen's Avatar


oaceen
09.15.2013 , 11:25 PM | #54
Quote: Originally Posted by KeyboardNinja View Post
The execute talent is a 30% increase to burning effects below 30%. I'm *guessing* this affects Plasma Grenade / Fusion Missile, but I don't actually know.
it does not.
oaceen assault specialist / oac scrapper / oacao kinetic combatant / oacianado tactician
[Guide] The Dirty Rotten Scoundrel's PvE DPS Compendium

KeyboardNinja's Avatar


KeyboardNinja
09.16.2013 , 11:08 AM | #55
Quote: Originally Posted by oaceen View Post
it does not.
Ok then. Removing the damage from Fusion Missile, as well as the initial hit from Incendiary Missile (the initial hit is not a periodic effect), we have a damage percentage of 21.22%. That gives us 0.2122 * 0.3 * 0.3 = 1.91%, and (1 + 0.0804) * (1 + 0.0191) - 1 = 10.1% overall.
Computer Programmer. Theory Crafter. Dragonslayer on The Ebon Hawk.
Tam (shadow tank) Tov-ren (commando healer) Aveo (combat sentinel) Nimri (df scoundrel)
Averith (hybrid sniper) Alish (lightning sorcerer) Aresham (jugg tank) Effek (ap powertech)

KeyboardNinja's Avatar


KeyboardNinja
09.16.2013 , 11:18 AM | #56
For Focus Sentinel, there aren't a lot of PvE parses floating around, so I'll use this one from A'jantis…

95.15% of the damage is K/E. With 30% armor pen as part of the spec, we have 0.9515 * ((1 - 0.35 * (1 - 0.2 - 0.3)) / (1 - 0.35 * (1 - 0.3)) - 1) = 8.82%.

Dispatch is slightly tricky to calculate. We can reuse our math from earlier looking at the value of VT vs VS for Rage Jugg, meaning that Dispatch is 57% more powerful than Slash. This is not quite right because of off hand damage, but it's very close. Slash was used 35 times in this parse, which is once every 8.5 seconds. Dispatch has a cooldown of 6 seconds, so we can assume that Dispatch is effectively replacing every Slash below 30%. This also isn't quite right due to the way the Focus rotation works (Slash issued consecutively in periodic bursts), but Cauterize gets displaced slightly below 30% to allow more frequent Dispatch usage, so I think Dispatch on CD with eliminated Slash is a pretty fair estimate. Slash is 17.02% of the overall damage when you include Force Lash. Dispatch is 57% more powerful, and thus we have 0.1702 * 0.3 * 0.57 = 2.91% boost to damage dealt.

Overall, that's (1 + 0.0882) * (1 + 0.0291) - 1 = 11.99%. Round to three significant figures and we have 12.0%.
Computer Programmer. Theory Crafter. Dragonslayer on The Ebon Hawk.
Tam (shadow tank) Tov-ren (commando healer) Aveo (combat sentinel) Nimri (df scoundrel)
Averith (hybrid sniper) Alish (lightning sorcerer) Aresham (jugg tank) Effek (ap powertech)

THoK-Zeus's Avatar


THoK-Zeus
09.16.2013 , 11:22 AM | #57
@Mathemagica: You should take in the dps from corrosive dart, during Takedown -FT cast (I think you forgot that). That gives about 800-900 dps during the 3 seconds.

KeyboardNinja's Avatar


KeyboardNinja
09.16.2013 , 11:58 AM | #58
Looking at Combat/Carnage spec, using Noo'dles's parse here…

96.13% of Carnage/Combat's damage is K/E. However, the spec has a 4.5 second buff obtained twice every 20 seconds which provides 100% armor pen, which reduces the benefit from an external armor pen. It is more than a little tricky to account for this, because generally the hardest-hitting K/E attacks get pushed into that window. For now, I'm going to cheat in the following way. The buff from Precision Slash/Gore doesn't precisely come once every 20 seconds, simply due to the fact that Hand of Justice / Slaughter is not a 100% proc. There is also drift due to GCD misalignment with the ICD. I will inflate the value of this proc by assuming it comes precisely on the ICD and never clips the natural PS/Gore CD, which is my hand-wavy way of accounting for the fact that pure time-averaging of this buff deflates its value by a significant bit.

If we time-average this buff, we have 9/20 = 45% of total time in which we do not benefit from an external armor debuff. Thus, we must consider 0.9613 * (1 - 0.45) = 52.87% of total damage dealt to be affected by armor pen. The rest is easy: 0.5287 * 0.1076 = 5.69% from armor debuff.

The execute phase is also a little different. Combat/Carnage gets a proc which resets the CD on Dispatch/Vicious Throw and makes it usable at any HP. Thus, the execute is in rotation. However, the lull between Gore/PS windows is a full 11.5 seconds, in which you will generally use a single Dispatch/VT when below 30%. This can (and often does) displace Rupture, but not always. I'm open to more rigorous ways of accounting for this effect. Right now, I'm just going to assume that we displace a Blade Rush / Massacre with a Dispatch/VT once every 20 seconds.

Dispatch does almost exactly double the damage of Blade Rush, including the forced Ataru hit, off-hand hit and the chance of an Ataru hit on Dispatch itself. Evidence:

Spoiler


Massacre was used 69 times in 345 seconds, which is once every 5 seconds (sanity check). That's 4 every 20 seconds. Thus, 25% of Massacres are doubled in damage below 30%, which is to say 0.2052 * 0.25 * 0.3 * 1 = 1.54%.

Altogether, armor debuff and execute phase should give us (1 + 0.0569) * (1 + 0.0154) - 1 = 7.3%.

Please check my work. Combat/Carnage is pretty complex due to the interaction of relative abilities, and I'm making a huge number of assumptions. Also, Nood'les parse was extremely luck on proc rates (which is part of why it's so high). This translates into fewer requisite Massacre activations. An average parse would have more Massacres, but would also be less able to swap some of them out below the execute phase due to periodic bad proc RNG.
Computer Programmer. Theory Crafter. Dragonslayer on The Ebon Hawk.
Tam (shadow tank) Tov-ren (commando healer) Aveo (combat sentinel) Nimri (df scoundrel)
Averith (hybrid sniper) Alish (lightning sorcerer) Aresham (jugg tank) Effek (ap powertech)

MGNMTTRN's Avatar


MGNMTTRN
09.16.2013 , 01:34 PM | #59
You are all of course free to engage in this sort of work, but I can't take any of these results seriously until I see some actual tests that demonstrate the 'dps scalar model' is actually predictive. You need to demonstrate that the damage will increase in the way you predicted for at least some of the specs.

Two kinds of tests need to be conducted in order to demonstrate that your model is actually predictive:
1) parse DPS of a spec without armor penetration, then parse DPS of that spec while someone else maintains the armor debuff
2) find a boss with a large HP pool and DPS him from 100% to 0%. See if execute effects behave approximately as predicted

Until these tests are conducted and the model is verified for the most common cases (shared DOT tree hp <30% execute effects and classes without any inherent armor penetration) I'm going to assume all this work is invalid. While the model is reasonable and intuitive, the reason it's not safe to assume it's actually a good model is because many rotations have emergent effects when executes or armor penetration are applied; e.g. Sharpshooters and Dirty Fighting Slingers both have large changes in their rotation due to the availability of Quickdraw, Shadows have different proc chances bound to their execute because it's a single attack rather than two attacks at once, all DOT specs have a point in the boss's HP where it becomes disadvantageous to reapply DOTs because the boss is going to expire before the DOT completes its ticks, Vanguards have different armor penetrations on HIB and Hammershot, etc.

KeyboardNinja's Avatar


KeyboardNinja
09.16.2013 , 01:40 PM | #60
Quote: Originally Posted by MGNMTTRN View Post
Until these tests are conducted and the model is verified for the most common cases (shared DOT tree hp <30% execute effects and classes without any inherent armor penetration) I'm going to assume all this work is invalid. While the model is reasonable and intuitive, the reason it's not safe to assume it's actually a good model is because many rotations have emergent effects when executes or armor penetration are applied; e.g. Sharpshooters and Dirty Fighting Slingers both have large changes in their rotation due to the availability of Quickdraw, Shadows have different proc chances bound to their execute because it's a single attack rather than two attacks at once, all DOT specs have a point in the boss's HP where it becomes disadvantageous to reapply DOTs because the boss is going to expire before the DOT completes its ticks, Vanguards have different armor penetrations on HIB and Hammershot, etc.
If you actually read the specific justifications, you will see that we're accounting for the effects you have listed (except for the lower proc chance on Mind Crush/Crushing Darkness, which is negligible).

The calculations are fairly easy to verify in part. Once we're done with the Vanguard/Powertech numbers, it will be pretty easy to verify their value quite precisely simply by parsing with an armor debuff. In any case…

If you would like to point out specific flaws in the math we're using, please feel free. That's why this is a thread. Otherwise, posting that you're considering all this work invalid until you see combat log verification is extremely counter-productive and borderline ignorant. All theory crafting proceeds in the manner we have followed, and it does so because statistical models are much smoother and produce far more precise results than practical testing. Of course, if theory crafting fails to predict actual bosses, then naturally the work is useless, but the point of good theory crafting is to account for all factors in a smooth, stochastic environment. If you can show that we have failed to account for the factors involved in a statistically sound manner, then please do so and we will adjust the work. Otherwise, randomly sniping the math just because it's math is not welcome.
Computer Programmer. Theory Crafter. Dragonslayer on The Ebon Hawk.
Tam (shadow tank) Tov-ren (commando healer) Aveo (combat sentinel) Nimri (df scoundrel)
Averith (hybrid sniper) Alish (lightning sorcerer) Aresham (jugg tank) Effek (ap powertech)