Raid Optimization

Malazaar · 02/28/08, 7:51 AM

In order to quickly find the best possible raid setup for your given raid group, i wanted to write a little (possibly web based) tool that optimizes your raid dps with a given setup (best possible group setup and which curses to use).

As i don't have time to read all the theorycrafting threads i ask you to give me information about your class, how certain buffs effect your dps. This could look like this:

Class: Feral Druid (DPS)
Base DPS: 1500
1% Crit: 0.732 % increased DPS
1% Haste: 0,399 % increased DPS
100 AP: 1.442 % increased DPS
100 Str: 3.267 % increased DPS
100 Agi: 4.952 % increased DPS
100 Armorpenetration: 0.681 % increased DPS
Windfury: -

This will make it possible to determine the effect of all possible melee buffs for a feral druid in a linear approximation.
This doesn't have to be this format (Casters need other stats for example), a link to a good, reliable spreadsheet would also suffice.

Caesar · 02/28/08, 8:54 AM

Base dps is without external buffs? (ie shouts, blessings, totems, unleashed rage and all that gravy?) I would like to meet the druid that do 1500 dps unbuffed and recruit him straight away!

I started writing a tool like this some time ago and I just wanted to give you a little heads up it's EXTREMELY complex and difficult to get it to produce accurate numbers. There are som many variables that it's depending on (getting those base dps values, exactly how much they scale with different stats etc for every class and spec etc).

Remember that you probably need one setup like that for every class and spec, an arms warrior do for example give BF, an aff lock 3% improved CoE etc. And how do you model Curse of Recklessness and wich warlock uses it etc? If it's going to be a usable tool it has to be very accurate, and you have to model a LOT of stuff. All kudos to you if you do, but the system is way more complex than one initially thinks.

Malazaar · 02/28/08, 9:14 AM

Originally Posted by Caesar

Base dps is without external buffs? (ie shouts, blessings, totems, unleashed rage and all that gravy?) I would like to meet the druid that do 1500 dps unbuffed and recruit him straight away!

I started writing a tool like this some time ago and I just wanted to give you a little heads up it's EXTREMELY complex and difficult to get it to produce accurate numbers. There are som many variables that it's depending on (getting those base dps values, exactly how much they scale with different stats etc for every class and spec etc).

Remember that you probably need one setup like that for every class and spec, an arms warrior do for example give BF, an aff lock 3% improved CoE etc. And how do you model Curse of Recklessness and wich warlock uses it etc? If it's going to be a usable tool it has to be very accurate, and you have to model a LOT of stuff. All kudos to you if you do, but the system is way more complex than one initially thinks.

I do know that it's very complex, i'm not sure yet if i'll do a linear appoximation or a real modelling (which will take much more cpu time). I want to find out how much of a difference it really is first.

CoR is really easy to model, i just value it as 800 armorignore for everyone. Different specs all have different scaling factors, that's why i need your help. I also want to include leatherworking (drums) and somehow model the mt's threat.

I plan to make a simple version as fast as possible and include the more complex factors later.

What i need is not the unbuffed dps, more like dps without buffs that only affect your group (bs, totems, lotp, etc.). This should be with current endgame equip (and yeah, 1500 dps is possible as a feral without group specific buffs).

Carlos · 02/28/08, 9:17 AM

The thing is definitely not linear due to Raid-buffs and things try Toskks Calculator with different Parameters for the Cat and you will see very high differences from the lower to the higher levels of Equipment or use Rawr for the Cat DPS and you see the difference there.

Malazaar · 02/28/08, 9:30 AM

Originally Posted by Carlos

The thing is definitely not linear due to Raid-buffs and things try Toskks Calculator with different Parameters for the Cat and you will see very high differences from the lower to the higher levels of Equipment or use Rawr for the Cat DPS and you see the difference there.

I don't want to model different levels of equipment but rather assume everyone has endgame gear. I know that this may limit the usability but different gear levels would just be too complex (at least for now).

So the question about linearity is more like this.

Assume a rogue does 2000 dps, windfury adds 200 to that and battleshout maybe 100. So the question is, does a rogue with both windfury and battleshout do 2300 dps or significantly more ? (for this matter i wouldn't count 2305 as significantly more.

koaschten · 02/28/08, 9:50 AM

Rogue (19/42/0) based on Aldrianas current 0.9.4 sheet
assuming these buffs are always up:
- Improved Gift of the Wild
- Improved Blessing of Might
- Blessing of Kings
- Flask of Relentless Assault
- Spicey Hot Talbuk
- 5/5 Sunder Armor

Base DPS with known 2.4 Gear: 1466dps
- Improved Battleshout +120 dps
- Unleashed Rage +90 dps
- Leader of the Pack +60 dps
- Trueshot Aura +39 dps
- Improved Hunters Mark +35 dps
- Expose Weakness +71 dps (Hunter with 950agi)
- Improved Grace of Air +61 dps
- Improved Strength of Earth +36 dps
- Curse of Recklessness +80 dps
- Fairie Fire +60 dps
- Mangle +20 dps
- 1 Bloodlust every Cooldown from Shaman in the group +20 dps
- 1 Leatherworker with Haste Drums +12 dps
- Windfury +137 dps
- Improved Windfury 144 dps

Values derived from Base DPS
100str 34.5dps
100agi 69dps
100ap 31dps
100crit 54dps
100hit 66dps
100exp 28dps
100ArP 9dps
100haste 58dps

Rogues are very synergistic on the buffs if they get it. A usual group setup will be Warrior, Enh. Shaman, 1-3 Rogues plus the occasional additional Warrior, Hunter or Feral Druid.

Assuming a Perfect Melee group from a Rogue perspective you would take Warrior, Rogue, Enh. Shaman, Feral Druid and probably a Retribution Paladin or BM Hunter.

Assuming a Warrior, Enh.Shaman and Feral in the Group would result in a little different scales.
Buffs assumed: Imp. Windfury, Mangle, Fairie Fire, Curse of Recklessness, Imp Strength of Earth, Imp. Grace of Air (twisted), Leader of the Pack, Unleashed Rage and Improved Battle Shout

100str 48dps
100agi 99dps
100ap 43dps
100crit 78dps
100hit 113dps
100exp 48dps
100ArP 16dps
100haste 95dps

EDIT:

Replying to the post before this.
Said Endgame-Rogue stripped of ALL Buffs would to 1087dps
- Windfury + basic Str totem -> 1212 dps (+125dps)
- imp. Battle Shout -> 1186 dps (+99dps)
- Windfury + basic Str totem + imp Battle Shout -> 1322 dps (+235dps)

so the synergy of a Warrior and Resto-Shaman for a Rogue is basically +10dps which will scale through ANY buff the rogue will get.

As a figure for going all out on possible buffs this would end up at about 2400dps, a plus of 140% over unbuffed just by group and raid buff synergies.

edit: corrected some too hasty made equivalency values. those go really crazy, i wonder what i mixed up now, perhaps i would be better if Aldriana would look over that.

Caesar · 02/28/08, 9:54 AM

Originally Posted by Malazaar

CoR is really easy to model, i just value it as 800 armorignore for everyone ... more like dps without buffs that only affect your group (bs, totems, lotp, etc.).

well this is the complex part! If we for example assume CoR for all bosses, and CoE and CoS (it's not group specific buffs) you need 3 warlocks in the raid. And should their dps be calculated with or without the dps from their curse? And what about a raid setup where you dont have warlocks? And how will you model shadowpriests? If all melee base dps is calculated with BF for example, your calculator will show that a fury warrior brings the most raid dps, while in reality it's very likely not the case.

If you want it to be accurate you have to do base dps without any buffs, even without your own buff (BS for a warrior, motw for a druid etc) and calculate it all. And then as it's not a linear scaling, the numbers you will put in will play a major impact on what the result will be, and to get it accurate is very very hard.

Anyways gl with the project.

Malazaar · 02/28/08, 10:24 AM

Originally Posted by koaschten

Detailed Rogue Buff Scaling Stuff

Thanks for the quick and detailed answer. It seems that (at least for rogues) the scaling is very much non-linear so i need to consider how to best model those buff synergies without having to completely model each class.

Please don't bother posting more scaling info for now as it may be useless untill i've solve the nonlinear scaling issue

Melthar · 02/28/08, 10:31 AM

Originally Posted by koaschten

Rogue (19/42/0) based on Aldrianas current 0.9.4 sheet
assuming these buffs are always up:

Values derived from Base DPS
100str 34.5dps
100agi 34dps
100ap 37dps
100crit 22.5dps
100hit 120dps
100exp 80dps
100ArP 84dps
100haste 49dps

I fail to see how it's possible that agility provides less DPS for a rogue than strength, considering 100 agility = 100AP along with some crit, and 100str = 100AP, no crit. Not to mention 100strength should be greater than 100AP simply due to BoK (which you listed as being included)

koaschten · 02/28/08, 10:43 AM

Melthar well thats the funny thing i stumbled over too, those values are directly taken from the calculation page of the sheet. It boggled my mind but pushed it on the "unbalanced" raising of a single stat and i admit i didn't further investigate that issue.

edit: i looked over the way i derived the values for the stats... quite stupid error... the sheet supplies the value for +1 increments and i simply multiplied them... after the +1 went through all calculations... will update my post to reflect the corrected values.

MonikaRed · 03/14/08, 9:32 AM

I was searching since few days how to calculate a theorical value for ArP (why for example in my own gear spreadsheet S3head was better than Cowl of defiance or t4), this seems to answer to this question, thank you very much for this work.
I have to conclude that 1 ArP = 0.29-0.37AP / 0.14hit / 0.166-0.205crit (Base dps-"Perfect melee group").
I am nearing truth with theses numbers ?
My second question is : globaly thinking of 5man dungeon and 10raid, is ArP losing or gaining interest in regard of ap/hit/crit ?
(I know it depends on which class you play with, but i still hope we can get a general lesson about that. I personally think that ArP looses of his interest while your dps is getting smaller, but how much ?)

P.s.: very first post, apologize for my english

galzohar · 03/14/08, 6:42 PM

The only real way to solve the nonlinear scaling issues is basically to have a macro run all the possible combos and return the one that gives the best DPS. Of course that would have to rely on a spreadsheet for every single class and they all have to be accurate taking everything into account (many spreadsheets neglect DPS downtime, partial resists etc) to get actual cross-class DPS comparisons.

Then you'll have to add non-DPS requirements of your raid (number of tanks, number of healers, classes of tanks and healers). For example you will have to consider if bringing a hunter/mage/lock in a leftover group and putting the resto shaman in the melee group would beat putting the resto shaman in the healer group and an enhancement shaman in the melee group. Same kind of issue goes with elemental shaman and caster group. Even shadow priests (beyond the obvious 1st for the mages/shadow weaving/misery) will need some consideration unless you like playing it safe and running with 3 SPs (infinite healer mana and some extra healing on the raid on top of the dps boost of the mana regen).

Basically if you actually want to get anywhere with this you will need to combine every class' spreadsheet and probably also contact the person who made it to make sure you can "normalize" the dps it gives to make a fair cross-class DPS comparison. Then make a mega spreadsheet that will run all possible raid combos within the limitations and return the one with the most DPS, picking classes from the available pool of guild memebers...

Exbox · 03/14/08, 7:35 PM

kaoschten, did you take into account the possibility of one of those being a draenie ( I realize it doesnt apply to you but to the circumstances of a melee group, it matters greatly

)

Malazaar · 03/14/08, 7:59 PM

Originally Posted by galzohar

The only real way to solve the nonlinear scaling issues is basically to have a macro run all the possible combos and return the one that gives the best DPS. Of course that would have to rely on a spreadsheet for every single class and they all have to be accurate taking everything into account (many spreadsheets neglect DPS downtime, partial resists etc) to get actual cross-class DPS comparisons.

Then you'll have to add non-DPS requirements of your raid (number of tanks, number of healers, classes of tanks and healers). For example you will have to consider if bringing a hunter/mage/lock in a leftover group and putting the resto shaman in the melee group would beat putting the resto shaman in the healer group and an enhancement shaman in the melee group. Same kind of issue goes with elemental shaman and caster group. Even shadow priests (beyond the obvious 1st for the mages/shadow weaving/misery) will need some consideration unless you like playing it safe and running with 3 SPs (infinite healer mana and some extra healing on the raid on top of the dps boost of the mana regen).

Basically if you actually want to get anywhere with this you will need to combine every class' spreadsheet and probably also contact the person who made it to make sure you can "normalize" the dps it gives to make a fair cross-class DPS comparison. Then make a mega spreadsheet that will run all possible raid combos within the limitations and return the one with the most DPS, picking classes from the available pool of guild memebers...

Yeah, i thought about that. Basically i'd have to model each class/spec seperately. For some that's easy, for many already done. However, it will take time to implement. I'll also most likely have to simplify some things at first.

I don't plan to make a spreadsheet, it's a complete program in itself. It's in java so i could make an web based applet. The optimization routine is ready and working i 'just' need to finish an evaluation routine for every single class/spec - which will take a LOT of time.

Basically, i have moonkin, feral druid and a half baked rogue model working right now.

Khurzog · 03/18/08, 6:54 AM

Well another question. We run normally with an enh. sham, one rogue, one feral an two warrior in the melee grp. Problem is, we are 3 dmg warrior actively raiding. Me as Fury, one BF-MS and one 2h-fury (each of the 2h wears CE, me Rising tide and swiftsteel Bludgeon now until I get the S3-2.6-OH-axe) equip is comparable and I do the most dmg of us when in the same grp. My problem is that usually I have to leave the grp because common opinion is that 2h-warrior profit more from an enh. shaman.
Has anyone some numbers proving or disproving that issue? The one getting out of our melee grp has still an restosham dropping wf, thats why I think I would get most benefit of the enhancer due to 10% AP, str. totem affecting my dmg more than the MS-dmg (and my dmg much depending on crit = flurryuptime, our enhancer always twists wf and agi while the resto in our tank group just drops wf)