Leulier's Warlock DPS Spreadsheet - Page 17

• Zakalwe · 02/04/09, 8:30 AM

[Edit: Deleted. I missed something obvious.]

Abidos · 02/04/09, 11:59 AM

I still find two mistakes on the spreadsheet
- Demon values don't represent the last patch changes, I believe Whitetooth has posted the correct ones in post 152# of this thread
- Demonic empathy is not triggerd by auto-attack crits, only by the demon's abilities and spells.

Bakka · 02/04/09, 1:35 PM

Originally Posted by Abidos

I still find two mistakes on the spreadsheet
- Demon values don't represent the last patch changes, I believe Whitetooth has posted the correct ones in post 152# of this thread
- Demonic empathy is not triggerd by auto-attack crits, only by the demon's abilities and spells.

The numbers White posted were based on PTR values. I don't perform changes based on PTR values. If someone wants to confirm they are all accurate and repost them then I'll update them.

Will investigate demonic empathy further, thanks.

nenad · 02/04/09, 2:44 PM

Few more comments:
- based on some posts in Demo thread (posts #438-#441) I changed base melee pet attack to start with (412.5+fg_ap/7)*1.05 , instead of old (597+fg_ap/7) . Same start should be for all melee pets, only coeficient is different (same 1.05 for Succy, but 0.80 for FH and 0.86 for VW).

- STA field for pets show too low value. I don't know what is real formula, but changing coefficient from 0.3 to 0.7 shows much closer numbers. Maybe someone can come up with real formula, but until then I would suggest changing to 0.7 and coloring it orange. That result in closer values for DK

I included those changes in warlock_dps_demo.xls, and also fixed that DemonicPact stacking issue and added Meta+ImmoAura.

I also included there one feature that I would really like to see in next version of XLS - 0/1 switch for all raid buffs (H22). Whenever I test something at target dummies, I dont have all raid buffs and without that switch its really slow to find all places to turn off ;p

nenad · 02/05/09, 6:04 AM

After intensive testing, I think I got exact formula for pets STA and INT:

STA= (baseSTA+0.75*PlayerSTA*(1+0.05*synergy)+Fort+GoTW)*(1+fel_vit*0.05)*(1+0.02*BoK)
INT= (baseINT +0.30*PlayerINT*(1+0.05*synergy)+ AI +GoTW)*(1+fel_vit*0.05)*(1+0.02*BoK)

For FH, VW, Succ and FG baseSTA=328 and baseINT=150, so XLS should be:
B32=B41=B50=(328+0.75*stam*(1+0.05*synergy)+C9+C11)*(1+fel_vit*0.05)*(1+0.02*C12)
B33=B42=B51=(150+0.3*int*(1+0.05*synergy)+C10+C11)*(1+fel_vit*0.05)*(1+0.02*C12)

For Imp baseSTA=118 and baseINT=369, so XLS should be:
B26=(118+0.75*stam*(1+0.05*synergy)+C9+C11)*(1+fel_vit*0.05)*(1+0.02*C12)
B27=(369+0.3*int*(1+0.05*synergy)+C10+C11)*(1+fel_vit*0.05)*(1+0.02*C12)

Those numbers give exact INT and STA that I could test unbuffed.

Edit: I tested buffed (BoK, MotW) and formula is confirmed for buffs - although I tested while afflic again, so not 100% sure if Fel Vitality multiply buffs also.

nenad · 02/06/09, 6:12 AM

Another issue that could be fixed are short term effects that could be used once (or few times) during fight. All of those have effects that significantly depend on fight duration. Because of that I propose that we rename "OOM time" to "Fight time" (cell should remaoi named "oom_time"), and use it for all such cases.

There is number of cases in different areas. One example are all Potions. They can be used usually once per fight, but effect should be based on fight duration. Now its fixed on 300sec, and with many fights now lasting 150sec it means real effect of those is almost double of what is shown. Solution is to divide with /oom_time instead of /300. For Potion of Wild Magic on 155sec fight it improve effect from 13 to 26.

Another , slightly more complex case, is when short term effects has CD shorter than some fights and can be reused more than once. For example, Demonic Empowerment is now "(15/60)*0.2" or around 5%. But in fight of duration of 135 sec it should be 6.7%. Similar case is with many effects/abilities on CD. For example on above Potion of Wild Magis, if its CD is 60sec it could be used twice in 155sec, so effect is 200*30/155= 39 ... much more than 13 !

Solution for that is to make function that take effect time, CD time, fight time, and return effect uptime percentage. I made such example function in XLS, in Alt-F11, new module3 :

Function uptime_perc(effectDurationSec As Double, cooldownSec As Double, fightDurationSec As Double) As Double
'Return uptime of effect, given its duration, cooldown and fight duration (in seconds)
    Dim full_N, remains, added_N As Double
    full_N = Int(fightDurationSec / cooldownSec)
    remains = fightDurationSec - full_N * cooldownSec
    If (remains > effectDurationSec) Then
        added_N = 1
    Else
        added_N = remains / effectDurationSec
    End If
    uptime_perc = (full_N + added_N) * effectDurationSec / fightDurationSec
End Function

It can be used on all those places, for example:
- instead of 15/60 for DE if field F41 (frequency for FG), use uptime_perc(15,60,oom_time)
- instead of 15/300 in Potions for WildMagic and others, use uptime_perc(15,60,oom_time)

It could even be used on more delicate trinket uptimes. Its bit tricky since it need estimated time for trinket to proc (usually based on your crit chances or number of casted spells etc). But if you have calculated such "TriggerTime", then uptime for trinket would be:

TrinketUptime%= uptime_perc( EffectTime, TriggerTime+CD, oom_time)

For example, Dying Curse is trinket with 15% chance to trigger on your casted spell and 45sec CD, and have effect lasting 10sec. One way to estimate uptime was to estimate TriggerTime (Models!$C$30/0.15) at 13sec, and say that its used once in 58sec (13+45), giving uptime of 10/58 = 17.24%, and estimated effect of 765*17.24% = 132dmg

But in fight of length 140sec, that should be instead 3xDyingCurse/140sec = 3*10/140 = 21.42%*765= 164 dmg!

Since difference is not small, correct way to calculate effect of trinket would be:

DyingCurseDmg= uptime_perc(10, Models!$C$30/0.15+45, oom_time)*765
SundialOfTheExiledDmg= uptime_perc(10, Models!$C$30/0.10+45, oom_time)*590

Those are just some of examples where more accurate numbers could be obtained by taking fight duration into consideration on timed effect.

Abidos · 02/06/09, 9:48 AM

After a few tests with the sparing dummys I confirmed that demonic pact does not take into acount the spellpower increase we get from our spirit with fel armour, only its static spellpower

Bakka · 02/06/09, 10:15 AM

Originally Posted by nenad

Another issue that could be fixed are short term effects that could be used once (or few times) during fight. All of those have effects that significantly depend on fight duration. Because of that I propose that we rename "OOM time" to "Fight time" (cell should remaoi named "oom_time"), and use it for all such cases.

I will never add any sort of fight duration to the spreadsheet. The entire spreadsheet is based around averages. It is not a real time dps calc and was never intended to be. The purpose of this sheet is not to exactly predict what sort of dmg you can put out for all situations but to help people get a good idea about what does what and how things interact with each other. People need to realize this and understand that with shorter duration fights, fights which require movement, etc. these numbers get skewed.

nenad · 02/06/09, 10:38 AM

Originally Posted by Bakka

I will never add any sort of fight duration to the spreadsheet. The entire spreadsheet is based around averages. It is not a real time dps calc and was never intended to be. The purpose of this sheet is not to exactly predict what sort of dmg you can put out for all situations but to help people get a good idea about what does what and how things interact with each other. People need to realize this and understand that with shorter duration fights, fights which require movement, etc. these numbers get skewed.

That is understandable, but fact is that spreadsheet already have "sort of fight duration", and it is called "oom_time". Which is good, since without that many calculations (like LT effects) would be meaningless. Also spreadsheet has many implicit/hidden fight duration constants, like "300" appearing in many formulas (potion effects for example).

Renaming already existing cell would not make spreadsheet deviate from being "based around averages". Even now when oom_time is changed from 360sec to some more expected value for Patch fight for example (~160sec), it change DPS numbers.

Problem is that without taking into consideration fight duration spreadsheet is giving wrong messages/answers to people who would like to use it. For example:
- should I use Potion of Wild Magic on Patch? XLS says 13 dmg - I should not bother... or?
- does Meta and Immolation Aura significantly influence my DPS on Patch? on XLS right now seems not - I should not bother ... or?

Saying that "People need to realize this and understand that with shorter duration fights, fights which require movement, etc. these numbers get skewed" is correct, but it is just listing limitations of spreadsheet. It is similar to saying "this model/spreadsheet is not accurate for different fight durations and movements" - which is true.

But with only little effort it could accurately model different fight times - which in my opinion would be significant since right now majority of DPS discussions and comparations is done based on Patchwerk fights, and those are significantly shorter than 360sec present in spreadsheet or 300sec implied in some formulas.

Trickykid · 02/06/09, 1:24 PM

Originally Posted by nenad

That is understandable, but fact is that spreadsheet already have "sort of fight duration", and it is called "oom_time". Which is good, since without that many calculations (like LT effects) would be meaningless. Also spreadsheet has many implicit/hidden fight duration constants, like "300" appearing in many formulas (potion effects for example).

Renaming already existing cell would not make spreadsheet deviate from being "based around averages". Even now when oom_time is changed from 360sec to some more expected value for Patch fight for example (~160sec), it change DPS numbers.

Problem is that without taking into consideration fight duration spreadsheet is giving wrong messages/answers to people who would like to use it. For example:
- should I use Potion of Wild Magic on Patch? XLS says 13 dmg - I should not bother... or?
- does Meta and Immolation Aura significantly influence my DPS on Patch? on XLS right now seems not - I should not bother ... or?

Saying that "People need to realize this and understand that with shorter duration fights, fights which require movement, etc. these numbers get skewed" is correct, but it is just listing limitations of spreadsheet. It is similar to saying "this model/spreadsheet is not accurate for different fight durations and movements" - which is true.

But with only little effort it could accurately model different fight times - which in my opinion would be significant since right now majority of DPS discussions and comparations is done based on Patchwerk fights, and those are significantly shorter than 360sec present in spreadsheet or 300sec implied in some formulas.

The primary use of the spreadsheet is to define what gear is worth to different specs. Secondarily it let's you decide (within a spec), where to put specific talents, glyphs, and which spells to use. In general the fight length will affect those issues identically. The main factors that contribute to DPS change when fight length changes are already included (life tapping and I believe Heroism). That said, if you really want to see those additional numbers change, it's only 1-2 cells that need editing.

nenad · 02/07/09, 5:04 AM

Originally Posted by Trickykid

The primary use of the spreadsheet is to define what gear is worth to different specs. Secondarily it let's you decide (within a spec), where to put specific talents, glyphs, and which spells to use. In general the fight length will affect those issues identically. The main factors that contribute to DPS change when fight length changes are already included (life tapping and I believe Heroism). That said, if you really want to see those additional numbers change, it's only 1-2 cells that need editing.

Well, I guess that limiting use of spreadsheet to primary, secondary etc is debatable. Bakka mentioned "to help people get a good idea about what does what and how things interact with each other" which is less limited.

But fact is that it is called "Warlock DPS" and that ultimately DPS is what we want to maximize. Everything else is secondary - we talk of gear only because it influence DPS. We discuss talents only because they differently influence DPS.

Spreadsheet is attempt to model warlock DPS. In order to accurately do that, everything that influence DPS should (ideally) br modeled. That include gear, talents, used spells, consumables, buffs, fight time, spell rotation/priorities etc.

Of course, it is not easy to model all of those. Alternative way to analyze DPS is using simulations. Each of them has limits.

But goal of every model is to be as acurate as possible. Limiting to some imagenary "primary,secondary" goals, or saying "this model is not for that use" are different ways of dealing with model shortcommings - first try to stear you away of them, and second one tells you to live with them.

Saying that fight time do not influence results (even those limited to "primary,secondary") is not correct. If you now in 308b change oom_time from 360 to 155, you will get not only different DPS, but different values fro +hit, +haste etc. Effect on comparing talent values is even greater (for example, value of Meta is much influenced by fight duration). That is not to say difference is huge, but it is noticeable. I presume that same line of reasoning resulted in spreadsheet now having itemlist, effect lists, raid buffs ... even if at start most of those were missing. When people back then asked for those things, answers were similar to these now , along lines of "spreadsheet is not intended for that", or "be aware that its not ideal and live with it" or "make your own changes if you need it" - and yet today all of those are in spreadsheet.

I can easily accept that making accurate model is hard - after all that is why many people on this forum try to help and improve it. But I find it harder to accept not improving model when we have information how to do that, especially if its not hard to do.

Maybe my mistake is in thinking about this spreadsheet as "community model" - where all of us work to make it better. Alternative would be all of us posting on some general "warlocks theory" thread, and everyone (or at least everyone inclined ) makes his own spreadsheets. But I dont think that is right way to do, and even people who "defend" not adding things in this spreadsheet are among those who are first to ask whenever someone makes their own "why you need it when we have one already?". And I fully agree - as long as WE have one already, and that one is not held back (ie not improved when we know how) without good reasons.

Trickykid · 02/07/09, 2:37 PM

Originally Posted by nenad

Well, I guess that limiting use of spreadsheet to primary, secondary etc is debatable. Bakka mentioned "to help people get a good idea about what does what and how things interact with each other" which is less limited.

But fact is that it is called "Warlock DPS" and that ultimately DPS is what we want to maximize. Everything else is secondary - we talk of gear only because it influence DPS. We discuss talents only because they differently influence DPS.

Spreadsheet is attempt to model warlock DPS. In order to accurately do that, everything that influence DPS should (ideally) br modeled. That include gear, talents, used spells, consumables, buffs, fight time, spell rotation/priorities etc.

But goal of every model is to be as acurate as possible. Limiting to some imagenary "primary,secondary" goals, or saying "this model is not for that use" are different ways of dealing with model shortcommings - first try to stear you away of them, and second one tells you to live with them.

Maybe my mistake is in thinking about this spreadsheet as "community model" - where all of us work to make it better. Alternative would be all of us posting on some general "warlocks theory" thread, and everyone (or at least everyone inclined ) makes his own spreadsheets. But I dont think that is right way to do, and even people who "defend" not adding things in this spreadsheet are among those who are first to ask whenever someone makes their own "why you need it when we have one already?". And I fully agree - as long as WE have one already, and that one is not held back (ie not improved when we know how) without good reasons.

Want to start by saying I didn't claim fight length is unrelated to DPS. I said that the two biggest factors that contribute to DPS change with respect to fight length area already accounted for.

On a more general level, however, I would put forward that the goal of this modeling is not to predict perfectly the outcome given the same variables in game. The goal is to create a tool that allows for actionable decisions to be made in game. Accurately demonstrating DPS is not useful by itself, you can just go play the game to see that number! The goal is to let the tool enhance your playing by doing the calculations that otherwise are in the napkin-math realm of this forum smarter and automatically. The areas where we need a tool to help us include (in no particular order):
1) Gear/gem choices based on stat weights
2) Glyph choices
3) Marginal talent choices (1/3 MC 4/5 ISB or 0/3 MC 5/5 ISB)
4) Spell usage (immolate or not?)

Beyond any factors which affect these, making the DPS number on the sheet as close to reality is not useful. This is the reason why including partial resists isn't necessary: they affect every spell the same amount, so it won't add any new insights to the choices we have to make.

Now I am not saying we shouldn't improve the work we have so far. I think it's a fine goal to make the sheet more accurate. I just think we should check on what we want to add to make sure it's accomplishing something that will affect the use of the tool. Changes take time to implement, QA (generally greater complexity leads to more bugs and hard to track down dependencies), and some of them can make the sheet more difficult to use for the important choices. So, I apologize if I came off sounding like I was just crapping on your suggestion -- I think you have good goals. It's just important to make sure the fixes made are necessary and useful to the community.

Immortus · 02/07/09, 3:14 PM

Bakka thank you and the rest of the community for a great spreadsheet. I have made good use of it for a very long time.

I have noticed that the weighting of gems on the items tab could use a better model. For example,

Titanium Spellshock Ring int 33 Sta 48 Dmg 49 Crit 42 One Red slot with a 5 dmg slot bonus. DPS value on my sheet 115.82

Now a custom item with the same ring with a 19 dmg gem in it
Custom Int 33 Sta 48 Dmg 73 Crit 42 DPS value on my sheet 127.89

A nice addition to the sheet would be a preferred gem selection per slot color.

Thank You
Immortus

nenad · 02/09/09, 4:28 AM

Originally Posted by Trickykid

The goal is to create a tool that allows for actionable decisions to be made in game. Accurately demonstrating DPS is not useful by itself, you can just go play the game to see that number! The goal is to let the tool enhance your playing by doing the calculations that otherwise are in the napkin-math realm of this forum smarter and automatically. The areas where we need a tool to help us include (in no particular order):
1) Gear/gem choices based on stat weights
2) Glyph choices
3) Marginal talent choices (1/3 MC 4/5 ISB or 0/3 MC 5/5 ISB)
4) Spell usage (immolate or not?)

We basically agree on major details.

I still thinks , as I mentioned in previous post, that ultimate goal of people using such tools is DPS. And we look for tools to help us select best gear/glyph/talent/spell in order to predict and maximize DPS. In other words, DPS is thing by which we measure "value" of gear/talents.

So we agree that calculating DPS for our current gear and talents is not goal of any modeling tool. But being able to calculate difference in DPS if some gear or talents or whatever is changed is goal of modeling tools - in order to help us decide on that talent, gear etc.

Unfortunately, if we want that calculating difference to be meaningful, we need even more accuracy than if we just wanted to calculate total DPS. For example, most people using tools like spreadsheet will change one glyph/talent with other if they gain 20 DPS. Therefore our "margin of error" in predicting DPS needs to be as low as possible.

Above is my general position, not strictly related to fight duration but to spreadsheets and DPS models in general - in order to be more useful, goal of every model should be to get as accurate as possible (of course, within limits of feasibility - if something is too hard to model, then its too hard).

Related strictly to fight duration, I already pointed where it does make difference in what people would select using spreadsheet. Some examples:
1) lock who is deciding between Meta+Ruin or FG+Ember is getting wrong (too low) info from spreadsheet about Meta effect. One reason is because Meta is highly dependant on fight duration. There are other reasons too (not including Immo Aura, not calculating BL effect on Meta etc). BTW, I'm not claiming that end result would be that Meta is better than Ember - just that difference is not that high as in spreadsheet.
2) lock who is deciding if it is worth paying for and using Wild Magic potions on all boss fights. Spreadsheet right now shows estimated 12 dmg/crit (or ~17DPS gain) from those, so maybe some people would decide its not worth paying gold for that. But better calculation that takes into consideration fight duration, BL etc, can get it to 36dmg/crit or ~60 DPS - and that may be worth paying gold

As i mentioned in post #246, there are other cases and examples where it can make difference. But my main point here is that it is easy to incorporate those changes in spreadsheet. This is not something hard to do - I know because I already did it and I posted whats needed to do.

So I still have open question:

If we know how to improve accuracy of spreadsheet model in such way that it is not too hard to implement and it may influence actionable decisions of people who use spreadsheet, why decide not to implement it?

But I would like to point again that we are agreeing on major positions here, so consider this as another attempt to improve spreadsheet. Even if it does not succeed, people like me will have option to include their "corrections" in new versions - maybe time consuming, but doable. Also, there is possibility that in future those differences would become impossible to ignore and will be added to spreadsheet, just like it happened historically to item/gear lists, potions/food lists, gear sets/effects, and many other features that spreadsheet has today but did not have before. Fact that spreadsheet today has much more features that it did year(s) ago, and that most of those new features were incrementally improving accuracy (rather that revolutionary fixing things) support my belief that accuracy should be goal of every model - even if we ignore it, eventually we need to do it or tool will lose its usefulness.

Bakka · 02/09/09, 10:19 AM

Let me state once again that I will not be adding a fight time to the spreadsheet. It's not as simple as adding a few formulas because you did it. And all the justification in the world for adding it isn't going to make me change my mind on doing so. I'm the one who currently supports the spreadsheet and I'm the one who doesn't want to make this 'small' change because I don't want to have to make the next 'small' change which then attempts to model bloodlusts, trinkets, pots, meta, etc. all at the same time as opposed to an average. And then is asked to make another 'small' change which based on fight duration adds multiple metas and BLs. And then gets asked to make an initial cast sequence because this it really matters for short fights and it would be a small change. I'm staying away from anything that makes this spreadsheet anymore accurate based on real time because if I plan on going that route I may as well scrap the spreadsheet entirely and come up with a real time model.

RE: Immortus
There is. There's a sheet called Sock_bonus that needs to be unhidden and adjusted to fit your preferences.