I'm very well aware of the general consensus that UA is better raid DPS than ruin. I set out to prove it because I don't take assumptions at face value and I want to see the numbers. I used the lock sheet available here in the forums and stumbled across some results that I found quite interesting. First things first, apologies for my epic wall of text I had a lot to say.
Assumptions (I hate these but you have to have them):
No differences in build other than Ruin and UA
Hit capped at 16%
Full raid debuffs (CoE, CoS, Misery, Shadow Weaving at 10%, and Imp Scorch)
Lag of 200ms
No outside ISB sources
0 DPS from pet
Self Sufficient i.e. including down time for life tap and dark pact with a net loss of close to 0 hp and 0 mana
1000 Shadow 800 Fire 20% crit rate
Total ISB uptime = 38.88%
Immolate 2281 damage/cast over 17 secs 1.7 sec cast time 134 DPS
CoD 10175 damage/cast over 62 secs 1.7 sec cast time 164 DPS
Corruption 3574 damage/cast over 20 secs 1.7 sec cast time 178.5 DPS
Siphon 2522 damage/cast over 32 secs 1.7 sec cast time 78.7 DPS
UA 3481 damage/cast over 20 secs 1.7 sec cast time 173.9 DPS
NF Procs 2406 damage/proc every 75 secs 1.7 sec cast time 32.08 DPS
DP every 34 secs
LT every 15 secs
Total % of cast time consumed on DoT + DP + LT + NF cycle = 53.96%
Leaving 46.04% cast time for SB
SB 2406 damage/cast every 2.7 secs cast time 2.7 sec cast time 46.04% cast time 410.35 DPS
Total DPS = 1171.8
Total ISB uptime = 44.20%
Immolate 2347 damage/cast over 17 secs 1.7 sec cast time 137.9 DPS
CoD 10275 damage/cast over 62 secs 1.7 sec cast time 165.7 DPS
Corruption 3610 damage/cast over 20 secs 1.7 sec cast time 180.32 DPS
Siphon 2547 damage/cast over 32 secs 1.7 sec cast time 79.55 DPS
NF Procs 2651 damage/proc every 75 secs 1.7 sec cast time 35.34 DPS
DP every 42.5 secs
LT every 14.4 secs
Total % of cast time consumed on DoT + DP + LT + NF cycle = 44.77%
Leaving 55.23% cast time for SB
SB 2651 damage/cast every 2.7 secs cast time 2.7 sec cast time 55.23% cast time 542.28 DPS
Total DPS = 1141.0 + 1% to raid shadow damage (ISB uptime difference of 5.32% of the time at 20% boost to shadow damage)
Notes and observations:
-Ruin is more mana efficient because it is mana free damage i.e. fewer LTs and DPs
-Ruin is less desirable in threat limited situations because the damage has greater variance
-UA is more desirable in movement heavy fights
-1% raid shadow would roughly close the gap in DPS with 3 other raid shadow damage sources at 1000 damage
-more shadow damage with no change in crit favors UA while the opposite favors Ruin. Given that +shadow and even +damage are cheaper than +crit rating item budget wise it seems like UA, properly geared, would scale better.
Should we really assume self sufficiency? For example, you don't necessarily want to waste time LTing and DPing to stay at roughly 100% mana because time spent on LT or DP is time lost on casting damage causing spells. However, in order to break the assumption on self sufficient you would need to make up some arbitrary mana pool size and some kind of fight duration. What might be good values for these assumptions?
Is hit capped at 1000 shadow and 20 crit (after devastation) really a good assumption? If not what would be a better approximation? I did run the same calculations at a lower hit percentage and hit capped favors ruin but not by a ton.
At 6% hit from gear (16% total for DoTs) I came out with UA on top by 45 DPS in personal terms but ruin on top in raid terms with 4 shadow sources at an average of 1070 DPS each. In a 25 man raid it is not uncommon to see three locks and two shadow priests so it seems like ruin would be better for overall raid DPS even at a considerably lower hit rate on SB. However...
With 3 locks in a raid, even if they're all affliction spec, how much ISB uptime is any one lock actually contributing? Obviously if they were all generating ISB uptimes of 30% or so then ISB would regularly be up 100% of the time. More shadow damage sources typically means ISB falls off faster so it seems more than possible the ISB uptime improvement of ~5% is overstated.
Would it be useful or more accurate to compare DPS when ISB uptime is a constant value (more realistic in a raid setting with several locks) instead of being totally dependent on the differences in cast rotations between the specs?
Lastly, and probably most importantly, did I do something wrong? I only ask because I was not expecting the two values to be so close. Am I just an idiot for assuming UA was pretty much always better than ruin at attainable crit levels?
Last edited by tetracycloide : 05/31/07 at 8:49 PM.
Your problem is your assumptions. You assume that there is CoS and CoE, but no other damage sources while you yourself are running CoD. You also cannot assumed Shadoweaving and no damage source reduction, while some Shadow Priests will just stick to MF when ISB is up, some will still throw the occasional MB and SW for a spike in healing/mana production. So you cannot make the assumption that CoS, CoE, and Shadoweaving are in effect if you're in a vacuum situation like you've done. This will show and increase in the amount of +crit needed for Ruin to close the gap with UA.
I do see where you are trying to go and I think a lot of Raids and Warlock will head that route as they get more gear from SSC, TE, and onward, that being that you will see one Affliction Warlock in the raid and any other Warlocks running Dest/Demo or Demo/Dest to keep the number of debuffs down while still maintaining a high DPS due to the amount of +spell hit and +spell crit on the cloth armors out there. Overall, the Tier sets favor Destruction and Demonology over Affliction due to the itemization of spell crit rating.
I think you made two assumptions above that are not exactly accurate, and influence your results:
1) you had unrealistic stats that favors ruin , with more hit% (and even crit%) than it is usual
2) you overestimate effect of ISB on multiple shadow casters
I got my numbers slightly different from yours, but then you probably used different XLS calculations ( I used http://www.savefile.com/files/766070 ) . But anyway, results that I got are:
- for 20% crit/hit, UA best Ruin in DPS by 1.40%, while ISB is more efficient in Ruin by 1.07%
- for 15% crit and 10% hit, UA best Ruin in DPS by 3.60%, while ISB is more efficient in Ruin by 0.91%
- for 15% crit and 10% hit and 20% of fight unable to cast, UA best Ruin in DPS by 4.46%, while ISB is more efficient in Ruin by 0.89%
Now, last one is probably something more realistic for fights like Gruul where you need to run from AE or shatter lot of time. But second one is more realistic in general, since +crit for warlocks is poor return in DPS compared to =dmg, and most warlocks chose dmg items/gems over crit, so 20%crit is not common. Also, 20% to hit is definitelly more than average.
As you can see, UA is better even with unrealistic stats, and gets better as you try more realistic situations.
Related to (2), I think you overestimate effect of that 1% ISB bonus on other shadow users. While it would seem that , in case where UA is better 3.60% and Ruin has better ISB for 0.9%, that we need only 4 other shadow users to compensate for difference, it does not take into account that more shadow users will reduce uptime calculated in XLS.
Basically, XLS (at least in my case, but I guess it is same for other spreadsheets) count on ISB being active 12sec once it is triggered. It is reasonable assumption if warlock is only shadow user (since average time for lock to cast 4 needed SB to remobe ISB is about 16sec, which is more than 12sec duration). But it is not valid if there are more shadow users nukers, because then 4 nukes in 12sec or less are very possible to happen.
Therefore, you can consider bonus from ISB as valid for warlock as single shadow user. Also you can consider it valid if other shadow users are other warlocks, since they will both use your ISB time, but will also create their ISB buffs, so it average to same numbers/benefits for all warlocks. In cases where you have 1 lock and 3 shadow priests, you can consider that other shadow nukers would simply "steal" ISB benefit from your DPS numbers shown in XLS, since ISB will not last 12sec, but same amount will increase their DPS - ie, total raid DPS will not change by much.
Conclusion is that you can not compesate advantage that UA has over Ruin in DPS by counting on ISB.
I'm working on the numbers for overall ISB uptime with 2 affliction locks and 1 destruciton lock. It's a completely artifical situation to be sure but it is a more realistc articifical situation than ISB uptime models that only count one lock's worth of SBs.
I'm willing to conjecture that 20% hit is way above average although would like to point out that the assumption was hit capped which is only 16%. Also willing to conjecture that hit capped is above average as well. For my new ISB model with 3 locks I'm going with 6% hit from gear on the affliciton locks and 10% hit form gear on the destruction lock.
I don't think 20% crit on SB is really all that uncommon though. Devestation is 5% by itself and your base crit from int is roughly 5% as well which leaves only 10% from gear or roughly 220 crit rating.
I'll post again when I have my ISB numbers for multipul warlocks, math should be tons of fun on those...
To model raid ISB uptime I assumed 2 affliction locks and 1 destruction lock with only SB eating ISB charages. Frankly anything more complicated than this, like including SPs, is quite beyond my will to calculate.
Using the sheet I cam up with the following numbers for % time spent on SB spam for each spec:
UA 2.2% on NF procs and 48.47% on SB spam
Ruin 2.2% on NF procs and 55.23% on SB spam
Destruciton 62.45% on SB spam
I assumed the following crit rates
UA - 20%
Ruin - 20%
Destruction - 23%
% ISB uptime = total uptime / total time
total uptime = total time * ISB/sec * AVG duration of ISB
% ISB uptime = ISB/sec * AVG duration of ISB
This looks a lot more right now, a lot closer to what I saw modelling a solo Warlock (which required 34.4% crit to hit the break even mark). I bet if you went further you'd probably see around 25% crit would be the break even in this raid situation.
I've looked at it a lot more dynamically than the static numbers I've actually taken the time to put down here may indicate. I'd agree that flat increases in crit would improve ruin and eventually you would have ruin topping UA but at the same time you don't really get to add crit to you gear without giving up the opportunity to add damage and even though the scaling for DPS on UA and ruin with damage is close to the same the differences that extra shadow makes in LT and DP rotations is significant.
Basically what it boils down to is that by the time ruin is actually better than UA based on crit you'd actually see a 1/21/39 destruction build way out ahead of both of them. It's basically a giant cascade with UA as the lynchpin, once you drop it for more SB time it's worth it to drop SL, once you drop SL you're better off with DS than SM, ect ect ect.
hmm... uptime of ISB with multiple shadow users is fairly hard to correctly express with formula.
There may be some problems with above approach resulting in 81% uptime - for example, if same logic is applied on 3 locks with 40% crit, it would result in over 100% uptime. Also not sure how well it scale with number of warlocks.
I made small simulation program that simulate situation depending on crit %, number of afflic locks, number of demo/destro locks and number of shadow priests (only use ISB, not create)
Results that I got differ depending on how I simulate possibility of ISB landing while another ISB is already up - and i must admit I dont know how it really behave. Question is: if ISB land while old ISB still has 2 counters up, will new ISB have 4 counters or 6?
If I go with first option (new ISB overwrite old, with only 4 counters), then I get following ISB uptimes with 20% crit for all ( A=afflic, D=demo or destro, S=shadowpriest)
and some with shadow priests:
note that total uptime is capped at 61%, due to new warlocks not only making new ISBs, but making ISB duration shorter.
If we go with second option (new ISB adds counter to old ISB if existing, so we can have 6 or more), results have similar progression, but its capped higher, at 84%:
and some with shadow priests:
I believe that first assumption (and first set of results) are more correct, although i expected to be able to reach more than 60-80% uptime with lots of warlocks.
BTW, notice that if we introduce shadow priests, uptime goes down, but benefit from ISb is remaining about same. Take for example one usual raid setup , 2A+1D+2S, in first scenario. If only (2A+1D) were there, uptime would be 61%, but bonus would be 61%*20%*3players= 36%. With (2A+1D+2S) we have uptime lower at 38%, but benefit is still close at 38% (38%*20%*5).
Also, both of above scenarios show that even with 2 warlocks we have 60% or more uptime on ISB, so when i use my XLS to calculate warlock DPS for raid environment (Warlock DPS XLS), maybe i should increase uptime from usual 42% (if 20% crit) to above 61% , as more realistic raid scenario.
note that total uptime is capped at 61%, due to new warlocks not only making new ISBs, but making ISB duration shorter.
Can you show your work on this? Just posting numbers does not mean a lot to me if I cannot go back on disect your formulas and inputs.
I suspect that the reason you are coming up with a hard cap at 61% is because you are assuming 20% crit rate for each lock, which is more than likely wrong since deep destruction has backlash at 3% more crit in the same gear. If you vary your crit rates accross specs and assume intelegent gearing (like destruciton locks taking 2 crit rating over 1 damage for gear) then you'll end up with a higher ISB uptime.
as I said in post above, for this i didnt use formulas, since it is bit hard to model with multiple locks and shadow priests.
I made small program that simulate situation , by assuming afflic lock cast SB avery 5sec, demo/destro every 4sec (which frequency i took from my XLS, and include all DoT refreshes and LTs) and wildly guessed that shadowpriest also nuke every 5sec ... although if they use MindFly, its more like 6sec but anyway shadowpriest casting is not relevant for those 61%.
My simulation results in practically same numbers for 1 lock (either afflic or demo/destro) as my XLS calculations, for different crit % values, and that was one way that i could check if it was valid. Also, if i increase crit% hugely (like 50% crit), ISb uptime gets to 99%, not to some impossible values, which also suggested that simulation was correct in that aspect.
But as any program simulation, it is hard to *show* how it gets to numbers. One way to see how it works is to download ZIP from
and inside you will find both EXE which you can start and change values for crit%, number of Allfic/Demo/Shadowpriests, and cast times for their nukes, and see how numbers change. Also inside you can find source file (PAS) and you have possibility to check for possible errors.
As for having cap at 61% because I assumed all locks have same crit%, it is true that different number will be cap if different crit% was used, but important point is that there would be cap - for example if all locks had 23% crit, cap would be at 65%, not much different. Also, assuming that all locks have same crit rate has same level ot probablility as assuming all afflic locks have same crit rate ;p But for purpose of decidint ISB overtime, its quite acceptable.
In honest truth, it's hard to model Affliction and when the SBs/LT-DPs are cast because of the varying length of DoTs. With UA and Corruption both being 18 seconds, CoA being 24, Siphon Life being 30, and if you throw it in there, Immolate being 15 seconds, you're quickly going to start seeing either overlap, small gaps, and large gaps between DoT casts where you may not get an SB/LT-DP off, get one SB/LT-DP off, or multiple SB/LT-DP off. For Destruction it is far more easy to model your SB-Incinerate/LT rotation because you're going to have very defined cast times where you put up Immolate and/or Corruption and have 12 to 15 seconds of time to cast SB-Incinerate/LT. In essence, you're guarenteed gaps for non-DoT casts with Destruction making it easy to model, but you are not guarenteed gaps when you can better model Affliction due to so many DoTs with large variance in time the DoTs last.
i based my assumption on more than "i think", if you are referring to average cast time for SB
I based it on my XLS calc Warlock DPS XLS where i set up lock stats, and got exactly how many SB one build (afflic or demo) use in period of time (and how much its spent on refreshing DoTs or LTs).
That is where i got those 4sec for demo or 5sec for afflic (due to afflic having 2 more DoTs to keep up). Incidentally, that result in same number as tetracycloide used, ie 2/5+1/4=0.65, and his SB/sec was 0.6548 - but again, i didnt use neither simple formula, nor "i think", but XLS calculations ;p
Also, for destro , you should still use all available DoTs to maximize DPS, so you need to calculate time for immolate, CoD, Corruption - so destro gets similar number of SBs as demo unless you use conflagrate, in which case your number of SBs will go down (and probably your DPS too , but that is discutable )
How can you have an XLS that simulates ISB uptime in a raid given crit ratings and SB/sec as inputs that generates a result without using a formula or a combination of formulas? At some point the simulation has to have been programmed with what math to perform on those inputs to arrive at a conclusion. If it wasn't your simulation I could understand if you were not actually aware of the exact formula that it used but you describe it as 'my simulation' so I assume you wrote it. It should not be any more difficult to post how it arrives at those results given those inputs than it was for me to do so in my post.
Also, if the simulation is suggesting 99% uptime at 50% crit with 3 locks using 40% cast time each to cast SB then there is something wrong with the simulation. With those stats 3 locks should be putting out an SB every ~2 secs or so with an average ISB duration of 8 secs and a crit every 4 secs. If the average time between crits is smaller than the average duration of ISB then you're going to see ISB uptime at 100% on average.
What would really be interesting (and also disgustingly complicated) is a confidence interval on ISB uptime given the probabilities of crit, SB/sec, number of locks, ect.
Last edited by tetracycloide : 06/05/07 at 3:33 PM.
Reason: spelling, the bane of my existance