SimulationCraft for Warlocks (4.0.3a numbers at 80 and 85) - Page 97

Injez · 05/10/09, 9:24 AM

Originally Posted by Zakalwe

If you want to test what spec is more latency dependent, you need to adjust *all* the lag parameters, not just queue_lag, which specifically increases only the lag penalty applied when queuing up spells after spells with a cast time. Since destro casts more spells with a cast time than affliction, of course you're going to end up with a disadvantage for destro.

I took parameters*2.

queue_lag=0.15
13/58: -206
41/30: -203
53/18: -125
-----------
gcd_lag=0.3
13/58: -213
41/30: -199
53/18: -133
-----------
channel_lag=0.5
13/58: 0
41/30: 0
53/18: -33
-----------

Originally Posted by Zakalwe

That being said, dots are inherently less latency dependent than direct damage, so I don't dispute the fact that destro is more latency dependent than afflicton.

As you can see, not only than affliction

And, to be honest, around 25% of damage in 41/30 spec is FG, whose latency = 0.

Warlocomotif · 05/10/09, 9:47 AM

Originally Posted by Injez

I took parameters*2.

snip

As you can see, not only than affliction

And, to be honest, around 25% of damage in 41/30 spec is FG, whose latency = 0.

In terms of rotation I think 0/41/30 is worst, however as you pointed out- due to 25% of their damage coming from the felguard, their dps doesn't take as big a hit as destruction does.

Copola · 05/10/09, 11:40 AM

Originally Posted by Pyralissa

Fire Stone outperforms the spell stone on live. The only reason the spell stone is superior for Demonology is the huge ratings boost granted by Master Conjurer. Without that boost in ratings, the largest DPS gain from the stone comes from the DD or DOT portion of the stone, which obviously favors the Fire Stone for Destruction.

Every bit of testing I've done on the target dummy shows spell stone as the superior choice for destruction. I have a fairly significant dps drop every time i try with fire stone. I think the haste portion as well as the way conflag works now with it scaling off of immolate is the difference.

The reason it's soo much better is the new Destro tree benefits soo soo much from direct damage, with the HUGE boost to incin/cb damage in the patch.

Thank you, I figured it might have to do with the new fire and brimstone.

Heeno · 05/10/09, 2:32 PM

Target dummies aren't a very reliable source of testing some small change in gear, spec, etc. You aren't raid buffed and numerous sources of RNG can throw off your results. Even if it were reliable, you wouldn't see a "significant dps drop". It should only differ by less than 50 dps.

Eyry · 05/11/09, 12:53 AM

I am still trying to learn how to read these simulationcrafts. The one thing that I do not understand is how these simulations are pulling 6-7k+ dps. I am 41/30 atm and I can barely pull 4.5k in a 25 man raid. Given my gear avg is around 216. I am very good at watching my timers and reapplying dots and spaming incinerate. Can someone explain the basis on which these simulations are done? How their DPS is so high? Thanks guys.

Copola · 05/11/09, 2:39 AM

Originally Posted by Eyry

I am still trying to learn how to read these simulationcrafts. The one thing that I do not understand is how these simulations are pulling 6-7k+ dps. I am 41/30 atm and I can barely pull 4.5k in a 25 man raid. Given my gear avg is around 216. I am very good at watching my timers and reapplying dots and spaming incinerate. Can someone explain the basis on which these simulations are done? How their DPS is so high? Thanks guys.

The simulation assumes perfection, meaning you never mess up and that you don't have to move or avoid anything in the boss fight. You are constantly dpsing and never running out of range of totems or auras etc. It is an extreme scenario that is nearly impossible to pull off.

Darkmantle · 05/11/09, 3:38 AM

Originally Posted by Copola

The simulation assumes perfection, meaning you never mess up and that you don't have to move or avoid anything in the boss fight. You are constantly dpsing and never running out of range of totems or auras etc. It is an extreme scenario that is nearly impossible to pull off.

It should also be noted that people have done testing on Patchwerk under near ideal conditions and achieved dps higher than simulationcraft predicts so it is definitely possible to pull close to those numbers under the right conditions.

Cempa · 05/11/09, 4:06 AM

Originally Posted by Eyry

I am still trying to learn how to read these simulationcrafts. The one thing that I do not understand is how these simulations are pulling 6-7k+ dps. I am 41/30 atm and I can barely pull 4.5k in a 25 man raid. Given my gear avg is around 216. I am very good at watching my timers and reapplying dots and spaming incinerate. Can someone explain the basis on which these simulations are done? How their DPS is so high? Thanks guys.

If I can refer you to a recent Blue post (Ghostcrawler MMO-Champion BlueTracker - A Moonkin's Plea)

“Yes. This is a tremendously important concept that a lot of players seem to miss. When you only focus on Patchwerk or training dummies or simcraft, you can come up with a number that may be higher or lower than other specs or classes. The problem is that number is meaningless in many other fights. Do you have to move? Do you have to deal with adds? Do you have to swap out gear to improve your health? Do you run OOM? Is your threat too high? Do you need to free NPCs? All of those are "real" considerations on the actual bosses you're trying to kill.

The real questions, which have somewhat frustratingly fuzzy answers, are: Do you justify a raiding slot? Are you pulling your weight? Would your group be better off replacing you? Does the rogue beat you because your class is undertuned or because it's a perfect fight for a rogue (or because you need to L2P)?”

Simcraft is a great tool because it allows a relevant point of comparison. Don’t take the absolute end number and assume that will be ‘your’ DPS. Rather compare the numbers between specs and then take ‘your’ live DPS and assume it should go up or down in ratio as what Simcraft is reporting. I hope my logic and how I view Simcraft is right? Helpful?

• Zakalwe · 05/11/09, 4:12 AM

There are a few of ways to outperform the sim:

1. Make the fight much shorter than the simulated fight's 5 minute duration, which increases the portion of time you spend affected by heroism and other cooldowns.

2. Be lucky with the RNG and get higher rates than expected on crits or other procs.

3. Gain buffs or procs that aren't covered by the sim, such as any of the multitudes of buffs available in the Hodir fight.

If the sim outperforms you, it's likely because the sim models a completely stationary patchwerk-style fight. None of the fights in Ulduar are like that.

Seleyana · 05/11/09, 4:37 AM

2. Be lucky with the RNG and get higher rates than expected on crits or other procs

The difference of your real dps to the average dps value, obtained from simulationcraft by averaging over many iterations, should be symmetric positive-/negative-wise, shouldn't it? I.e. if you have a patchwerk-like fight and mantain a perfect rotation, your dps will be higher than the average value given by simcraft in 50% of the cases and vice versa. It just depends on the rng for crits, spells that do x to y damage and certain procs (trinkets, mc etc.), or am I missing something (excluding the effect of (uptime bloodlust) / fight time) ?

Heeno · 05/11/09, 7:11 AM

Fifty percent of the iterations will be above the average dps value for the simulator. Real DPS on the other hand, on average will be lower than that (unless as Zakalwe mentioned, you obtain some unexpected buff), provided the fight is at least five minutes. The sim makes the assumption that you play to perfection and never make a mistake in your rotation. And by "playing to perfection" I mean not humanly possible. There will be cases where someone can outperform the sim, this is the result of either RNG, short fight length, or receiving some unexpected buff.

• Zakalwe · 05/11/09, 7:13 AM

Originally Posted by Seleyana

The difference of your real dps to the average dps value, obtained from simulationcraft by averaging over many iterations, should be symmetric positive-/negative-wise, shouldn't it?

Yes, and I didn't mean to imply otherwise - I was just listing the possible ways to get a *higher* DPS than the sim predicts.

Arnath · 05/11/09, 1:01 PM

Well the Simcraft numbers are in a 100% ideal situation, e.g., no movement, perfect play, etc. It shouldn't generally be possible to reach the Simcraft numbers unless you get really lucky with crit rates or find an ideal fight.

Nadagast · 05/11/09, 8:06 PM

Just out of curiosity, would it make sense to add some parameter called mistake rate, which causes the sim to skip the 'correct' move and do the 2nd most 'correct' item in the priority?

For example say the mistake rate is 20%, and the sim is modeling FG Emberstorm, and Corruption and Immolate are on the target, but not CoA, in this situation, 20% of the time the sim would cast Incinerate instead of CoA. I don't know the code so I don't know how hard this would be to implement but it seems interesting to me

Demi9OD · 05/11/09, 8:23 PM

Assuming your mistake is to always cast additional fillers rather than refreshing dots on time, as I believe would mirror actual ingame play, you need merely look at the filler DPET to determine which specs maintain high performance with sub-optimal play. Destruction builds suffer the least due to the high DPET of Incinerate, with mistakes on refreshing immolate causing more harm. Demo comes in the middle, and affliction suffers the most with additional shadowbolts not only being low DPET but haunt mistakes quite costly to overall damage.