SimulationCraft for Warlocks (4.0.3a numbers at 80 and 85) - Page 79

Kalku · 04/15/09, 6:39 PM

Originally Posted by Zakalwe

I've updated simcraft to reflect this. The OP contains the latest results.

I guess destro is officially dead

Is this change in the downloadable version of simcraft?

If not, could you please do these sims in the T8 gearsets so we can see how the specs scale? Based on the scale factors, it appears that 0/41/30 will fall behind deep aff fairly quickly, but it is hard to say for sure.

imtsensational · 04/15/09, 8:00 PM

Originally Posted by Aeiri

So is the conflag hotfix actually live right now? Or is it in the process of being implemented.

It's live.

Weazus · 04/15/09, 8:38 PM

If its live then how am I still hitting for about 10k conflags in PvP, its no more than I was last night? I'm pretty sure the fix hasn't gone live yet.

pfooti · 04/15/09, 9:28 PM

Originally Posted by Weazus

If its live then how am I still hitting for about 10k conflags in PvP, its no more than I was last night? I'm pretty sure the fix hasn't gone live yet.

I spent some time working on the target dummy just about 30 minutes ago, and my conflag was hitting for about exactly 5 ticks of immolate. So at the current time, I don't think it's live. Sure to change soon enough.

Pach · 04/16/09, 3:26 AM

Originally Posted by heemo

but how does that explain the boost that 0/41/30 saw in the latest run?

please answer this. I cant figure out how a non-meta spec that can't even cast conflag would have gained personal dps by a nerf to conflag. Not just a higher rank in relation to other specs but actual dps.

Copola · 04/16/09, 4:50 AM

The hot fix is indeed on live. On the PTR my conflags were hitting for nearly 19k now they are hitting for nearly 13k. That is about a 30% drop in dmg. I was still getting some good dmg with the spec though, I'm playing around with 41/30 right now but may try going back to 40/31 and affliction.

nenad · 04/16/09, 11:07 AM

I decided to check simulator results for 3.1, and only "easy" way I could think of was:
1) remove all raid related things from *.simcraft configs (no pally/dru/war.. buffs, no "optimal raid", only one lock)
2) set up my own gear instead of one in Warlock_T7_Gear.simcraft
3) set up my own talent and glyphs in my Warlock_T8_xyz.simcraft
4) run simulator that way
5) do some TargetDummy tests
6) compare simulator results and my results on target dummy ;p

But I run into problem - when I did above, my Simcraft TXT result shows only pet damage - no damage for me.

After playing a bit with that ("playing" = change line by line untill i found where it stopped showing my damage), I found that if I remove "glyph_life_tap=1" , my damage is also removed.

Then I downloaded source trying to find what is reason, and I found it in:

(sc_warlock.cpp) 
life_tap_t.ready()
...
    if( glyph)
      return( ! p -> _buffs.life_tap_glyph );
...

It is combined error with code and my simcraft file - since in input file it still had "actions+=/life_tap,glyph=1/...", it was checking every time for active Life tap glyph effect, and was casting every time only Life Tap, because without actual glyph, no effect was possible.

One solution would be to always change simcraft file when I change used glyphs, but better one would be to replace code with:

 
   if( glyph && p->glyphs.life_tap )
      return( ! p -> _buffs.life_tap_glyph );

After I replaced that, it worked without problem (BTW, maybe same should be done with "tier7_4pc"). I still didn't get to check TargetDummy results with simulators, but at least now I got some results that I can try later on to compare. At first glance, simcraft DPS gives me numbers (~3450 DPS) that look lower that those I remember from last dummy tests , but it was only 1min test so I need to do it again.

Question for those who know Simulationcraft is if there is maybe some other reason why trying to simulate single player without raid buffs would not work correctly? I could understand getting slightly lower numbers on dummy than on simulationcraft (due to ideal situation in simulation, and non-ideal situation on dummy like not being able to have ideal decimate weaving, or to use DoTs during it ), but I wouldn't expect lower numbers on simulator.

Grimrage · 04/16/09, 11:39 AM

You can create a second "actions+=" line without the Life Tap action and then just comment out the first line.

#actions+=/haunt,debuff=1/life_tap,glyph=1/corruption/curse_of_agony/unstable_affliction/haunt
actions+=/haunt,debuff=1/corruption/curse_of_agony/unstable_affliction/haunt

If you have no desire to preserve the original you can just remove the Life Tap action with no further effort required. No need to alter the code when you can simply take the action out of the line.

Testing on the boss dummy for any spec with a bonus at low life will artificially inflate your numbers. Affliction is an example of this. (not sure which spec you were testing) The sim calculates a real target with real health loss so a higher portion of the fight will be lower DPS and that DPS only gets boosted below the required health threshold.

nenad · 04/16/09, 12:27 PM

Well, obviously I know that I can comment line, and I can also temporarily delete "/life_tap,glyph=1" part. Problem here is that it was not obvious that it is reason for wrong simulation. People who could try same spec, only changing glyphs, would get that error without knowing what is reason.

So, for me after I spent time to find reason for problem, there is really "No need to alter the code" , but in order to make simulationcraft more "user friendly" for those less inclined to hunt reasons for strange behavior, code change is advisable.

As for testing, I was planning to test simulator and dummy both with and without "execute". Way to do that is:

1) set simulation time at same as time you can do DPS on dummy (so not 300sec, 150sec is more realistic)
2) find out when would "execute" start on simulator if fight is 150sec. If simulator is not giving that data, estimate at 110sec (take into consideration that DPS in execute is higher ~30%)
3) test on dummy by doing normal DPS in first 110sec, and execute in last 40sec, stop at 150sec by /recount pause

To test other way around, ie simulator without execute, just dont use "/soul_fire,decimation=1" and ",health_percentage>=35", and do normal non-execute damage on dummy for those 150sec. BTW, I was testing 0/41/30 , but will test also Haunt build.

But as I said, I still didn't have time to check on dummy above things ;p

dedmonwakeen · 04/16/09, 1:18 PM

Originally Posted by nenad

I decided to check simulator results for 3.1

snip

After I replaced that, it worked without problem (BTW, maybe same should be done with "tier7_4pc"). I still didn't get to check TargetDummy results with simulators, but at least now I got some results that I can try later on to compare. At first glance, simcraft DPS gives me numbers (~3450 DPS) that look lower that those I remember from last dummy tests , but it was only 1min test so I need to do it again.

Question for those who know Simulationcraft is if there is maybe some other reason why trying to simulate single player without raid buffs would not work correctly? I could understand getting slightly lower numbers on dummy than on simulationcraft (due to ideal situation in simulation, and non-ideal situation on dummy like not being able to have ideal decimate weaving, or to use DoTs during it ), but I wouldn't expect lower numbers on simulator.

The change you recommended has been made in r2169.

Thanks for doing some "real world" testing. The most effective way to start is to ignore DPS and look at hit values and crit percentages of each spell individually. Once those line up, take a look at spell cast frequency of each spell. Once that checks out, run with log=1 to make sure you are happy with the sequence of spells cast.

Odeen · 04/16/09, 3:19 PM

I tested last night, and demonic pact still doesn't take into account the extra spell power granted by fel armor from the talent demonic aegis. Has anyone managed to find a life tap glyph and been able to see if they fixed that bug where is wasn't scaling at all from the buff?

Nnayr · 04/16/09, 6:43 PM

Originally Posted by Darkstarrz

I still feel 40-31 is top spec. Although simcraft stated otherwise I am still able to pull much better dps than affliction and the felgaurd hybrid spec that sim states is top besides aff with a DG. Even with the conflag nerf I still have an average conflag hit for 20k over 6 hours of raiding last night, and was easily able to keep myself at the top of the meters with the rogues and one of our hunters.

I have a really hard time believing this. Most of the fights have a lot of moving and that is why affliction will seem to surpass these hybrid specs. I myself have never liked playing hybrid specs, but just like everyone else I'm sad going back to affliction. If indeed you are seeing numbers like that please post SS, even some parses.

Darkmantle · 04/16/09, 8:57 PM

Originally Posted by Nnayr

I have a really hard time believing this. Most of the fights have a lot of moving and that is why affliction will seem to surpass these hybrid specs. I myself have never liked playing hybrid specs, but just like everyone else I'm sad going back to affliction. If indeed you are seeing numbers like that please post SS, even some parses.

Affliction has one instant CoA which you cast every 28 seconds. It has two 1.5 second casts(using unmodified by haste since everyone has haste, affliction may have more with eradication but demo specs may have more with imp conj spellstones), then 2.5 or 3 second cast fillers. All other specs have a 1.5 second cast, 1 instant dot, 1 instant curse probably cod and either 2.5 or 2.25 fillers. Deep Demo has the highest damage pet and super fast soulfires. Any conflag spec has a really hard hitting instant. Deep destro has backdraft for faster fillers, a fast chaosbolt and a high damage pet. Any decimate spec has fast soul fires.

There is nothing really special about affliction's damage when it comes to movement heavy fights these days. It is also especially susceptible to required movement just as haunt comes of cooldown and during drain soul ticks. Just as Decimate specs suffer from interuption of their soulfire weave.

• Zakalwe · 04/16/09, 11:31 PM

Originally Posted by heemo

but how does that explain the boost that 0/41/30 saw in the latest run?

This took some investigation, but it turns out an architectural change in simcraft changed how raid-wide spell power buffs interact with pets: Previously they were not taken into account when calculating how much spell/attack power the pet gains from the warlock's spell power. Now they are. I believe this new behavior matches in-game behavior though, so the new numbers are correct whereas the old ones were wrong.

• Zakalwe · 04/17/09, 4:25 AM

I've updated the OP to reflect new simcraft standards now that 3.0 mechanics are irrelevant: T8 is Ulduar gear (best known set), T7 is pre-Ulduar gear (best set). I'm aware that my T8 set is pretty outdated, so feel free to post better suggestions.

Is chardev updated with the latest known Ulduar gear now?