[SimC] WoD Arcane SimC Discussion

[Komma]

This thread is intended for discussing Simulationcraft for Arcane Mages. The goal is to get more players involved in simulation development, so we can better model in-game Arcane Mages and provide accurate statistics/analysis.

Please report any bugs or suspected inaccuracies regarding Arcane Mage simulations here. You are also welcome to post any thoughts on the current APL, gear/talent/racial selection for default profiles. If you are unfamiliar with SimulationCraft, Frosted has written a good introductory guide at http://altered-time.com/forum/viewtopic.php?f=2&t=620" target="_blank

All latest sample talent comparison charts can be found in my directory at:
http://downloads.simulationcraft.org/Komma/" target="_blank
The report filenames are timestamped, marked with the chosen specialization, and level of gear. Additionally, Frosted also has sample charts that are produced with Highmaul gear.
http://downloads.simulationcraft.org/Frost/" target="_blank

For convenience, the latest APL can always been found at:
https://code.google.com/p/simulationcra ... _T17N.simc" target="_blank
The sample reports also provide sequence tables that can be used to observe casting orders.

Please keep the thread on topic, and avoid raising issues that are not related to SimulationCraft, such as asking for gear or rotation advice. Those are better served by the Q&A forums, or class forums.

[Komma]

To start, let me announce a bugfix that was made last night that heavily affects talent comparisons.

One of the mysteries we've had in Arcane talent comparisons was that SN/PC combinations feel much more potent in game than observed in sims, while NT/AO seemed absurdly high. Last night, we discovered a bug where Arcane Missiles was not benefiting from 30% extra damage on Prismatic Crystal in simulations. This has been fixed today, and will be included in the next release.

As of January 4th, 665 itemlevel (T17N without set bonus) talent comparisons look roughly as follow:

Source: http://downloads.simulationcraft.org/Ko ... Bonus.html" target="_blank

This seems to correlate much better to what we're seeing in game, although damage done by Prismatic Crystal still seems a little lower than observed in game. We hope to catch more bugs/issues to close this gap.

[Soggs]

This seems to correlate much better to what we're seeing in game, although damage done by Prismatic Crystal still seems a little lower than observed in game. We hope to catch more bugs/issues to close this gap.

I think one in game bug contributes to this experience. Supernova sometimes can generate 2 Arcane Missile procs at the instead of just one. Since SN use is concentrated around PC use that only amplifies that effect.

I cannot reliably reproduce it, so I have no idea how we can model and frankly it is probably not intended behavior. All I can say is that it happens fairly often and is not a rare corner case.

[Komma]

I think one in game bug contributes to this experience. Supernova sometimes can generate 2 Arcane Missile procs at the instead of just one. Since SN use is concentrated around PC use that only amplifies that effect.

Already modeled!

[Soggs]

The APL is not proctected against lucky streaks throwing off Evocation and AP timing. If you skim through the top logs, it rather looks like that people start burning 15-20s before Evo comes off cooldown and then use Evo immediately. Regardless of the actual mana percentage.

I have to think about how implement that in APL, just wanted to share the thought.

[Killget]

So to clarify. on the sims, Has the bugs mentioned in the post been resolved?

Also, would there be any chance of getting some updated sims for higher ilvls?

And if at all possible, some sims for 2pc and 4pc talent combos?

Im getting better at simcraft but still have trouble putting in the intricies and actually displaying them correctly.

Thanks

[Chev]

Also, would there be any chance of getting some updated sims for higher ilvls?

And if at all possible, some sims for 2pc and 4pc talent combos?

Im getting better at simcraft but still have trouble putting in the intricies and actually displaying them correctly.

Thanks

This is not really the thread to ask for this information. This is to be used to discuss the Arcane module. If there are particular sims that you want to see, open a new thread and ask there.

With that said, the information you asked for is already available:

Tallent Combinations: http://altered-time.com/forum/viewtopic.php?f=3&t=833" target="_blank (needs to be updated for BRF gear but this will be done in the next few days.

Current gear vs 2 set vs 4 set (at different levels) http://altered-time.com/forum/viewtopic.php?f=3&t=1071" target="_blank

Trinket Comparisons: http://altered-time.com/forum/viewtopic.php?f=3&t=787" target="_blank (updated with BRF but not 6.1 trinkets)

Weapon Enchant Comparison: http://altered-time.com/forum/viewtopic.php?f=3&t=843" target="_blank (updated using BRF BiS)

If you can think of anything else you want to see, just ask, but not in this thread.

[Komma]

Recently I updated the Arcane APL with some cleave optimizations. Cleave refers to situations with more than 1 active target, but less than the number required to start using AOE spells such as Arcane Explosion. Robomage now uses Arcane Barrage and Arcane Orb more aggressively, even during burn phases. Also, the mana threshold for evocation. It is now dependent on the number of enemies. It means to represent the behavior we often see on Beastlord - staying at higher mana level for strong Nether Tempests and Arcane Barrages, while making sure we use Evocation for the tier 17 2-piece bonus.

Code: Select all

actions.burn+=/evocation,interrupt_if=mana.pct>92,if=time_to_die>10&mana.pct<30+2.5*active_enemies*(9-active_enemies)

Yes, that is a quadratic function based on the target number. It is a poor-man's version of an asymptotic curve. Basically the goal is to keep the single target threshold the same (50%, ie. ~2 ticks of evocation), while getting closer and closer to 75% (ie. ~1 tick of evocation) as the target number increases. It only works until 4-5 enemies, but after that the AOE action list takes over anyway.

[Chev]

A quick translation for those who don't want to work out the math behind this:

1 target = 50%
2 = 65%
3 = 75%
4 = 80%
5+ = AoE rotation

[TLTeo]

So essentially with AO/NT, when burning you want to keep AO on cd and Abar before casting it to cleave as much as possible?

[Frosted]

A quick translation for those who don't want to work out the math behind this:

1 target = 50%
2 = 65%
3 = 75%
4 = 80%
5+ = AoE rotation

Drawing any sort of concrete rules for mana %s / evocation times and that function doesn't make a lot of sense I don't think. The function is entirely made up, and is just serving as a means to non-linearly increase the evocation mana threshold the sim evos at without having a bunch of different evo actions in the APL for different target numbers. The values you come to may or may not be real - they just sort of work best for what komma is trying to accomplish here (which is a more simple/elegant APL for multiple number of targets and arcane cleave)

[Komma]

A quick translation for those who don't want to work out the math behind this:

It is a poor-man's version of an asymptotic curve.

Don't read too much into it.

[Frosted]

So there was a fairly significant bugfix that I just pushed.

There has always been questions about the way SimC was doing RPPM procs (particularly, the Fire T17 4pc). Today Komma and I were looking into the RPPM proc rates of the 4pc for Fire and noticed it was too high compared to what others were experiencing, and what the formulas for them said we should be having. Ended up concluding that the module was mistakenly giving haste scaling to the proc rate - causing it to be heavily inflated on the pull, and generally inflated elsewhere.

This is now fixed for both the Fire and Arcane 4pc procs. Resulted in about a 1.5% DPS loss for Fire overall, and a loss in 4pc strength from 6.35% down to 4.67% DPS gain - which is quite significant. This resulted in a 1% DPS loss for Arcane overall.

Ultimately, this is my failing. I had gone and investigated this weeks ago when the 4pc first went public and people were complaining it was overvalued in SimC due to it always procing on the pull - and indeed, I saw a >100% proc rate on it when I checked it. But I was stupid and didn't look into it further to figure out why, and just assumed it was right, or that there was some technical reason that was occuring (even though I knew it couldn't be right - since we never experienced 100% procing on the pull).

[Carwey]

Would it be possible to integrate equipping trinkets+ring at 5-7 seconds before the pull, resulting in RPPM trinkets proccing 3-5 seconds into the pull? From testing there's a 10 second ICD on RPPM trinkets after equipping them(assuming a boss pull with 2 min rppm protection). That results in not having procs during the first low AC stacks but guarantees full trinket uptime for crystal.

[Komma]

Would it be possible to integrate equipping trinkets+ring at 5-7 seconds before the pull, resulting in RPPM trinkets proccing 3-5 seconds into the pull?

Probably not. The rogues requested weapon swapping in combat, which also got denied due to how it's set up.

Shifting a trinket a few seconds on pull isn't going to make a significant difference for your overall DPS for a fight, even if it results in slightly prettier numbers on your pull.

[Komma]

It took quite a bit of work, but Arcane has now been migrated to a boolean-flag-based system for burn/conserve. Commands like "start_burn_phase" and "stop_burn_phase" should clearly indicate what the player is doing now. Problems such as dropping out of burn into conserve due to spending mana too quickly should no longer happen, and mana thresholds are more understandable.

Unfortunately, this has also created monster if-conditions in terms of when to "start_burn_phase". Here's a snippet to give you some idea of what I mean:

Code: Select all

# Regular burn with evocation actions.init_burn=start_burn_phase,if=buff.arcane_charge.stack>=4&(cooldown.prismatic_crystal.up|!talent.prismatic_crystal.enabled)&(cooldown.arcane_power.up|(glyph.arcane_power.enabled&cooldown.arcane_power.remains>60))&(cooldown.evocation.remains-2*buff.arcane_missiles.stack*spell_haste-gcd.max*talent.prismatic_crystal.enabled)*0.75*(1-0.1*(cooldown.arcane_power.remains<5))*(1-0.1*(talent.nether_tempest.enabled|talent.supernova.enabled))*(10%action.arcane_blast.execute_time)<mana.pct-20-2.5*active_enemies*(9-active_enemies)+(cooldown.evocation.remains*1.8%spell_haste)

The line actually makes a lot of logical sense to me, but it'll take a bit of explaining for folks to understand why it works. Long story short, it's an estimation of mana consumption versus mana regeneration, using the time remaining on evocation's CD.

As an additional benefit of these changes, Arcane gained another ~0.8% DPS on single target. Hurray!

[navv]

Would it be possible to integrate equipping trinkets+ring at 5-7 seconds before the pull, resulting in RPPM trinkets proccing 3-5 seconds into the pull?

Probably not. The rogues requested weapon swapping in combat, which also got denied due to how it's set up.

Shifting a trinket a few seconds on pull isn't going to make a significant difference for your overall DPS for a fight, even if it results in slightly prettier numbers on your pull.

The weapon swapping part is not strictly true anymore. Also, we could add a pair of options that lets you customise the USP (unlucky streak protection) and "time since last proc" values of the initial state. This would not "guarantee" anything though.

[beatle]

Why is the action list on the web report different from default one in simcraft ?
"action.arcane_blast.cast_time" replaced by "2*spell_haste"
"cast_time+buff.arcane_missiles.stack*2*spell_haste+action.arcane_missiles.travel_time" replaced by "cast_time"

Also for a small dps increase try setting mana treshold to 96 on crystal for arcane blast
actions.crystal_sequence+=/arcane_blast,if=mana.pct>96&pet.prismatic_crystal.remains>cast_time+buff.arcane_missiles.stack*2*spell_haste+action.arcane_missiles.travel_time

[Komma]

The sample profile is different because it uses the developer version, and updates may not be in public release yet.

I'll investigate the mana threshold change. I wonder why that would work.

[Dutchmagoz]

Managed to gain ~130 DPS by changing

actions.burn+=/evocation,interrupt_if=mana.pct>92,if=time_to_die>10&mana.pct<30+2.5*active_enemies*(9-active_enemies)

to

actions.burn+=/evocation,interrupt_if=mana.pct>88,if=time_to_die>10&mana.pct<30+2.5*active_enemies*(9-active_enemies)