Jump to content


Photo

Cataclysm Mage Simulators and Formulators


  • Please log in to reply
443 replies to this topic

#1 Lhivera

Lhivera

    Bald Bull

  • Members
  • 1,519 posts

Posted 03 December 2010 - 01:59 AM

This thread is meant to serve as a spec-agnostic resource for Mage theorycrafting tools. It should be used both for presenting and discussing data generated by these tools that would be of interest regardless of spec, and for discussing potential problems and theorycrafting errors in such tools and providing feedback helpful for improving them.

Bug reports and questions on the usage of these tools should be directed to the websites below.


Simulators and Formulators

Rawr
Method: Formulation
Platforms: Web (requires Silverlight)

SimulationCraft
Method: Simulation
Platforms: Mac, Windows (binary and source distributions), UNIX (source distribution)

Vontre's Magegraf II
Method: Formulation
Platforms: Web

(Please post information on additional Mage-friendly theorycrafting tools if you know of any.)


Results Collections

SimulationCraft output repository


Standardized Gearsets and Profiles

For the sake of testing the tools against each other, it's helpful to run them all with the same stats, specs, etc. We should build a collection of standard sets, probably including BIS blue heroics (for starting T11 content), T11 359 (for starting T11 heroic content), and T11 372 (for starting T12 content).


Standardized Run Options

Should be developed for consistency and comparison.
At Veridian Dynamics, we can even make radishes so spicy, people can't eat them. But we're not, because people can't eat them.

#2 Lhivera

Lhivera

    Bald Bull

  • Members
  • 1,519 posts

Posted 03 December 2010 - 02:04 AM

Found another bug I think, if you don't select Improved Hot-Streak but do have the normal version it still stops casting pyroblasts.

EDIT: also something wonky going on with the dots, might be with the dd spells as well.

As a L80 gnome mage, naked, no buffs or target debuffs, with the normal Fire spec, glyphs where MA, Pyro and Fireball (183 spp, character sheet and simcraft):
[TABLE]spell|live|simcraft
LB| 598|616
LB DoT| 702|739
Pyroblast!| 2009.5|2153
Pyroblast! DoT| 329|351
[/TABLE]

(1390 spp)
[TABLE]spell|live|simcraft
LB| 1015|1046
LB DoT| 1190|1254
Pyroblast!| 4037.5|4349
Pyroblast! DoT| 492|527
[/TABLE]


It looks to me like something is wrong with Critical Mass -- not in SimulationCraft, in the game. I did some testing myself; I stripped down my spec to eliminate any damage-multiplying talents, so the only multipliers that should have been in effect were +25% from the spec bonus and +20% from the mastery bonus.

180 spell power
Living Bomb tick: 604
Pyroblast DOT tick: 319

2838 spell power
Living Bomb tick: 1533
Pyroblast DOT tick: 666

Then I added 3/3 Fire Power, for an extra 3% damage:

180 spell power
Living Bomb tick: 622 (expected: 604 * 1.03 = 622.12)
Pyroblast DOT tick: 329 (expected: 319 * 1.03 = 328.57)

2838 spell power
Living Bomb tick: 1579 (expected: 1533 * 1.03 = 1578.99)
Pyroblast DOT tick: 686 (expected: 666 * 1.03 = 685.98)

Finally, this won't affect Pyro, but I put my last two points into Critical Mass for +10% damage to Living Bomb:

180 spell power
Living Bomb tick: 674 (expected: 622 * 1.1 = 684.2 ?? additive? 604 * 1.13 = 682.52 -- nope)

2838 spell power
Living Bomb tick: 1711 (expected: 1579 * 1.1 = 1736.9)

That's all I had time for, so I didn't get the chance to explore other possible causes.
At Veridian Dynamics, we can even make radishes so spicy, people can't eat them. But we're not, because people can't eat them.

#3 Jaerel

Jaerel

    Von Kaiser

  • Members
  • 86 posts

Posted 03 December 2010 - 02:50 AM

It looks like the critical mass bonus is additive with mastery.

622 / 1.2 * 1.3 = 673.8
1579 / 1.2 * 1.3 = 1710.6

2 other things

- Is ignite fixed on beta or are folks operating under the assumption that it will be fixed (so as not to have the refresh lockout period) in the first tier of raiding? Both models have considerably higher damage ratios from ignite than what you'd observe on live with similar ratings.

- Both models appear to have significantly higher hotstreak proc rates (almost double) than what you'd observe in game with similar crit rates (although I noticed some "hot streak bug fixes" got added today in simulationcraft. Is there some reason for that?

#4 Tyrian

Tyrian

    Bald Bull

  • Members
  • 2,376 posts

Posted 03 December 2010 - 06:47 AM

Repasting a few posts to get the ball rolling. Looking at Simulationcraft: It uses a talent build with 2/2 in Improved Fire Blast, which is unnecessary.

The sim doesn't use Fire Blast in rotation at all, and 35 yard Impact range (with 1 point) is more than sufficient. What improvements could we see, if that 1 point was shuffled over to Arcane Concentration instead? Furthermore, you don't need Pyromaniac for almost all fights. You could easily manage a full 3/3 Arcane Concentration by skipping 2 points from Pyro. How much could picking up 1/3 or 3/3 Arcane Concentration alter the Scorch VS Fireball ratio?

1/3 Arcane Concentration: Talent Calculator - World of Warcraft
3/3 Arcane Concentration: Talent Calculator - World of Warcraft

The difference is Pyromaniac: One build has it, the other doesn't. As many beta testers concluded, Pyromaniac can be considered too situational for inclusion in standard lvl 85 talent builds. The fights that see it get compelling use are few and far between. Whether any of the aforementioned makes any meaningful difference to the Scorch VS Fireball weaving ratio, should give a good insight as to how unnecessary / useful Arcane Concentration could be for Fire builds (that aren't explicitly speccing for max, albeit situational AOE power - like the current SimC version).

I'm also interested to know how Combustion is being handled by the sim. Is it making the (unrealistic) assumption that it will always be readily (and easily) used with a Pyro DOT, LB, and a Fireball or Pyroblast Ignite up? I say unrealistic because during heroism, and weakened boss phases, you just can't guarantee these will align on demand. I'm not the best at reading all the output, but I noticed this:

combustion,if=dot.living_bomb.ticking&dot.ignite.ticking&dot.pyroblast_hs.ticking



I see Lhivera updated Fire to now include Combustion. It appears Combustion is being used under ideal circumstances. But what can options do players have for when the stars don't align so easily?

1 - Use a weak Combustion immediately (Just Living Bomb will be guaranteed). Can use this exactly on 2 minutes every time.
2 - Wait for a medium Combustion. (Living Bomb and an Ignite) Shouldn't have to wait too long for a single ignite / crit, maybe just 3-5 casts. Whether you wait for a Scorch or Fireball ignite is another factor to be mindful of.
3 - Wait for a strong combustion. (The conditions the sim uses: LB, Pyro dot, Ignite) Doing this will mean you will not be using Combustion every 2 minutes, as the wait time for these conditions to align is substantial - compared to the 2 minute cooldown.

If the sim is modelling Combustion using point (3) - but treating it as though it can be used as easily as point (1) - players will need to be mindful of that when interpreting the results. Things just don't happen so easily that way in practise at level 85.

Edit: How long you can afford to wait for optional Combustion conditions, is also going to depend on Fight length. A 7 minute fight has 4 ideal opportunities to use Combustion (At start, +2 min, +4 min +6 min) , so you have some wiggle room with waiting periods - having a spare minute.

Maybe the 7 minute sim results, which have up to 1 minute of wiggle room for Combustion waiting if three are to be used, will be somewhat more accurate than I thought after all. Question will be whether ~15 seconds (x4) is enough waiting per Combustion to get those optimal conditions to align - yet still have 4 full Combustions used for the fight duration. But for a 6.5 minute encounter for example, this will change as having only ~5-7 seconds max to wait per Combustion (if you still want to use 4 total) is probably not enough to get the optimal stars to align.

#5 Ektoplasme

Ektoplasme

    Von Kaiser

  • Members
  • 50 posts

Posted 03 December 2010 - 09:35 AM

I did some testing with the lastest version of simulation craft on the Fire BiS ivi372 profile for a 5 minutes fight. It clearly shows that waiting for the 3 dots provides the best DPS. Here are the results:

[table="head"]Condition to be met||DPS||DPET of combustion||Description
All three DoTs ticking||26569 +/- 40|| 198k ||
Ignite and Pyro ticking||26536 +/- 40|| 197k || Same DPS as above. This expected considering LB uptime is very high.
LB and Pyro ticking||26498 +/- 40||194k ||Small drop in DPS, but still within the margin of margin. Will do further testing to reduce the margin.
LB and Ignite ticking||25388 +/- 40|| 118k||Huge drop in DPS, the additional combustion you might get doesn't offset the loss.
At least one of the 3 DoTs ticking||25044 +/- 40|| 97k ||Since you will always have LB ticking, this condition means that you use combustion on CD and hope for the best. Not good.


So these results mean that you should use Combustion when at least Pyro is ticking (LB is a given anyway). Ignite adds only a small amount of DPS. I will do additionnal simulation on larger samples to reduce the margin of error and will update the post.

Edit: One thing I also noticed is that the sim always managed to fit 3 combustions during the fight, even when the condition was the most restrictive. It means that at least in BiS 372, the 3 DoTs will be ticking together often enough that you can use 3 combustions during a 5 minutes fight.
I will do the same testing with lower gear levels to check if these results remain the same at lower crit rate.

Edit 2: changed the results with new ones with better margin of error. Added the resulting average DPET of combustion.

#6 Tyrian

Tyrian

    Bald Bull

  • Members
  • 2,376 posts

Posted 03 December 2010 - 09:48 AM

Which Ignites were you referring to, was it the same one for every condition? (Scorch, Fireball or Pyroblast ignite).

#7 Ektoplasme

Ektoplasme

    Von Kaiser

  • Members
  • 50 posts

Posted 03 December 2010 - 09:53 AM

The test is on any Ignite.

And there is no way to improve this without changing the code, because for now you cannot make a condition on the origin of the ignite.

#8 Zakalwe

Zakalwe

    The Chairmaker

  • Members
  • 1,466 posts

Posted 03 December 2010 - 10:05 AM

If the sim is modelling Combustion using point (3) - but treating it as though it can be used as easily as point (1) - players will need to be mindful of that when interpreting the results. Things just don't happen so easily that way in practise at level 85.

I don't play a mage, but I wanted to comment on this. The simulator is just that, a simulator, iterating through every single step of every one of the 10k simulated encounters and accurately simulating the results of every single action taken by the mage actor. Given the action list entry for combustion quoted in your post, the mage will never cast combustion unless all three dots are up. There's no difference between "in practice" and "in the simulator" when it comes to how often this condition will occur.

#9 Lhivera

Lhivera

    Bald Bull

  • Members
  • 1,519 posts

Posted 03 December 2010 - 12:59 PM

It looks like the critical mass bonus is additive with mastery.

622 / 1.2 * 1.3 = 673.8
1579 / 1.2 * 1.3 = 1710.6

2 other things

- Is ignite fixed on beta or are folks operating under the assumption that it will be fixed (so as not to have the refresh lockout period) in the first tier of raiding? Both models have considerably higher damage ratios from ignite than what you'd observe on live with similar ratings.

- Both models appear to have significantly higher hotstreak proc rates (almost double) than what you'd observe in game with similar crit rates (although I noticed some "hot streak bug fixes" got added today in simulationcraft. Is there some reason for that?


Thanks for the catch on the additive bit -- I've passed that along to the main devs.

I'm currently running the sims with the option "aura_delay=0.150" which causes the sim to model both the tick-munching bug and the tick-rolling bug, so it is currently not behaving as if it's been fixed.

There was a fix to Hot Streak yesterday that was causing it to have a chance to trigger on a single Scorch crit. Not currently aware of anything else that's going on, but I'll keep looking.

ETA: Are you looking at Hot Streaks from my output, or are you running the sim yourself? If running it yourself, are you building from the latest source in the repository, or from the release? If you're using last week's release, there are many fixes that have made since that you're not seeing in your results.
At Veridian Dynamics, we can even make radishes so spicy, people can't eat them. But we're not, because people can't eat them.

#10 Maje

Maje

    Don Flamenco

  • Members
  • 487 posts

Posted 03 December 2010 - 02:40 PM

A couple more things, gnome racial isn't working in Simcraft (ie. no 5% mana), and removing evocation from the fire action list leads to a dps increase of about 3% either because it's not worth the time or because it's not used properly in terms of priorities.

#11 Lhivera

Lhivera

    Bald Bull

  • Members
  • 1,519 posts

Posted 03 December 2010 - 03:32 PM

Most likely just not worth the time. I tried with and without in several different configs, and it was always iffy, easily within the margin of error. Recent fixes have probably settled things down. i'll test it in the full profiles and remove it if it matches what you're seeing.
At Veridian Dynamics, we can even make radishes so spicy, people can't eat them. But we're not, because people can't eat them.

#12 Ektoplasme

Ektoplasme

    Von Kaiser

  • Members
  • 50 posts

Posted 03 December 2010 - 03:34 PM

ETA: Are you looking at Hot Streaks from my output, or are you running the sim yourself? If running it yourself, are you building from the latest source in the repository, or from the release? If you're using last week's release, there are many fixes that have made since that you're not seeing in your results.


Yes I run the sim myself. And I built the version I got this morning from the repository (the cataclysm branche).

I did some further testing with a fight length of 250 seconds. It means that only in case 5 (no condition on combustion so combustion is used on cooldown) the sim manages to use combustion 3 times consistently. With the more restrictive conditions (for example case 1 where all 3 DoTs need to be active for combustion to be cast) the sim only uses combustion twice during the fight. Well even in this case 2 strong combustions are better than 3 weakers ones.

It's easy to see if you compare the DPET: 2*198k > 3*97k

So I think for BiS ilvl372 gear, the question is settled. Combustion should be used when the 3 DoTs are active, or at least LB and Pyro Dots.

Now I need to do some testing on ilvl 346 blue gear to see if the ranking changes.

#13 Althor

Althor

    Great Tiger

  • Members
  • 792 posts

Posted 03 December 2010 - 03:49 PM

The gnome racial has now been added although it was too late to make the 403-4 release. Look for it in the 403-5 release in a couple of days time.

#14 Maje

Maje

    Don Flamenco

  • Members
  • 487 posts

Posted 03 December 2010 - 04:12 PM

Actually it seems more to be a case of actions, using this:
actions+=/mana_gem,if=mana_deficit>12500
actions+=/scorch,debuff=1
actions+=/combustion,if=dot.living_bomb.ticking&dot.ignite.ticking&dot.pyroblast_hs.ticking
actions+=/mirror_image
actions+=/living_bomb,if=!ticking
actions+=/pyroblast_hs,if=buff.hot_streak.react
actions+=/flame_orb
actions+=/scorch,if=mana_pct<5
actions+=/fireball,if=target.time_to_die<60
actions+=/fireball,if=mana_pct>39
actions+=/evocation
actions+=/scorch
Seems to yield a dps increase over not using Evocation. Although it is quite small.

EDIT: adding 3 points in Arcane Concentration is worth about 2% dps increase, these changes actually allow you to cast a healthier ratio of fireballs/scorches and for some reason shift the value of mastery above haste although it doesn't yet include a fix for mastery+critical mass interaction which would probably lower mastery value slightly.

#15 Maje

Maje

    Don Flamenco

  • Members
  • 487 posts

Posted 03 December 2010 - 05:24 PM

A couple more things regarding Master of Elements and Arcane Concentration;
1. simcraft doesn't proc MoE for Living Bomb explosions it costs 718 mana @80 and refunds 265 mana (not sure why, should be 215 but 265 it is) on explosion crit.
2. Improved Scorch scorches do infact proc MoE (not sure if that's intended or not, but they do)
3. Any clear casted spell still procs MoE despite not costing anything
4. Pyroblast! doesn't grant any mana from MoE ever, which is logical given it doesn't cost any mana and they baked a special spell for HS.

Basically this part in trigger_master_of_elements:
if ( s -> resource_consumed == 0 )
    return;
makes sure you don't get any feedback if the spell didn't cost anything.

#16 Dejah-Thoris

Dejah-Thoris

    Von Kaiser

  • Members
  • 27 posts

Posted 03 December 2010 - 07:41 PM

MoE refunds 30% of the base cost of a spell. It doesn't matter if the spell is free (like with Clearcasting) or if it costs extra (Arcane Blast stacked). So a crit with a spell while Clearcasting is a mana gain.

The base cost of Pyroblast! is 0, so we get nothing back from MoE. It probably should have a base cost equal to Pyroblast's, with HS making it free in the same way Clearcasting does. A Hot Streak Pyroblast! should refund mana with MoE.

#17 dedmonwakeen

dedmonwakeen

    Bald Bull

  • Members
  • 1,302 posts

Posted 03 December 2010 - 07:50 PM

Master of Elements support has been updated in SimC revision r5904 to proc on "free" spells. It will also proc on Living Bomb explosions.

#18 Maje

Maje

    Don Flamenco

  • Members
  • 487 posts

Posted 03 December 2010 - 08:34 PM

Thank you looks to be working ok now, the only issue for fire I know of at the moment is the interaction between Critical Mass and Mastery as in they stack additively.

#19 Silverwind

Silverwind

    Von Kaiser

  • Members
  • 67 posts

Posted 03 December 2010 - 09:12 PM

I've been able to improve the Arcane rotation a bit (~800 dps) by:

- Making sure the sim doesn't spam AB until OOM, instead it stops spamming at 25% mana and evo's back up.
- Trying to end the fight at zero mana by spamming AB in the last 20s of a fight, as well as using all cooldowns that pop up in that phase. This works most of the time, except when the end of a burn phase happens to be at <20s before the end. The sim can run OOM some seconds too soon in that case. I don't really know how to prevent this.
- Using cooldowns only at 3 or more stacks. Seems like a small DPS gain.
- Using PoM + AB every cooldown. Another small DPS gain.

The new mana timeline looks like ~80% average mana over the whole fight (7m in this case):
http://chart.apis.go...,20&chco=2459FF

Here's the data:
Spoiler
The BiS lists could use some work (like using Crown of the Twilight Queen as off-piece to T11, and maybe Shard of Woe instead of the crit trinket), is there some tool to create these or are they edited by hand?

edit: forum seems to be messing with the code, delete the space in "buff.arcane_ blast.stack<3"

#20 Lhivera

Lhivera

    Bald Bull

  • Members
  • 1,519 posts

Posted 03 December 2010 - 09:42 PM

I'll put your rules into the profile when I get home - I should be able to prevent the OOM issue as well. Thanks!

Edit:

Updated the source file and the profiles with Maje's changes to Fire priority list and Silverwind's changes to Arcane. Also implemented conditions for casting Molten and Mage Armor, which allowed me to have Frost switch from Molten to Mage at 15% mana. It will never switch back, and will evocate if mana drops below 5%. It can go literally forever like this, and it gains some 60-80 DPS in the five-minute encounter.
At Veridian Dynamics, we can even make radishes so spicy, people can't eat them. But we're not, because people can't eat them.




0 user(s) are reading this topic

0 members, 0 guests, 0 anonymous users