Cataclysm Mage Simulators and Formulators - Page 16

dedmonwakeen · 03/04/11, 2:42 PM

Originally Posted by Maje

There are still some errors with MoE implementation, namely Living Bomb is returning 1411 mana on crits rather than the normal 30%, Flame Orb ticks return 20-21 mana per crit tick (not implemented).

The MoE-FlameOrb interaction has been fixed (r7913), but I believe LivingBomb is working correctly. I see it returning 888 mana in my runs. I've also moved some resource reporting from "debug" to "log" so that this is easier to track without the debug wall of text.

I have also added two new action conditionals that affect channeled spells:

interrupt=1
After every tick the AI will look to see if there is a higher-priority action that is ready. If so, the channel will be interrupted.

interrupt_if=expression
After every tick the AI will evaluate "expression". If it evaluates to a non-zero value then the channel will be interrupted.

All the many options available to if= are also available to interrupt_if=. You can query the state of the player (actions, buffs, dots, cooldowns, etc.)

Maje · 03/04/11, 3:33 PM

Sorry I wasn't clear, yes theoretically Simcraft is working correctly however the game returns 1411 mana for LB, ie. the game is being inconsistent (be it a bug or intended I don't know), that's how it works since 4.0 came out if I remember correctly.

Power · 03/04/11, 6:00 PM

Fire's mana management in 406-11 seems strange:

The simulation switches to Mage Armor, uses Evocation and switches back to Molten Armor.

Evocation without Armor switching gives me a 100 DPS increase.
Armor Switching without Evocation gives me a 200 DPS increase.

dedmonwakeen · 03/04/11, 10:47 PM

Originally Posted by Power

Fire's mana management in 406-11 seems strange:

The simulation switches to Mage Armor, uses Evocation and switches back to Molten Armor.

Evocation without Armor switching gives me a 100 DPS increase.
Armor Switching without Evocation gives me a 200 DPS increase.

There have been several community-contributed action lists for Mage recently and they were coming so fast I must admit I did not provide as much auditing as is my usual wont. Please open an Issue on our site so that I can track the changes and make sure we get this right.

Nyphur · 03/05/11, 4:05 PM

I've been using Rawr to simulate an arcane mage in my guild who can't work out what her best gear choices are, and I think I've hit something of a snag in the way Rawr values trinkets in an arcane rotation. The first issue I've had is with the Soul Casket, which on paper gives 321 spellpower on average over the course of a fight if used on cooldown. But for an arcane mage, it can be used right before a burn phase for added benefit. Am I right in thinking Rawr doesn't take this into account when valuing the trinket, or does it have some mechanism for working out the best time to use it?

The second issue I've hit is with the Volcano card, power torrent enchant and volcanic potions -- basically Rawr doesn't have a separate stat valuation for int gained in combat. Since arcane's mastery is dependent on percentage mana and int gained in combat decreases this, the two wouldn't have the same stat values. Mr Robot is valuing Volcano card very highly for the same character, so there seems to be some disparity between the way that int proc is being factored in.

Kavan · 03/05/11, 7:02 PM

Originally Posted by Nyphur

...Am I right in thinking Rawr doesn't take this into account when valuing the trinket, or does it have some mechanism for working out the best time to use it?

...Since arcane's mastery is dependent on percentage mana and int gained in combat decreases this, the two wouldn't have the same stat values...

I'm not sure where you got your information, but all that is accounted for in Rawr. It takes into account cooldown stacking with burn phases (or wherever it's most optimal) and for int procs it takes into account mastery penalty.

Kavan · 03/05/11, 10:48 PM

Originally Posted by Nyphur

I'm getting this from comparing Rawr against other simulators like Mr robot.

Take a look at How does Mr. Robot estimate trinkets?, perhaps that will explain why Mr Robot values that trinket so highly.

MrDarkflame · 03/05/11, 11:20 PM

Nyphur, you might need to refresh item data. I just tried Rawr online, Rawr b4.0.20, and Rawr build 58474 from source code check in, and all of them (with multiple arcane characters) show DMC: Volcano as a BIS trinket. Just in case, I refreshed my item data and still showed same results.

Kavan · 03/05/11, 11:53 PM

Originally Posted by MrDarkflame

Nyphur, you might need to refresh item data. I just tried Rawr online, Rawr b4.0.20, and Rawr build 58474 from source code check in, and all of them (with multiple arcane characters) show DMC: Volcano as a BIS trinket. Just in case, I refreshed my item data and still showed same results.

I'm actually surprised that you're seeing DMC: Volcano valued high. Even if I load your character I see it below many other 359 trinkets.

• Wizeowel · 03/06/11, 6:12 AM

Originally Posted by Nyphur

I'm getting this from comparing Rawr against other simulators like Mr robot. Rawr predicts that equipping a Volcano card over a Witching Hourglass is a DPS decrease, but all other simulators and valuations put the card as best-in-slot by a long shot. I'm trying to understand why Rawr is giving such a different result, because one of the simulators must be doing something wrong if they disagree so starkly. Is Rawr the only one accounting for the mastery penalty or something?

Firstly, Mr. Robot is not a simulator, it only uses the data that the author borrowed from a real simulator, namely Simulationcraft. Secondly, Rawr is not a simulator but a solver which calculates the dps of all the possible cycles/rotations you might use and picks the best combination of these. If you really want to understand why these tools give different results, then the source code is available for both. Your concern that there is "something wrong" is misplaced. They model the game in different ways and as such will naturally come to different conclusions. If anyone has some specific insight that any of these tools is calculating something incorrectly, they are welcome to post it.

MrDarkflame · 03/06/11, 10:34 AM

Originally Posted by Kavan

I'm actually surprised that you're seeing DMC: Volcano valued high. Even if I load your character I see it below many other 359 trinkets.

Ive been using a saved character file. Just for kicks, I tried reloading my character from scratch like you did. I got same results, but if I add the "Dark Intent (Warlock)" buff, DMC: Volcano becomes BIS for me. Maybe this is the reason why we're getting such different results?

Kavan · 03/06/11, 6:51 PM

Originally Posted by MrDarkflame

Ive been using a saved character file. Just for kicks, I tried reloading my character from scratch like you did. I got same results, but if I add the "Dark Intent (Warlock)" buff, DMC: Volcano becomes BIS for me. Maybe this is the reason why we're getting such different results?

I've figured out what was the problem. Enabling Dark Intent in your case triggered the mastery problem I've posted about before (Rawr.Mage). I've already set Arcane Light setting in advanced options to default enabled for next release, but in the mean time I would strongly suggest everyone to enable it for themselves to avoid this problem.

DrRumpus · 03/08/11, 8:02 AM

Originally Posted by Nathyiel

Conclusion:

The race with passive bonus increase the dps between 0.5% and 0.9%. Because of the rating of each stat, Worgen bonuses is better than Draeneï or Goblin but Draeneï can change some blue gem (20intel/20hit to 40 intel).
Troll racial is very good. It's the better DPET of all race.
Orc racial is the better. Deep Freeze is 5k above the other and elem's dps is 0.2k above (2500 versus 2300).

I've had a talk with a GM to confirm that Orc Mages receive 5% dmg bonus to their Water Elementals. Turned out that only Warlocks, Hunters and Dks get the bonus (according to him).

//Edit: Wrong! See update: Cataclysm Mage Simulators and Formulators

Nathyiel · 03/09/11, 8:51 AM

+5% dmg for the Pet is equal to +0.5% to the global DPS. If we remove it, we have the same DPS for Troll and Orc.

Another little upgrade is to use the 65 Mastery on Hands in place of 50 Hastes. It's about 80 dps gain.

Finally, I have found an issue with the "Unbuffed" stats. The Spell Critical is taking account of Molten Armor (2% buff) but not Piercing Ice (3% talent). The Spell Haste don't show Netherwind Presence (2% talent). I can confirm that is correctly used in the calculation, it's only an error on the calculation of the Unbuffed Stats.

Zimeron · 03/09/11, 10:45 AM

Originally Posted by Nathyiel

Finally, I have found an issue with the "Unbuffed" stats. The Spell Critical is taking account of Molten Armor (2% buff) but not Piercing Ice (3% talent). The Spell Haste don't show Netherwind Presence (2% talent). I can confirm that is correctly used in the calculation, it's only an error on the calculation of the Unbuffed Stats.

I corrected the reporting in the code. In doing so, the pet saw an increase in their critical strike chance, due to Piercing Ice. Now in SVN we have the pet's crit chance benefiting from Molten Armor, Focus Magic, and Piercing Ice. If this isn't correct, let us know so we can get it changed. I also removed the Orc racial benefiting the Water Elemental and we didn't have it benefiting MI either.