Arcane simulation spreadsheet

[Vog]

Hi all,

I've been working on a spreadsheet for Arcane to test out various gear, buff and rotation scenario's. Simulationcraft does a great job at this already, but to me it feels like a black box as I don't know C++.

The benefit of Excel is that it's more accessible and audit-able, but also more limited in calculation speed so I had to make some sacrifices. One major sacrifice is that I calculate everything at the beginning and end of each cast, rather than every (1/60)th of a second. Another is that I sometimes calculate based on a snapshot of the row above to avoid circular references.

Things I've implemented:
-2-piece tier 16
-4-piece tier 16
-PBoI and KTT trinket (chance to proc, amp factor, proc itself)
-Burning and Sinister metagem
-Legendary cloak
-Bloodlust (on pull or at the end)
-Arcane power (on cooldown for now)
-Raid buffs
-Mana regen model
-Mana adept
-spell and dot crits
-Arcane Blast, Missiles, Barrage
-Arcane charges
-Living bomb/nether tempest

Things I haven't yet implemented:
-Alter time
-Multiple targets
-AoE spells
-Hit model (might skip it since it will be obsolete in 6.0)
-Managem
-Jade serpent potion
-other trinkets

You can download the file here, be aware it's still a rough alpha version:
https://drive.google.com/file/d/0B1zYPJ ... 2hsTG9XZHM" target="_blank

Feedback and double-checking of my calculations would be much appreciated. <3

31-05-14: added the ability for individual missile procs to provide a chance to proc cloak, metagem, trinkets.
31-05-14: hacked in the possibility to test 4-stack camping AB and AM, see calculation sheet column DV.

[Frosted]

Are spells getting two chances to proc trinkets, one roll at the start and one at finish?

[Komma]

Obscurity is in the eyes of the observer, not the creator, haha. I've spent an hour looking at your spreadsheet, but it's been a frustrating experience figuring out how everything works!

I due to your implementation of AM and Mage Bombs, I can't help but wonder how you're modelling individual missiles or bomb ticks triggering RPPM effects such as the cloak, meta, jade spirit, or trinkets.

In the sample calculation you uploaded it with, I also see "spamming missiles at 4 stack until you run out" behavior, which is rather different from what most mages are doing. I looked at the hidden MMBB vs MBMBarr sheet, but I don't really understand it, and don't see how it generates the conclusions written there. One thing that caught my eye is how you're comparing two spell sequences that both drop stacks with ABarr, when the goal is to avoided doing that and last a lot longer than 6-7 spells.

Still though, I'm really impressed by this spreadsheet!

[Vog]

Are spells getting two chances to proc trinkets, one roll at the start and one at finish?

I only do a roll at the finish, you can take a look at Calculation sheet cell CN5 to see an example.

I due to your implementation of AM and Mage Bombs, I can't help but wonder how you're modelling individual missiles or bomb ticks triggering RPPM effects such as the cloak, meta, jade spirit, or trinkets.

This is a great point, so far I've only modeled those as one chance to trigger upon spell finish, because the timeline is based on cast times rather than on milliseconds, I can't model anything that happens between a spellcast and a spellfinish. I'll have to think about implementing this, but I'm afraid that even if I model the ticks or missiles triggering, the buff gained would only start at spellfinish

In the sample calculation you uploaded it with, I also see "spamming missiles at 4 stack until you run out" behavior, which is rather different from what most mages are doing. I looked at the hidden MMBB vs MBMBarr sheet, but I don't really understand it, and don't see how it generates the conclusions written there. One thing that caught my eye is how you're comparing two spell sequences that both drop stacks with ABarr, when the goal is to avoided doing that and last a lot longer than 6-7 spells.

Very early into development of the spreadsheet I did a test to check if alternating AB and AM at 4 stacks is more dps than AM AM AB (2-pc) Barrage, and it turned out the latter was better. The extra chance of proccing AM didn't appear worth the high mana cost of AB4 and also your AM's are stronger due to queuing them in a row, benefiting more from Mana Adept.

The way I created the 'rotation' is basically to write down which spell to use in which combination of AM procs, Arcane charges, and 2-pc procs, in order to retest the 4-stack camping I'll have to add the previous spell cast to the permutations, which would rapidly multiply the amount of possible scenario's, but I'll see what I can do.

[Komma]

Very early into development of the spreadsheet I did a test to check if alternating AB and AM at 4 stacks is more dps than AM AM AB (2-pc) Barrage, and it turned out the latter was better. The extra chance of proccing AM didn't appear worth the high mana cost of AB4 and also your AM's are stronger due to queuing them in a row, benefiting more from Mana Adept.

The way I created the 'rotation' is basically to write down which spell to use in which combination of AM procs, Arcane charges, and 2-pc procs, in order to retest the 4-stack camping I'll have to add the previous spell cast to the permutations, which would rapidly multiply the amount of possible scenario's, but I'll see what I can do.

I question this result, because arcane charge building from scratch is the biggest cost to DPS we can have. The way you handle missiles also undervalues it, because AM doesn't snapshot. AM ticks will do increasing damage as mana regen gives us increased mastery effects. ABAM weaving is slightly mana negative, but bomb refreshs, RoP recasts all give us room to easily return to 100% mana.

You might want to check simcraft's implementation of Arcane Missiles for reference: https://code.google.com/p/simulationcra ... e.cpp#1503

After looking into the spreadsheet some more, I think you might also be missing temporary haste buffs providing increased mana regen due to Nether Attunement. Time warp and meta gem procs in particular should make huge differences.

[Vog]

I question this result, because arcane charge building from scratch is the biggest cost to DPS we can have. The way you handle missiles also undervalues it, because AM doesn't snapshot. AM ticks will do increasing damage as mana regen gives us increased mastery effects.

Good point on the snapshotting of AM, I agree I undervalue it, although I wonder if it's significant; even in the ideal situation of BL + frost armor + Tempus repit proc I regenerate 12.9k per second, which adds 3.6% to my missile damage per second.

After looking into the spreadsheet some more, I think you might also be missing temporary haste buffs providing increased mana regen due to Nether Attunement. Time warp and meta gem procs in particular should make huge differences.

I'm taking these into account, see Calculation sheet cell I5 and H5.

Thanks a lot for your feedback, I added the mechanic of missile ticks having a chance to proc trinkets etc, although still slightly flawed as the buff is applied at the end of cast. Not sure how to do it for dot ticks yet.

Rotation wise it's indeed odd I get a different result than the consensus, but I wrote it off thinking people never proved it. I'll see what I can do to test the 4-stack alternating AB and AM rotation.

[Shangalar]

I have no knowledge whatsoever of how to design simulations, but if possible try programming the following situations:

1 - AM AM AB ABarr without any active procs
2 - AM AM AB ABarr with intellect procs
3 - AM AM AB ABarr with Tempus Repit
4 - AM AB AM AB ABarr with Tempus Repit and intellect procs
5 - AM AM AB ABarr with Time Warp without other procs
6 - AM AB AM AB ABarr with Time Warp with other procs

I'm not sure now because after so much time I do this instinctively, like driving on familiar roads, but that looks like how I play. Haste isn't the only prerequisite for alternating Missiles and Blasts, I need an intellect proc with a haste proc to do it. Plus this also assumes that it was a rotational cycle started around 95% mana, for lower mana levels it's always AM AM AB ABarr. And Mage Armor of course.

[Kver]

Awesome work. It would be nice to reach a conclusion on 4 charge behaviour since it is one of the main key points of Arcane I feel. I actually always alternate as long as my mana allows it.

[Vog]

Thanks for the feedback, added a "was last spell AM4?" scenario so 4-stack camping with alternating AB and AM can be tested. Turns out it's indeed significantly higher dps than AM AM AB Abarr, I guess my early testing lacked some factors that made 4-stack camping superior.

I have no knowledge whatsoever of how to design simulations, but if possible try programming the following situations:
1 - AM AM AB ABarr without any active procs
2 - AM AM AB ABarr with intellect procs
3 - AM AM AB ABarr with Tempus Repit
4 - AM AB AM AB ABarr with Tempus Repit and intellect procs
5 - AM AM AB ABarr with Time Warp without other procs
6 - AM AB AM AB ABarr with Time Warp with other procs

I've been thinking of adding something that allows you to fill in a separate rotation in case you have these procs. The problem is that I use a vlookup concatenate of every combination of

• Arcane charges
• AM procs
• 2-pc procs
• Is bomb up yes/no?
• is previous spell AM4 yes/no?

for every condition that I add the amount of possible combinations is doubled, if I add too many then the priority list will be too cumbersome as you will have to through hundreds of combinations and assign each spell to use in that situation. I'm not really sure how to solve this yet, I want to add at least:

• is mana level below X yes/no?
• # of targets

[Kver]

Am I the only one who is amazed that you can do this with Excel..?

I think the rotation can be simplified a bit more than Shangalar suggests, having different scenarios for ONLY Time Warp and Tempus Repit makes no sense because they have the same intensity, just a different duration. The exception is real Burst Haste scenarios but that is the 'big' exception regardless.

It might also be easier to just keep it purely single target I suppose, because multiple targets means bomb uptimes, which can give AMs, etc..

[Frosted]

I have no knowledge whatsoever of how to design simulations, but if possible try programming the following situations:

1 - AM AM AB ABarr without any active procs
2 - AM AM AB ABarr with intellect procs
3 - AM AM AB ABarr with Tempus Repit
4 - AM AB AM AB ABarr with Tempus Repit and intellect procs
5 - AM AM AB ABarr with Time Warp without other procs
6 - AM AB AM AB ABarr with Time Warp with other procs

I'm not sure now because after so much time I do this instinctively, like driving on familiar roads, but that looks like how I play. Haste isn't the only prerequisite for alternating Missiles and Blasts, I need an intellect proc with a haste proc to do it. Plus this also assumes that it was a rotational cycle started around 95% mana, for lower mana levels it's always AM AM AB ABarr. And Mage Armor of course.

There is an old thread on the EJ forums about doing exactly this type of thing. I'll see if I can dig it up.

Alternatively, you could go in and edit the SimC APLs to respond to each of those scenarios and see how the DPS changes.