The old Mage spreadsheet comes out, as I'm at a point where I'm contemplating spec choices again.

http://www.sigilguild.net/hamlet/Mage2.xls

Not very user-friendly yet, but should be very functional for people looking to examine deep theorycraft. Just point out any mistakes. So far, it's based on the info people gave me on the TBC Mage Theorycraft thread.


EDIT: User-friendliness update. Basic usage:
Shaded areas are for user editing.
1) In either column 1 or 2 (or both, if you want to compare two setups), enter your gear bonuses, spell choice (Fire, Frost, or Arcane), talents, and buffs.
2) In the box labeled "miscellaneous," there are a few more parameters, most notably an estimate of your casting lag.
3) In the box on the right "Fight length" you can choose which fight durations you want to compute for, and whether to account for evocation, gems, or Major Mana Potions.
4) In the same box, read the computed DPS numbers for both setups over all of the fight durations.

General note:
Accurately estimating caster DPS from first principles is quite hard; there are a number of odd factors at play. The resulting inaccuracy is much greater than for, say, Rogue DPS. However, the relative information (comparison of different builds, behavior of a particular setup over different fight durations, the marginal values of different talents, etc.) should be very accurate, and that's the main function of the chart.

------------

Not currently in:
--Timers (trinkets/combust/WE/etc.)
--non-Clearcasting builds (will work on that; it's kind of baked-in right now).

------------

More advanced:

The confusing part is "casting cycles." These are the basis of my Mage theory. Basically, a casting cycle is a mode of operation, defined by:
1) What to I do on a Clearcast?
2) What to I do on a non-Clearcast?
If you look in the middle of pages 2 and 3, there's a block of 16 cycles, with DPS and MPS computed for each (cycle 1 is Frostbolt/Frostbolt, cycle 16 is wanding). Many of them may be obsolete now, due to the downranking nerf.

It's expected that, in a given fight, you will use two cycles to maximize your damage while running mana precisely to 0 (usually, one high-mana cycle and one low-mana or regen cycle). The mathematical justification for this is here: http://forums.elitistjerks.com/viewtopic.php?id=7742.
The chart lets you input four such pairs of cycles for each setup, and will return the highest DPS of the four. You input the cycles you want to try out on page 1.

If it becomes clear that Mage downranking is definitely obsolete, I can simplify it a lot.

Example:
In the currently uploaded version, I have my current Fire setup in column 1, and my planned Frost setup for tomorrow in column 2. The gear/talents are input into each column.
In the block labelled "cycle pairs," I've input four pairs of cycles that might be reasonable to compute for each build. In the Fire build, the first is 10/11, where 10 is Scorch/Fireball, and 11 is Fireball/Fireball. The fourth is 12/16, where 12 is Fireball/Missiles, and 16 is wanding. If you look at the block on the right of page 2 or 3, it will compute the DPS of each your four cycle pairs for each fight length, and return the highest one to the front page. So if you ever want to see which spell choice is responsible for the reported DPS on the front page, you have to check that block.

Some things the posted chart is telling me right now:
--My maximum DPS over a 60s period will drop by about 34 tomorrow, when I respec Frost.
--My maximum DPS over a 180s period without Evocation will increase by about 41. If I examine pages 2 and 3, I can find out why. In the Frost build, I can chaincast Frostbolt for nearly the full 3 minutes, but in the Fire build, I've reverted to a Scorch-based cycle over 70% of the time.
--In a Patchwerk situation (6min/evoc/3gem/3pot), I can expect to chain Frostbolt the entire time. Right now, I'm on cycle 10 (scorch-based) 45% of the time.


------------

For simpler usage, you can input one or two setups, ignore the whole "cycle pairs" thing, and just look on pages 2 and 3 to see the DPS and MPS of every cycle. That's a good quick way to evaluate different setups without the whole segment that automatically relates mana usage to fight length.

If I use the same spell for w/ CC and w/o CC I get a bunch of DIV/0 errors. You're dividing by delta in MPS of both spells and in this case it is zero.

Also the stats calculations are for humans only I guess?

As far as I could determine the correct formula for stats is:

INT((INT(Base*(1+ratialBonus)*(1+talentBonus))+GearAndBuffs*(1+ratialBonus)*(1+talent Bonus))*(1+kingsBonus))

You still have Arcane Mind as increasing mana, now it increases Int.

Sorry, to be more clear:

If you want to say, "I'm casting Frostbolt and only Frostbolt," put a "1" in for one cycle, and pair it with some trash cycle (You're assuming here that the mana constraint doesn't bind). The "1" by itself signifies "cycle 1, in which Frostbolt both on CC and on non-CC." You need to put something different to pair it with, because the math tries to hit both the time and mana constraints. If the cycle you pair it with is irrelevant, then it will return 100% cycle 1 and give you the number you're looking for. If it returns less than 100% in cycle 1, then the cycle you picked actually improves your DPS.

Yeah, human only for now, but I'm pretty sure it's correct.

Thanks.

I'm really confused after looking at this chart. One thing I noticed is that you don't have the option for clear casting in the talent columns, I take it too mean that you are assuming the mage has clear casting?

And secondly this cycle thing, I'm not getting what your numbers are showing.

Yeah, haven't included non-Clearcasting builds yet. When I first made this up, it wasn't really a reasonable raid option.

-------------------------------

Ok, here's what's going on. My goal was to not only compute the DPS of certain chaincasting patterns based on gear/talents, but also to figure out what the maximum possible damage over a certain fight length, based on mana.

So, say my base cycle is Fireball/Fireball (Fireball on both clearcasts and non-clearcasts). The chart computes the DPS and MPS of that cycle, based on you setup. You will find in column 11 of the large center block on page 2 (for setup A) or page 3 (for setup B). This is the first way to use the chart--just input your gear/talents, and read off the DPS/MPS of any cycle by looking at the corresponding page.

I should probably make that part more prominent, actually, since that's often all somebody is interested in.


The more complicated part I'll try to explain better at some point. The basic idea is that instead of simply evaluating the damage and mana usage of one cycle, you choose two, and the spreadsheet fits them together to do as much as damage as possible with a certain amount of time/mana.

Some explantion is here:
http://forums.elitistjerks.com/viewtopic.php?id=6825

------------------------------

Can Mages who have played TBC/Test much tell me whether downranking is ever considered as a serious option anymore? If it's not, I think I can simplify things greatly and eliminate the part where you need to input your cycles up front, since that's what most people find confusing.

Okay, I simplified the usage somewhat. Now, instead of having to input the cycles yourself, it simply asks you whether you're mainly going to be casting Arcane, Fire, or Frost, and uses a predefined spell set for each. Advanced users can edit the spell choices on the first page.

Have I missed something, or is it really impossible to enter a spec without Clearcasting in the spreadsheet?

That might be an interesting addition, since a 2/46/3 fire spec seems a viable choice for raiding in 2.0... It's what I'm planning on taking, anyway, and I was interested in seeing how it measured up when you factor in everything.

At the moment, it is, sorry. When I first made this, non-CC specs just weren't serious raid builds in any case. Unfortunately, since CC has such complex effects, its presence is really baked into the machinery of the sheet. Next time I look at it, I'm going to see if there's way to handle that without a major overhaul. I'd like to see the potential numbers of something like 0/48/3 as well.

Looking ahead: what do people think would be a good way to handle Arcane Blast usage? It's basically the first time that Mage behavior might be something fundamentally different from "chaincast one spell, possible reacting to Clearcasts."

I haven't raided in beta, but in 5-mans, the only time I'd downrank was to use Rank 1 Frostbolt to snare a mob or get Winter's Chill up.

From a coding point of view, it might be easiest to derive a function of time that returns the damage/dps/dpm of arcane blast. It should be doable in Excel I think. But I don't know if it'll affect that optimization algorithm you use since the dps and dpm of the spell isn't independent of the duration of that stage.

The answer will probably just be to implement custom cycles for any casting patterns that people that using. Such as "3 Frostbolts, then a first-order Arcane Blast, then repeat." Or "third-order Arcane Blasts, with Missiles on Clearcasts." I'm guessing there won't anything much more complicated than that. In the latter case, the ramping-up of AB is an edge effect to just ignore, much like Imp. Scorch.

thanks for the great work. i was wondering how you are going to factor arcane blast into your spell cycles? it might be worth looking into both the effectiveness of chain casting it and alternating between arcane blast and say arcane missles.

I made a quick note on this above, haven't thought about it much more since then. It's still a little while until the 70 raid game, and we'll have a much better familiarity with Arcane Blast by then.

As to errors pointed out above: I don't think I've fixed either of those, and probably won't for a little while--out of the country at the moment.