TLDR Version:
Fully improved and empowered talents work as follows:
Frostbolt: 82.2857% = (((3/3.5*0.95) {base 81.x%} + 0.1 {empowered}) * 0.9 {improved})
Fireball: 103.5% = ((1 {base 100%} + 0.15 {empowered})* 0.9 {improved})
AoE coefficient is 0.5, consistend for all mage spells.
Some Waterbolt testing.
Note to Kalman about odd cycles/rotations.
TLDR End.
Hey Vontre (and all other mage TCs),
first a quick note about waterbolt. In your sheet, this line is for the bonus damage on waterbolt from your spell damage gear.
=G447*'Game Stats'!$B$93+WENN('Main Sheet'!$B$47="yes";'Game Stats'!$E$37;0)+F$8+WENN('Main Sheet'!$B$27="yes";'Game Stats'!$E$14;0)
This looks pretty messy to me:
(fireballdamagebonus*WEmultiplier + ((elixir of frost power)) + frostdamage + wrath of air totem).
I'll hop back to the PTR to check the coefficents, it should be something like this:
((damage+frostdamage)*Water_Elemental_multiplier + wrath of air totem{The elemental gains double benefit, right? The totem buffs its damage directly and indirectly because you get buff and increase the elemental's damage})*Waterbolt_multiplier)
Went back to the PTR and checked out waterbolt:
Tooltip at level 70: 224 mana, 2.5s cast, 486-558 damage (522 average, 72 delta) (note: wowwiki has a typo, top end damage is not 588 but 558)
(For some reason, you entered the base damage as 464.5 in your sheet)
Test with +0 spell damage (no spell school damage either): 486-557 damage on waterbolt. So we know it has no hidden scaling with your/your pet's intellect/spirit/whatnot.
Test with +1000 damage/frost damage (had another +50 arcane/fire fromSunfire):
764-835 damage (71 delta), that means +277-278 damage from gear, i.e. 27.7% +/-0.2% from your gear is added as damage.
Assuming waterbolt has a 2.5/3.5 coefficient and the elemental gets 40% of your spell damage, we'd get a coefficient of 28.57%. Close but not exact. 40% WE, 69.5% WB would work, but makes no sense. 39% WE, 2.5/3.5 WB isn't any better. I don't know what else it could be.
Then I spent half a day on the PTR (again

...) to test some stuff. I wanted to find out the current AoE coefficients, since some of the wowwiki data is outdated
Spell Damage Coefficients - WoWWiki, the Warcraft wiki What I stumbled over though was that I have always miscalculated the improved/empowered talents.
The empowered talent is added. The improved talent is a multiplier, and is applied after the empowered talent is added. So, for fully specced talents, the coefficients are as follows:
Frostbolt
(((3/3.5*0.95) {this is the base 81.x%} + 0.1 {if 5/5 empowered}) * 0.9 {if 5/5 improved}) = 82.2857%
Fireball
((1 {the base 100%} + 0.15 {if 5/5 empowered})* 0.9 {if 5/5 improved}) = 103.5%
I did the frostbolt test myself on the 2.2 PTR and can guarantee that it works that way.
Fireball is annoying to test (it has a 172 damage delta between high and low) and someone else did it before, the link linked from wowwiki is
WoW Forums -> So I chaincast frostbolt on the PTR today...
What I actually wanted to test were the AoE coefficients. The AoE penalty now is 1/2 instead of 1/3, which it was before. So, the coefficients are:
Arcane Explosion: 21.46% = 1.5/3.5 (instant) * 0.5 (AoE)
Frost Nova: 4.2857% +/-0.1% = 1.5/3.5 (instant) * 0.5 (AoE) * 0.2 (special multiplier)
Cone of Cold: 19.2857% +/-0.3% = 1.5/3.5 (instant) * 0.5 (AoE) * 0.9 (special for slow)
Blizzard: 114.2857% +/-0.4% = 8/3.5 (8 seconds channeled) * 0.5 (AoE), 14.29% per tick
Blast Wave: 19.2857% +/-0.2% = 1.5/3.5 (instant) * 0.5 (AoE) * 0.9 (special for daze)
Dragon's Breath:Cone of Cold: 19.2857% +/-0.3% = 1.5/3.5 (instant) * 0.5 (AoE) * 0.9 (special for disorient)
Flamestrike - the damage on the base spell is split 55% direct damage - 45% damage over time (+/-1%, couldn't figure it out exactly)
Flamestrike (direct): 23.57% +/-0.2% = 3/3.5 (3s cast time) * 0.5 (AoE) * 0.55 (split)
Flamestrike (over time): 12.00% +/- 0.3% = 8/15 (8s DoT, 15s is the normalised DoT duration for calculation) * 0.5 (AoE) * 0.45% split, 3% per tick
I did the test by not picking any damage multiplier talents, note the tooltip damage of the spell and note the damage delta (difference between minimum and maximum). I then tested the spell on mobs until I got a similar damage data of my test points (fixed +damage gear, no procs, no buffs, no debuffs, etc.) with ~1k damage gear. In most cases, my test data delta was 1 less than the tooltip delta, but it was larger for some spells that are horrible to test (self-centered AoE with cooldowns are not fun), those have a higher listed error margin in the list above.
Edit: To Kalman
Thanks for clearing up the cycles.
I think that actually, the best/only(?) way to optimise is calculate "AB spam" and all "X*AB/6.5s filler" cycles for X=0,1,2,3. If all the "X*AB/6.5s filler" cycles are sustainable (they should be for most fight with a support group), then the maximum should be part "X*AB/filler" and part "AB spam". For each X and each filler, the "AB spam" ratio is easy to compute (excess mana/mana burn difference between AB and the cycle).
3*AB/filler isn't always going to be your low-burn rotation though: Take a fire heavy spec wit 2/5 T5, and maybe toss in high hit/crit and fire damage, and make it a short fight (or have very high regen rates). Then your (1?),2,3*AB/2*Fireball cycles will be inferior to Fireball spam damage and mana wise.
Your optimal rotation then could be Fireball spam as DPM cycle and AB spam as burn cycle.
It's a bit counterintuitive and can only happen in very skewed/biased settings where only the 1.5s AB beats FB DPS.
Just some food for thought about what may happen in those cycles. (I'm not even sure if my example is correct or if I'm wrong at some point there.)