1) If it has a casting time (or is any mage spell), and does direct damage, the base coefficient for the DD is casting time before talents / 3.5.
2) If it's channeled and is not a mage spell, and does direct damage, the base coefficient for the DD is casting time before talents / 5.0. (Apparently there are some exceptions tho, such as Starshards.)
3) If it has a DoT, the base coefficient for the DoT is duration before talents / 15.
4) If it's both direct damage and a DoT, the direct damage part gets its base coefficient multiplied by DD / (DD + DoT), and the DoT part gets its base coefficient multiplied by DoT / (DD + DoT), where DD = average direct damage before talents and gear, and DoT = average damage over time before talents and gear. (NOTE: fireball seems to be an exception to this rule; there are probably others.)
5) If it's AOE, multiply by 1/2; except for healing spells multiply by 1/3. In some cases, apparently when the effect lasts for 8 seconds, additionally multiply by 0.83. (Note: this used to be just 1/3 for everything before patch 2.0.3, after which things make much less sense.)
6) If it has a snare component (frostbolt, mind flay, etc), multiply by 0.95 or 0.9.
7) If it does both damage and healing (life tap, devouring plague etc), multiply by 0.5. (NOTE: only in some cases - compare life tap, siphon life, drain soul, vampiric touch etc.)
8) There are probably other rules as well but I can't think of them right now.
First off, even if a spell gets 0% of the character's spell damage, it will be affected by talents, debuffs on the target, etc. Example: The DoT portion of a rank 12 fireball normally ticks for 19 damage, but with both the fire power (+10%) and playing with fire (+3%) talents, it will tick for 21-22 (19 * 1.1 * 1.03); and with improved scorch (+15%) on the target it will tick for 24-25 (19 * 1.1 * 1.03 * 1.15), etc.
Second, all damage (and healing) increasing talents, as well as all mana cost decreasing talents, now seem to be multiplicative (see above). All empowered talents however are additive, as said on the talent: "gains an additional X%". For example, the normal coefficient for arcane missiles is 5/3.5, and the max rank of the empowered talent increases this to 5/3.5 + 0.45. One exception is the druid talent empowered rejuvenation, which increases the relevant coefficients from X to X * 1.2; this also affects the initial heal of regrowth. Other exceptions might exist.
Each rank of a spell has a "maximum level"; this means the level before the next rank of the spell is learned. For example, Dragon's Breath rank 2 has a max level of 63, as rank 3 is learned at 64. However there seem to be some exceptions - with some spells and ranks, the max level differs by up to 3 in either direction (source: Gagorian: personal communication). This isn't entirely clear yet and more research is needed.
Anywho, assuming we know the maximum level of each rank of each spell, the following coefficients then apply:
Healing spells: (max level of spell + 6) / player level
Damage spells: (max level of spell + 4 ) / player level
That is, the final effect of the spell is multiplied by this coefficient.
These are from Gagorian and are (apparently) tested and verified by him. Some discussion, though not necessarily on this matter, can be found here.
Deviations from the base damage
Spells (well, at least damage spells - probably healing spells too) scale when you go up levels. When you learn a spell, its base damage (before talents, gear, etc) is what you see on Thott or Alla; however, when you then level up, the same spell will do slightly more damage, around 1-10 points. There doesn't seem to be one simple number to explain how much additional damage each spell receives; rather, from some preliminary testing done by Gagorian indicates that most spells have two unique coefficients that determine this scaling, according to a common formula. Note that the game's default tooltip does not take all damage-increasing talents into account (but does take some), and is therefore also inaccurate in reporting these differences. In addition there are rounding issues (we do not know which values are rounded and how). All in all, unless you just learned the spell and didn't level up after that, it's difficult to tell exactly how much damage it's supposed to do... until someone (myself, Gagorian, whoever) figures it out. ;)
TODO: non-shadow priests; druids; shamans; warlock pets; damage spells downranking; and of course all the unknown and uncorfirmed stuff
They appear to be moving away from case-by-case and towards an absolute system with 2.0 though. Mind Flay, which has always been lower than any system would dictate, now fits within the above. (3/5*0.95 = 57%)
I haven't seen a good table anywhere. If one doesn't exist, perhaps it would be a good idea to compile one and link it from the threads of significance? I'll do some quick tests to see if the priest spells all follow the pattern.
There are more, but those are the ones I remember right now.
Lifestealing: 0.5. Health gained from lifestealing is affected by neither +/- healing gear, +/- healing effects like Wound Poison or Amp. Magic nor is it affected by % healing effects like Mortal Strike. Incidentally, lifestealing does not generate healing aggro in WoW 2.0.
Freezes have a very small modifier of as yet indeterminate size.
Channeled AoEs have an additional 0.33 modifier - e.g Blizzard gets 8/3.5*0.33 = 0.76 modifier (spread over 8 waves).
Not sure about AoEs with burst components or Drain Soul... if Drain Soul got a 15/3.5 modifier, I would be vastly amused.
I actually sat down and collected data to come up with numbers for a mage guide for Ergo. I found something I didn't understand too and meant to ask you guys anyway, so it's a convenient thread!
The way I measured most of them was to go into Deadmines and AoE a whole lot. I mixed up the number of targets cause I wanted to confirm that the AoE buff/nerf wasn't live, which I confirmed (if it were live, the damage range in practice would be much higher than the tooltip damage range). I line up the damage range from the tooltip and from what I measured and just do the math from there.
Arcane Explosion - (1.5/3.5) * (1/3)
Cone of Cold - about .182965 (116/634). It's not exact since the damage range on my data was only 40 as opposed to 43 for the tooltip. However, this coefficient is pretty close to 1.06 * (1.5/3.5) * (1/3) * 1.20 - which seems to say to me that the 1.20 was the part added by Blizz when CoC was buffed.
Blizzard - (8/3.5) * (1/3) * 1.06
Flamestrike - (87+44)/553 total, 87/553 on the direct damage and 44/553 on the DoT. (see below)
(Can you tell I was writing a section on optimizing AoE dps and dpm? heh)
So the thing with Flamestrike is that the total gained from +damage was lower than the expected (3/3.5) * (1/3). The closest coefficient I could get from using "sensible" numbers was (3/3.5) * (1/3) * (3/3.5) * 0.95. It was closer with the 0.95 than without. I thought to throw that in there from the Frostbolt coefficient even though it made no sense. I could sortof understand if Blizz had a little bug and counted the cast time twice somehow, but well, whatever, I'm stumped.
I didn't verify how I think Empowered talents work or the old coefficients for main spells, but I did some analysis of the Water Elemental for the same guide.
I used a lazier methodology for this - combat log my WE in Silithus with unchanging gear/buffs and tell WWS that he's a mage, then compare the max and average damage to the tooltip. Lining up the max damage, he got 27.6% of my +frost. Lining up the average damage, he got 26.8% of my +frost. I only had a sample of 43 Waterbolt casts to work with, the average and max alignments should be equal with a larger sample size. I figure human coders like nice round numbers, so I call it 27% or 27.5%.
While I was at it, I was curious whether it got any of my intellect, so I summoned one with my damage set on and one with my mana set on (it turns out you can swap gear while it's out and it's max HP/mana change accordingly). I found that it's mana was 1889.5 + (my int) * 1.5. In other words, it gets 10% of my int to use for the normal mana formula.
Anyway, if anyone wants more of the details of the data or collection or whatever, I've got those sitting around.
I've been getting some confusing numbers, unless there's known errors with Spelltips i've been occasionally hitting for higher than should be possible with Frostbolt and Fireball (not in molten fury territory) which has led me to be completely clueless about the changes in 2.01.
Im pretty sure CoC and Fire Blast are 43%, thats what i remember it was before the patch and I dont think it was changed. I believe it was 43% (rather high for an instant spell) because it is on the global cooldown and with a seprate cooldown of its own.
I was indeed looking for a table; I'm aware of the generic guidelines for the derivation of the coefficients, but I'm also aware that they are (now much more so than before) just guidelines and not accurate. :)
I'll go do some preliminary testing on my mage now; I'll edit my initial post later with any significant info I get (or leech from you guys, heh - thanks).