So how does Blizzard do it
Ghostcrawler from the Blizzard forums has posted a few articles now about class balancing. Here is just on of the good ones trying to explain the difficulty of class balance.
"Holy expletive! DK dps is through the roof. They're getting 10K SS crits in greens!"
"What do we do?"
"Geez. I don't know. Let's just change a bunch of numbers and hope something sticks."
"Okay, how about Scourge Strike? Should it be 20% of weapon damage?"
"Or 15%? I have no idea. Flip a coin?"
"What about Might of Mograine?"
"Just pick a number. I'm sure it will be fine."
"Should we compare them to other classes?"
"No time for that. Just make some changes and we'll try again."
That's not actually how we do our jobs. ;)
We have spreadsheets -- huge ones. We have values for characters in greens and raid buffed with epic gear. We have estimates of mob damage and health, character downtime, the benefits from enchants, gems and glyphs. We have conversion rates of mana to runic power to energy to rage. We look at damage per second, efficiency, button presses per second, typical rotations, movement vs. standing still, mobs vs. players and every other variable we can think of (and there are a LOT of them).
In the end we come up with an estimate for what an attack should do. We come up with a budget for talents and spells just like we have a budget for items at a given level. Sometimes those estimates are wrong because we forgot to take something into account, or because there's a bug in a talent or spell somewhere that messes up the calculation.
And then we do lots of testing, and to get a reality check on our tests, we compare them to the numbers people are reporting from the beta. Repeat as necessary.
I've said this in a few posts: that our numbers can sometimes be wrong (as is the case almost any time you deal with numbers), but the methods we use to arrive at them are absolutely not sloppy.