Well I've given it some thought on the matter of balancing classes and gameplay in general. There has been a great deal of conversation on this issue given Wizard's work on the D&D next. Even Monte Cook recently asked the question, "It is difficult (not impossible) to create a character creation system for an rpg that allows for customization of capabilities that doesn't greatly encourage optimization." By optimization we mean min-max issues that lead to over specialized characters that unbalance the game or take too much "action" from the others.
This begs me to ask the question: If you could would you be able to prove it?
Ok, so what if I can come up with a character creation system that allows for customization without encouraging optimization. How do I know it discourages optimization given all the possible combinations a character may take? Do I play test all possible scenarios? Well that is truly impossible. It's like knowing all the RPG adventures that will ever happen before the rules go to print. Or limit adventure/character options to such a limited set the game becomes boring. Which is against the ideals of RPG in the first place.
So this got me thinking about the whole objective of building a character generation system and its relevance in game balance. I arrived at the conclusion that it is totally irrelevant to game balance. It might appear so, but in truth it isn't. It's like shooting and unguided rocket to the Moon and hoping it lands safely there or at least enters Moon orbit. It won't. It will fly towards it. It will seem to get close, but it will miss by a long shot. No matter how good you do the initial trajectory calculations without onboard flight control you will miss the Moon.
The secret I believe is to create "onboard flight control" for the RPG. That means less concentration on the character generation and more concentration on the gaming side. Define "balance" in the game and give it some numeric value. Then rank characters against it and keep tabs on things as they evolve. This requires a different character sheet system (or recording system if you will). One that keeps tabs on how strong the character is. Numerically!
So far I've seen no system that does this. Point buy systems provide XP or means to adquire more skills, but there is no clear "feedback" link between the amount of skill points given and the way XP is supplied. Class based systems rely on levels to "supply" skills which are acquired at each level and the level-class combo is presumed to be balanced at design time when in truth it isn't so. Because skills can cause unwanted synergies and not all skills are the same even if they can be acquired at the same level. There is no "skill weight" if you will allow me the expression. With such a value you could rank the character, but D&D lacks it.
So I propose the following. Quit trying to balance the game from the character generation point of view and look to balance it during game play. One way I see of doing this is to create and index or value that shows how strong the character is.
To implement this index I coined a term called XP Tax. Once a skill is added to the character the character's XP tax is increased. The XP tax is in turn a % of the XP gained in every adventure that is "lost" to upkeep of current skills. The greater the skill count the greater the tax and the slower the progression is. It is also a good index to measure characters with. If you have a party with members having 18%, 22% and 47% XP tax you have a balance issue. This XP tax mechanism provides what is known as "negative feedback" on the character progression mechanism. It is negative because the stronger the character the more negative the feedback value (increased XP). In spite of the term "negative", negative feedback is a good thing. Like the rocket shooting to the Moon. The negative feed back in the guidance system works to reduce the error between the rocket's path and the Moon's position. In a similar way this feedback keeps the character at check.
So now instead of best guessing the character class progression during class design time you have a mechanism that allows you to measure the class as it advances. More so it "decouples" skills and skill design from character class design. If at some point you find out that some skill or power is actually too strong you just need to increase its XP tax to account for its bigger impact on the character class. So you can fine tune the system during playttest without having to go back to the character class drawing board, and you don't have to redesign the whole class to fix an issue. This is what "decoupling" means. Things are no longer tightly bound together and small changes to the parts don't affect the whole as much.
With this it is an easy matter to create an exponential XP tax progression for some skills. For example basic proficiency in a weapon might cost you 1 XP tax point, expert 2, master 4. Two weapon fighting might double your XP cost for the used weapons. So if you have 8 XP tax for long sword and 4 for dagger thats 12 XP tax points base. Add two weapon and that becomes 24 XP tax points. That's a lot of XP lost every adventure to have such a skill. It's a great skill to have, but it has got to be worth it for the player. Maybe those 12 points would be better used somewhere else?
The graph below shows the impact of XP tax on various XP incomes and the possibility of purchasing skills. This measures the skill purchasing power and it is plotted on a logarithmic scale. It goes from 2 on the outer rim to -2 on the center. The inner circle represents 0 or "unity gain" (one hundred times less progression speed than the outer rim). Each color represents an average XP income per adventure. Starting characters would average 1000 XP points and as they progress they will be able to tackle harder adventures that bring in more XP (2000, 4000,... etc). The percentage values represent the XP tax. The graph starts with 5% a very low tax value representing starting characters. As characters increase in skills their XP tax increases as well going counter clockwise on the graph.
If a starter character with 5% XP tax could bring in 32000 XP points per adventure it's progression would be monstrous as you can see the light blue triangle is nearly on the 2 (100 times faster than the inner circle of 0). But lets be realistic. How can such a weak character gain so much XP in a single adventure? Not going to happen. So we take the more realistic value of 1000 XP points which is closer to the inner circle of unity progression, but slightly above it.
As it gains experience the character then increases its skills and goes up to 10%, then 15% and then 20%. At that point the blue square has fallen nearly to the inner circle and the orange diamond is now where the blue square used to be. XP tax is such that progression at a rate of 2000 to 4000 XP per adventure is now equivalent to the early 1000 XP tax. This coupled with ever more expensive skills keeps the system under control. If players spend too much on certain skills their XP tax can skyrocket too early in their game and they find themselves doing an uphill climb from there on. If the points cross the inner circle to negative values the progression is practically impossible. So it is good to keep the number between the inner circle and halfway to the outer circle.