OK -- so there's a problem with this no class system, and that is that you can't easily implement a consider system. Traditionally in MUD, there was a consider command that you used on a mobile before fighting it to see if it would be too easy, too tough or just right.
The first thing is I've never seen a "con" system that actually works. The one in EverQuest and EverQuest 2, for example, doesn't take into account archtype, class, subclass or AA trees -- all of which are factors in how difficult or easy a fight is going to be. The one in WoW and the one in Warhammer are, similarly, broken. It's difficult any time you allow player customization at all -- so we need another approach other than looking at a flat level.
First, we're going to keep three scores. One is the number of wins or losses against easier oponents. Each time you win against an easy con, we add a point (saying we were right to con it easy). Each time you lose against an easy con, we subtract a point (saying we were wrong).
The second number is the number of wins against a medium con. Again, each win adds a point and each loss subtracts a point.
The third number is the number of wins against a hard con. Again, each win adds a point -- and each loss subtracts a point.
This forms an index that tells us how to modify cons. If you have a ton of points against hard con mobs, then we ramp down what we're conning the difficulty at. If you have lost a ton of points against easy con mobs, then we ramp the con down. If we're getting things perfect, we know it. Ideally, medium would be at zero, easy would be positive and hard would be negative -- and so we adjust the cons to try to achieve that, so that things that con as "even" give you about a 50/50 chance.
But how do you determine the initial con?
The answer is we do things differently from everyone else. Normally a stat -- lets call it strength -- feeds into a skill. You determine the stat at character creation, and maybe there's a way to move it through game play (gear, etc.). There's some magic stat that for your archtype you want to be maxed, and the other stats aren't so important.
So we flip the problem around -- the skills determine the stat, and the stat determines the archtype. But that's not sufficient -- there's a world of difference between an item you get easily, and an item you have to work for. Each item has skill requirements to equip.
Only skills that have the necessary gear are useable -- so you take the total of all those skills, and that's the "effective" stat. You total the "effective" stats, and that gets you a general power level. But again, that's not enough to get a good con. Someone who had only support abilities would con the same as someone who could instantly kill anything, and that would be bad. So the con has to be based on the rock, paper and scissors -- the strongest stat has to be compared to the stat it dominates, and the weakest stat needs to be compared to the stat it is dominated by. That makes it so that if the mob is really good against, say, ranged then it will con a lot higher versus ranged players than versus melee players.
This, then, helps the LFG system. A balanced group would have the stats equal, so that no one stat dominated. So if you're looking for more, the system can try to balance out the stats.
The key is the consider system must consider more than just the character level, and the consider system must consider more than just the points. Every single piece of equipment and the skill of the player sitting in the seat are factors, and if you're going to ramp the difficulty level to an appropriate level, you're going to have to take all of that into account. Otherwise there is no way to achieve a balanced game.
I want to emphasize -- all this is in generalities because I don't have specifics yet. But my personal opinion is that if you look at "Lord of the Rings" as an example -- each member of the Fellowship has their own way of doing things, but when they're fighting they are all equally successful. It shouldn't be about one guy repeatedly mashing the taunt button, while the rest very carefully avoid grabbing the mobs off of him. It should rarely be "just one guy," and boss fights should be relatively involved.
I have some more ideas -- some pulled from D&D -- that I'm going to share in a couple of days. The basic thought is -- the basic buckets should be "ranged," "melee" and "support." Stealth is applicable to all of those, but in different ways -- a melee stealth would be the backstabber, a ranged stealth a sniper, and a hacking stealth more of a spy. But I'm pretty sure the three buckets should be ranged, melee and support -- the reason being I can see a clear way to set up the rock, scissors and paper there that is fair to all of them. Obviously, certain choices will involve more strategy and thought, while other choices will be plow through -- but both choices should be equally effective. And that's really the key.
Anyway, I'll talk about it some more in a couple of days.