I think almost every balance problem Warband and cRPG had, currently have and will always have are related to the following two facts:
1.: Medieval combat was not about balance, if anything the opposite. You tried to keep it as unfair as possible, by all means you had, to get the upper hand over your enemy. This means it is not constructed around a "stone-scissors-paper"-system, which you can see by the different amounts of counters and prey classes the different classes have. Some are countered by more, some by less classes than they do counter. The high and late medieval knight, a highly trained, well equipped and usually really hard to kill fighter was the jack-of-all-trades at that time, at least the one which came closest to it. A single knight was always worth more than any other single fighter they had in their armies. Something which doesn't work in a MP game.
2.: The skill system introduced in M&B was not created with the idea of a MP game with constant characters, it's about SP. Different classes and builds need different amounts of skills and thus skill points to work. While a 2hd infantryman basically only needs power strike and weapon mastery to effectively swing his sword, a horse archer needs power draw, horse archery, riding skill and weapon master. Same for the equipment. A 2hd infantryman basically only needs his sword. To be a horse archer you (at least) need a bow, arrows and a horse. You can argue that with this basic setup (a naked 2hd infantryman without athletics vs a naked horse archer on a horse) the horse archer is more powerful than the 2hd infantryman, which represents the skill point cost. But since we are limited to lvl 30 in cRPG, and all players should be of about the same overall strength, the system breaks at this moment.
As we all know, you are not limited by the choice of skills, apart from the attribute requirement. But there does exist a limited amount of viable builds, especially due to the level cap, but also due to a few game mechanics concerning item weight, weapon proficiency and so on. There was a time when picking anything else than 18 STR and 6 PD for your archer was plainly stupid and inferior to this absolutely best choice.
I think it is interesting to give the option to try out different builds, but the choice is actually more limited than it looks like. You don't need to be a one man army, and usually you can't even be one, but you can be it more or you can be it less. If you want to be it less and more of a team player, supporter or something along these lines, you will undeniably be effective, and if people are playing properly with you you can even be super-effective, but usually on public servers a lot of frustration will be awaiting you. There are interdependencies between the classes, and sometimes strong synergy effects can be achieved, but again this varies from class to class.
As you notice I am constantly talking about classes, simply because I think there ARE classes. Your equipment heavily influences your playstyle, your possibilities and thus your role. You can have hybrids, but basically things stay the same, especially since a hybrid can always be only one class in a moment, never two at the same time. His advantage is that he can switch between those classes during the battle, giving him more flexibility at the expense of effectivity in his particular classes.
The last thing I like to mention is this goddamn market place, which actually entirely took out the equipment aspect of a build, since items are only restricted by their stats ever more, but not by something different. Short time after the implementation of the upkeep system I posted my alternative system where you can increase your (fixed) item budget by investing skill points, implementing a tradeoff between skills and equipment and thus adding more depth into the game, but it got ignored by the devs, although chadz once mentioned that my system would actually be "interesting"... this is not to lobby for my system, since cRPG is already dead in my eyes, it's just to show how the upkeep system is also part of the problem and how a possible alternative could look like.
