So every now and again you get an indie-type game whose character creation process is basically just writing down three traits completely freeform and then maybe you have a number. These have not taken off to dominate the market and mindshare of roleplaying games. Seeing as how a good character creation system is often a selling point for video games, it shouldn’t come as a surprise to anyone that abdicating the chargen system to the players didn’t really go too well. People like shopping for bits and bobs to customize their character with using a limited pool of resources.
That “shopping spree” sort of rush isn’t the only reason character creation is good for RPGs, though. It also helps to avoid Mary Sue and her overcompensating counterpart Anti Sue. The heart of the Mary Sue issue is that one character is loaded down with so many more special traits than other characters in the principle cast that it becomes jarring and usually comes off as the author showering their favorite character with power and attention like an old couple spoiling their grandkids. By having each special trait tied to some kind of cost based on how special it makes the character and how often that specialness comes up, you strongly curb the ability to make Mary Sue characters.
Likewise, by requiring everyone to spend X amount of character points or be minimum level Y, you can ensure that no one shows up with a character who’s bland and inept to the point where it’s hard to figure out why they’re even in this story. This second problem is more rare in standard fiction (although it comes up surprisingly often in roleplay contexts, whether freeform or system-based, from people who are overcorrecting for Mary Sue), but having a protagonist who is easily the most boring person in the room can be just as bad as your bog standard Mary Sue.
A lot of RPG chargen (and character advancement) systems are only loosely designed to operate like this, however. For example, in D&D races give certain bonuses and penalties, but there has never been a satisfying mechanism for playing as a particularly powerful/awesome race by sacrificing character power/awesomeness in other areas. Even just being an elf is slightly more awesome than being a human, which is usually modeled by giving humans bonus feats or some other very specific kind of character resource, but the problem with this is that #1 a player might reasonably want to spread their awesome around to things that aren’t feats and #2 a player might want to sacrifice some awesome in other areas in order to have a race that is way weirder than an elf, like a troll or a pixie or something.
As an example of the first problem, someone might want to play a wizard with mastery of two different schools. That’s not such a huge reach, and if you’re designing characters freeform, it’s easy to find other places to sacrifice some awesome in order to avoid making your character seem ridiculous for violating one of the established norms of the setting, which is that wizards only focus on one school of magic at a time (i.e. they’re diviners, or necromancers, or elementalists, or whatever, but not summoner/beguiler combos like your guy is). Main characters are a big deal and are allowed to violate some norms, so long as it doesn’t get out of control. Take someone who’s already doubling up on wizard powers and make them even a relatively bland elf and you’re already risking making them (or elves in general, depending on presentation) seem like a Mary Sue. Make them something really bizarre like a troll and you’re well into Sue territory (assuming that learning two different schools of magic is a sufficiently big deal in the setting – this stuff is relative). The closest D&D and its cousins get to modeling this is by having dual-class options that let you be kind of sucky at two things instead of regularly awesome at one.
Now, partly this is because D&D concentrates so much of what makes a character special into classes that there’s not really enough awesomeness left amongst everything else to be sacrificed for a full second class. Character classes come with a steady drip of new powers and abilities and often also imply quite a bit of fluff and background, and while there is obviously plenty of wiggle room for players to invent details, being a paladin or a sorcerer implies a huge amount of things about your character in a way that being a dwarf with the Blind Fighting feat does not. Even ignoring game balance, from a purely narrative point of view D&D concentrates a lot of what you’d use to describe a character in the class, which means having two threatens to dive into Mary Sue territory pretty much automatically. One thing to note is that in narrative terms, having two related classes, like being both a Ranger and a Druid, would be way less of an issue than having two unrelated classes, like being a Paladin and a Wizard, even though those would both be about equally unbalanced in terms of pure mechanics, in that both characters would have access to a full casting class plus a bunch of weapon and armor proficiencies plus some fighting styles or smite effects and such to make those proficiencies more useful. While the balance issues are important, they’re important for completely unrelated reasons that I’m not going to get into here. What matters for purposes of this post is that a Ranger + Druid combo would be narratively pretty acceptable whereas a Paladin + Wizard combo starts to veer into Sue territory almost immediately (however if we see someone start as a Paladin and then pick up Wizard powers over the course of the story, that’s fine).
Where D&D really collapses is on the second problem. If you want to play a D&D troll, that is a fairly weird and unusual race to play, and your character should have to sacrifice some amount of specialness in other areas for it. D&D doesn’t give you any way to actually do this (though it has, at times, tried). In terms of narrative design, it is easy to envision a troll sword master being no more unusual or spotlight hogging than those Ranger/Druid or Summoner/Beguiler combos we were talking about earlier – maybe even less, actually. Being a troll is a much bigger deal than being an orc, but it’s not really as big of a deal by itself as the suite of abilities that D&D gives you for being a Ranger. These aren’t just +1s that a narrative wouldn’t even notice because it doesn’t roll dice, this is things like being an expert marksman who can bring down titanic opposition with a bow and having a small library of nature-themed magical superpowers to boot, and it’s not like a bow-focused Ranger is particularly helpless in melee, either. A troll, meanwhile, is big and beefy and can regenerate damage that wasn’t dealt by acid or fire. That’s pretty awesome, especially that last part, but it’s not as good as the Ranger’s set, even from a purely narrative perspective. The Ranger’s abilities are the kind of thing you give to a protagonist that allows them to solve problems relevant to fantasy adventures for the length of a novel. The troll’s abilities are the kind of thing you give to a supporting character or a minor villain, and can only really support a few scenes on their own. Being a troll is not as cool as being a Ranger, and being a troll Fighter is not as cool as being a Ranger/Druid gestalt. It therefore follows that if being a half-orc Fighter is okay, it is narratively fine to sacrifice some of the Fighter’s abilities in order to upgrade from half-orc to troll. It should even work out in terms of game balance. There is no option to actually do this in D&D, however.
This is limiting in all kinds of ways. Every race has to be balanced against every other race, every class against every other class, every feat against every feat and every background against every background. This is true even when it narratively makes no sense, and indeed, some backgrounds sound way more impressive than they turn out to actually be in gameplay, because backgrounds are meant to be very minor additions to a character, even if your background is that you led a peasant rebellion against a tyrannical baron. That’s way cooler than going to college, but D&D prices them both the same: They are both your background pick, and they both give you a single skill, a bit of money, some gear you probably don’t care about, and a minor feature that will only occasionally be useful. Those are all perfectly fine things to have on a character sheet, and it’s definitely all the attention that the “went to wizard college” background deserves, but the peasant rebellion background should have a bit more oomph to it, and in order to avoid one character becoming clearly more important than the others (you almost never want your ensemble to have an obvious main character in D&D, because all the players expect to be main characters together and not to have someone else declared the story’s focus), you need to be able to give a little less oomph to something else.
Being a boring old human instead of an elf or a dwarf or something else kinda cool shouldn’t give you a feat that you may or may not want or need, it should give you a small pool of fungible character points that can spent on anything. Locking class abilities into distinct classes with steep entry costs is still fine in that it keeps things simple, so that rather than attaching a point cost to every class feature and letting people buy whatever they want, people buy the entire Fighter class, and can then sell off individual features from that class to pay for the troll race or a more awesome background or whatever – but importantly, there’s a large upfront point cost to being a Fighter at all, so if you buy the Fighter class and sell off every single feature, you still have fewer points. This encourages people to pick a single class and mostly stick to it, so that you can still say “I’m a Fighter” and have that actually mean something. Being able to communicate your character’s basic schtick quickly is a huge benefit of class-and-level systems, and it’s something that games like D&D might reasonably not want to give up, but I don’t see that as any excuse to make trolls unplayable.