If you don't mind, I'll use my typical response style, as I find addressing each point individually (when possible) to be more convenient and result in clearer responses.
MMXI said:
You only need a limited amount of actual mechanics to do a decent enough simulation to discourage the player from performing out of character actions.
The problem here is that good deterrent isn't automatically good simulation. In some cases, even though penalty is harsh enough, it may make no sense in-universe, which makes it poor simulation. Even if the penalty would make sense, the scenario leading to it may not if the character involved would simply not engage in such behaviour.
This is the main distinction between recruitable characters and ones created by the player (or otherwise forming a designated "core party"). Proper player characters are given minimal/broadest/least specific possible context in the gameworld to accomodate any character or variety of characters player may wish to create. The main point here is that for core party, the criterion of "would not do" is void, since the program has no way to tell what a player character would or wouldn't do.
The concept of core party or protagonist, while not strictly formal possesses enough markers to make it unambiguous without accounting for any storyfag baggage associated with traditional meaning of "protagonist", in particular it is easy to determine what characters are not core party - they are characters given some proper context in the gameworld.
A character you can find doing something, somewhere in the gameworld, who is motivated by certain goals (may have quest associated with him to reflect them better) and has certain history is *definitely* not a part of the core party since all these form a very specific context. In particular, the history and motivation of such character impose some "would" and "would not" restrictions in-universe and good simulation of this universe *has* to account for that. Filtering orders through AI which can veto them based on character personality and modified by characters willpower, PC charisma and relationship variables should give a good compromise between tactical control and well-simulated characters.
Argument from "It's my party WAAH!" is not simulationist - it's 100% gamist.
You have to understand that in the eyes of a devoted simulationist-narrativist (more of the former according to narration-through-simulation ideal) it makes you an inferior person.
This doesn't mean that every character has to be a collection of numbers and nothing more. You can still have developer created dialogue trees for characters. You don't have to generate those from their statistics or anything. This means that you can add enough background information and other things that would be very tricky to add as statistics. As long as there are mechanics to discourage paladins from slaughtering innocents and thieves from doing full frontal assault (to quote JarlFrank's example) then it's all good.
But that's the crux of the problem here. Those differences in behaviour are not determined by class nor stats, but by background.
The distinction between storyfag individually written and scripted characters, and non-storyfag generated ones, with background text and dialogue based on some templates and scripts parametrized or built from blocks appropriate to a given character is irrelevant here - what matters is relevant portions of the character background being in a form understandable by the game's logic.
Just think about how many statistics there are for combat in something like D&D. If you mirrored that for their "personality" or whatever you want to call it, I'm sure that's plenty. Think about all of those weapon proficiencies, saving throws, resistances and specific damage type armour classes. Then think about how many actual "personality" measures you can come up with. It'll easily be enough, combined with developer created dialogue, graphics and even voice overs to make a character both play uniquely (while still inside the game rules) and "act" uniquely in a more hand-crafted sense.
Except behaviour is far more complex and far subtler than giving some pluses when using certain type of weapon. Additionally, we're speaking of limiting or forcing behaviour - the area of behaviour that is normally regulated by players' role-playing in case of the core party.
It makes sense to split it from the stat part and merge it into the AI - recruitable characters can be generated with the same amount of detail and same range of personalities as personality generator allows, or manually scripted with detail required by their written background, so granularity of our "personality engine" will be non-issue in such cases. With the core party, however, you're bound to run into situations your "personality engine" won't be sophisticated enough to handle, which will be incredibly jarring and detrimental to the gameplay.
I'm not against partial inclusion of behavioral restrictions to character system - in particular you can have statistics like willpower that may influence the moment the character control is relinquished by controlling party (player in case of core characters, AI otherwise), and you can have optional traits representing well delimited characteristics (and, in case of traits enforcing behavioural restrictions, coming at negative cost during chargen), but the key word here is "optional" while the personalities in general aren't.
But what exactly is the challenge of an RPG? It seems obvious to me that the challenge is to exploit the strengths and weaknesses of your various characters in such a way as to allow you to finish the game.
Playing characters (ones you've created or ones you've recruited) optimally throughout the game is ultimately your goal.
Except that explicit notion of "goal" is inherently non- simulationist. It's purely gamist, though you can argue that in a highly linear storyfag game the "true" ending is the goal.
From simulationist POV, goal doesn't have to be in any way accounted for or enforced by the game. There are even games with no specified goal at all, which is pretty much as pure as it gets in terms of simulationism.
In such purely simulationist game goals would be up to player, and optimality of given tactics or strategy can only be meaningfully assessed in conjunction with goal.
If you fight with your fighters standing at the back and your mages meleeing in the frontline then you're going to fail repeatedly. This is what RPGs are all about really. The player should be allowed to do stupid things like that, but should be discouraged by the game mechanics. Mages should be able to rush past the fighters to smack a group of ogres around with their staves but having a crap THAC0, a bollocks AC, shit HP and doing low damage are those statistics/mechanics that discourage you from doing it. So you don't do it. Ever, really, unless you've cast some spell that turns you into some fighter equivalent temporarily. So why shouldn't the game give you the option of doing those things with recruited party members? The mechanics will discourage you from doing them, so what's the problem? The game will let you play a character out of character, but you'll be punished for it in the same way that you get punished for doing the same with the protagonist.
Since most goals at least implicitly imply survival, then retarded combat tactics will be pretty much universally non-optimal, but that makes it rather specific case rather than general rule.
But, let's agree that RPG revolves around strengths and weakneses of different characters. Wouldn't choice between recruiting mediocre, but disciplined fighter or a psycho that can hit people so hard he makes their mothers bleed, but also has propensity towards braining people for looking at him funny which inevitably ruins most of the party's attempts at diplomacy and its legal reputation add more depth to it?
Wouldn't you rather have a dilemma between recruiting a one-of-a-kind marvel of stealth and lockpicking and... wait where is that amulet we spent last month dungeoncrawling for? Or our "locksmith" for that matter?
What about joining with a paladin that makes a formidable combatant, valuable asset when it comes to dealing with <insert paladins religious organization here>, but doesn't tolerate legally questionable methods of fundraising, certain forms of magic and may skewer certain disagreable types even if they might be more valuable to the party alive?
Wouldn't BG be a much better game, if, say, Xan was a much better mage, but had actual detrimental effect on party morale and had really lowe morale of his own?
That can be solved quite easily. Either give competence measures to the AI controller itself, or add in some counterbalance statistics to divert the AI controller away from "perceived" optimal actions.
I'm not talking about some generic incompetence, but rather something along the lines of "blows all loot on hookers and booze" or "likes punching town guards in the faces" or "is a fanatic who skewers people on his sword if they deviate from his perceived righteous path in the least".
Also, discrediting JF's example on grounds of storyfaggotry is about as sensible as dismissing AK-47 when discussing firearm reliability because it was made by commies whose goeals and methods you don't agree with - in other words:
![Wast re-ta-dred maen? :retarded: :retarded:](/forums/smiles/retarded.png)