One of my primary goals when designing game mechanics is to make the game encourage actions which are genre-appropriate, fun, or fit well with the theme of the game. Encouragement can mean different things, but it most often means making that action one of the best possible actions from the competitive point of view (winning the game or defeating the adventure).
Let's create an example of what I'm trying to prevent. Imagine you have a fencer in a swashbuckling adventure game. The character concept, and the genre, imply that he should frequently use his weapon to disarm his opponents, rather than simply stabbing them.
In most traditional roleplaying games, the disarm maneuver involves the attacker making some sort of (often fairly difficult) roll to successfully force the defender to drop his weapon. So the game has fulfilled the mechanical requirement of permitting the character to disarm the foe.
The problem is that the disarmed opponent can then spend his action to pick the weapon back up. So all the fencer has done is spend an entire action to force the opponent to spend an action. And that is if he succeeds in the disarm; otherwise, he has spent an action for no effect. So overall, disarm costs an action for the attacker in order to cost the defender less than an action. Not very useful.
Now, one might point out that there are situations where this disarm maneuver is indeed useful. In particular, if there are two PC's fighting one opponent, one PC can disarm the opponent, and the other can pick up the weapon, leaving the defender unarmed and helpless. Or if you are fighting near a cliff or in a bog, you may be able to disarm an opponent and have the weapon be lost for good. Also, you can use disarm just to waste time, in case the situation doesn't require you to win, or if you are one-on-one with a more powerful opponent and want to delay matters until your allies arrive.
Unfortunately, none of this really helps the situation at hand, which is capturing the spirit of a swashbuckler movie where the fencer frequently disarms opponents in situations where none of these are true. Indeed, the fencer is often outnumbered and pressed for time, yet still disarms the opponents. But when playing a game with standard disarm rules, one would very quickly realize that this is totally ineffective and basically a waste of time, and the player would want to start stabbing opponents to win the battle.
One response to this would be to say that the swashbuckler has a psychological limitation that makes him enjoy disarming opponents even though it isn't really effective, just because he enjoys doing so. This limitation could give him the points to be such a good fencer that he can get away with some nonsense and still win the fights. But I don't really feel like this is true to the source material or the character. If this were true, you'd expect the other characters in the movie to say, "Wow, that guy's a good fencer. Too bad he's a grandstanding moron!" But they don't do so, and the movie doesn't convey that impression. Indeed, the character concept for the swashbuckler may well be that of a reluctant hero forced to fight for an important cause, not a that of an egotist enraptured with his own fencing skill. Also, if the psychological limitation theory were true, you'd expect the swashbuckler to drop the disarming when he has to save the life of his true love, but that doesn't really seem to happen in the source material.
A slightly different response would be to say that because disarming is appropriate to the source material, it is the player's responsibility to throw in some disarm attempts as "good roleplaying". The idea of expecting players to help make the game fun is a good one. I don't generally play roleplaying games in a highly competitive style, and in the roleplaying games I play in, it is understood that the players don't just go all-out to optimize every game mechanic to win the game, but try to do fun things and advance the story, or at least make the battles entertaining. There is no way to balance everything perfectly or guarantee that the most entertaining move is the most effective. You expect the players to mix things up and put some variety in the game even if a careful analysis may reveal that a more boring strategy is somewhat more effective. But I generally think this sort of thing works best when the game balance between the different actions being considered is pretty close. A lot of the time, you may suspect a certain action isn't optimal, but there are pros and cons both ways, so it isn't really clear. So you really feel free to do whatever you think is cool.
For instance, in Torg you had the ability to take "approved actions" each round, like taunting or tricking the opponent, instead of just attacking them. Success would give you a minor advantage over the foe and an extra card. We loved approved actions, they were cool, and we did them over and over. We always strongly suspected that just attacking the foe would be more effective; ending the battle early is generally a good thing. But the benefit of cards was hard to quantify; they could set you up for a big attack later on, and could potentially be saved for use later in the adventure. And approved action were fun, and Torg characters were so full of possibilities (hero points) that you weren't that scared of combat anyway. So it all seemed to work OK. Actually, though, I should admit that in this case I did make some rules changes to make the approved actions a bit better and encourage them more.
However, when one action is just flat out way inferior to another, even in cases where it ought to be useful, that is just going way beyond the scope of what should be expected from the players. In the case of the disarm example, using disarm in most situations isn't an interesting choice with pros and cons; it is just a way to make your character less effective. The more you use it, the less effective you are. It is basically as if the GM is saying, "I'd like you to vary your actions to make the combat more interesting. Whenever you do so, I will punish you by making you less effective. The more interesting you are, the more I will punish you." It is perverse and unfortunate, and even though players can and often do make games more entertaining this way, it would be much better if the rules were changed so that they were not made ineffective by doing so.
On a related note, the players can bypass various forms of rules abuse by following conventions. For instance, in Champions, when an enemy strongman jumped up to you and started trying to smash your face in, it was most effective within the rules to ignore that person and attack a weaker enemy. Nothing in the rules required you to pay any attention to enemies attacking you; even if you were virtually surrounded on all sides with angry swordsmen, you were free to run off to a different part of the battlefield as if they were all paralyzed and attack the vulnerable boss behind them. Since this was totally inappropriate to the genre (and reality too), we made a convention that you were expected to defend yourself when attacked and had to do something if you wanted to fight your way past the attacker. This was a good convention; conventions can be useful to fix bad rules or substitute for rules that don't exist. But it is even better to fix or create the rule, rather than having a convention. Then you know exactly when the rule applies and what the penalty is for breaking it. In any case, the idea of making a convention doesn't work well for disarm example because it isn't clear how you would apply the convention. Saying that you had to disarm every other attack would just be way too unnatural.
Another way to deal with the issue here is for the GM to compensate by rewarding genre-appropriate maneuvers. I highly approve of having the GM reward genre-appropriate maneuvers with cool custom bonuses. But this is best for things done infrequently. If the genre rarely involved disarming opponents, and you suddenly had a good reason to do so as a surprise maneuver to liven things up, it would be very appropriate for the GM to make up some sort of bonus on the spot. But if you disarm constantly as part of the genre, custom bonuses aren't very practical. If you give the same bonus every time, you've created a rule, and the rule might as well be written down. If not, you start to play a game where the GM is just making up the rules arbitrarily. You can do this, of course - you can roleplay without any rules at all if you want to. But the assumption here is that we are playing a game with rules, and the premise behind playing a game with rules is that, most of the time, it is better to have a rule than rely on pure GM arbitrariness.
So for all of these reasons, I would want to devise an improvement to the disarm rule. One could argue against this by pointing out that boosting the disarm power as written shouldn't be done because it would be too strong and would break the game in the situations where disarm is already a useful ability. This is true, but simply means that attempting to fix the problem requires rethinking the rule rather than simply boosting it. Part of the skill in modifying game rules is making sure that you don't create more problems than you fix. Just because my game analysis indicates that a game has some sort of problem or imperfection doesn't mean I will make a rule to fix it. I only make a rules change if I think the new rule will be better overall than the previous rule.
The truth is that every game is going to have problems simply because of the choices made in meeting various conflicting design goals. And sometimes those elements that make the game fun also seem to have disadvantages too. For instance, one of my favorite board games is History of the World. But it has the disadvantage that more than 6 hours to play. It can be hard to get people to play for this reason. But the length of the game is related to the fact that it plays out the "History of the World", and that is part of what I like about it. I haven't really thought of any clever way to speed up with game without detracting from the epic quality I like about it. So I haven't tried to make a rule to speed up the game, I just consider the length part of the pros and cons of a game I really like overall. I feel that putting in a simple-minded rule to speed up the game - like playing for only 3 turns - would make the game worse rather than better.
But if I did think of a way to make the game just as fun but twice as fast, I wouldn't hestitate to try it out. Just because a rule is hard to improve upon doesn't mean you shouldn't try. Sometimes I try a lot of experiments, and a lot of those experiments fail. Sometimes I try experimental rules that I know won't work perfectly, just to get information. But in the end, the goal is to craft a new rule that is better overall than the previous rule.
By the way, I've never really come up with what I would call a perfect solution to the disarming problem, only various ideas. One example of an idea would be to use the 4th edition D&D power design philosophy and have disarming be a special attack that causes damage, with the special effect that the damage is totally abstract and the attack looks totally non-violent. Attacks which fail to kill have the bonus effect of temporarily disarming the enemy; attacks which kill the enemy either look like kills on-screen, or the enemy looks like he is still conscious but is counted as "defeated" and no longer has any game effect on the battle.