Because its easier to assess a role than it is to assess someone's actions. An instance of "bad role-play" might take 5 minutes in real-time. Even one of these, if its egregious enough, should be enough to damn a character. But what are the odds its caught?

Similar problem for differentiating between "average guy" and "awesome guy". Awesome guy's awesomeness isn't going to be obviously apparent 100% of the time. Every once in a while some situation will come up, or he'll have some interaction with another character, that really demonstrates how awesome he is. But if nobody happens to be snooping him at the time? Then its like it never happened.

There are also characters who have bouts of awesomeness and bouts of craptaculosity on the exact same character. If the staff happens to notice the good stuff and not the bad, and rewards that guy for "observed rp" then the playerbase raises hell.

At the end of the day I kind of agree with you: role-contest rewards should be scaled back in significance, since its possible to write a tremendous role but then not really play it out. And its the playing it out that matters.

Rewards should be given for repeatedly observed good role-play. One good thing maybe gets you a little imm xp. Lots of good things piled up over time, with few-to-none bad things, gets you a bigger reward.

The corollary to this would be that for a character to have "observed good role-play" he must have a somewhat detailed role entered that he's actually playing. Hard to have "role-play" without a "role".