Over at his blog, Richard Bartle has layed out a very primative sketch of how an effective player driven reputation system might work. The short version is “Amazon’s You-might-also-like Lists”. You would rank people when you get to know them, and based on your selections and the selections of other people, someone you have never met before might be “recommended” to you because people who like the same people you like also like this new person.
Simple example ripped from Richard’s post:
You like A, they like A; you dislike B, they dislike B; you haven’t met C, they like C, so C is probably a decent person. The greater the insersection between lists, the higher the chance that you’ll share their opinions.
The only drawback he found was in the server resources it might take to maintain and display this data. Well, as far as maintaining it, there isn’t much I can suggest, it is going to be a huge amount of data – potentially if you have X players and all X players rank all other players (X-1), you have to store X*(X-1) records. Taking a game like EVE Online that runs a single server with over 100,000 subscribers, that is potentially 10 billion records. Calculating the data, however, could be contained by imbedding the “score” of a player to a “Looking For Group” tool, or as a given command (inspect player), so as to reduce the amount of processing done with these numbers, as opposed to his suggestion of having the results display all the time by a player’s name.
It is definately a good idea, I think, and merits a deeper look at the possibilities and realities of implementation.