If you give lower-skilled players more XP for killing better players, then what incentive does the better-skilled player have to play the game? BTW: etpub team member josh has done extensive, PhD-level work on this (he does have a PhD in statistics), so you might want to check out his work. He uses player rankings not to give out XP, but to calculate who is the “better” player given a bunch of players that have not faced each other before. So he built a great global player ranking system.
Here is a very simple explanation of thet different player ranking methods/stats:
http://stats.etpub.org/explain.php
And here si the nerdy version of the reasoning and methods behind the statistics (all 26 pages of it!):
http://axon.cs.byu.edu/~josh/etstats/update_bayes.pdf
If you would like to give out additional XP for killing the better player, you would start off where josh left off: The hard part is determining which player is better and then, how much better? Once you know that answer (which josh provides), then knowing how much more (or less) XP to give to the killer is easy.
EDIT: Determining who is the better player actually open up a huge can of worms… What about the panzerwhore or flamerwhore who keeps spawnkilling AFK players? Shouldn’t a kill with an MP40 count more than a kill with an airstrike? What about a player’s effect on a team? If the player is playing and his team wins more often isn’t that a sign of being a better player? If a player is a huge fragger (think rambo-medic) but his team does not win more game because of his presence, that player is not really a “better” player (team-win wise anyway).
EDIT2: Throw in the fact that maps are skewed so that a certain team will win more often (for example, given a large enough time limit [and it does not have to be very large], allies will always win SupplyDepot2). You have to build a ranking system that takes into account this team advantage per map (and at least per server).
All of this stuff has been looked at by josh in his player rating implementation, so it might be very interesting to look at how etpub has its ranking system implemented. Regardless of what you think of pub servers (or the etpub mod for that matter), it might be interesting for you to see the etpub global stats page ( http://stats.etpub.org/ ), it has almost 122,000 ranked players, on 171 servers, with 297,000 matches ranked.
FWIW: Only matches with 8 or more players are taken into account, and any matches with 1 or more bots are ignored.