Imagining games to have a "skill component" and a "luck component" is the wrong way to conceptualize games which involve randomness. A much better way conceptualize these games is to consider how many "units" (games, matches, hands, tournaments, whatever) must be played before the distribution of players by (score/place/points) becomes indistinguishable from that expected based on the players' ability levels.<p>For instance, if a GM rated 2800 plays a GM rated 2700, he may lose. In fact, he may lose several games in a row. However, if the two play a 30-game match, the probability that the 2700-rated player will win is very low.<p>Now, if you take the best heads up no limit hold em player in the world and have him play a series of hands against the hundredth-best heads up no limit hold em player in the world, he may very well lose the first hand. He may very well be down after the first thousand hands. But if they play, say, 1,000,000 hands, the probability that the weaker player will be up on the stronger player is as low as, if not lower than, the probability that the 2700 will beat the 2800 in their match.<p>It's not about "skill" versus "luck"--the question is simply, "How many (hands/games/matches/etc.) must be played before the probability that the weaker player (has won/is ahead/etc.) becomes sufficiently small?