The graph of random data looks like that since its capped. The first place person cannot overestimate his position, and the bottom place person cannot underestimate his position. So any randomness at all will replicate the effect unless there is much more overconfidence in the high achievers than low achievers.