To illustrate, imagine two people in the IARPA tournament, Frank and Nancy. In year 1, Frank does horribly but Nancy is outstanding. On the bell curve below, Frank is ranked in the bottom 1% and Nancy in the top 99%. If their results were caused entirely by luck—like coin flipping—then in year 2 we would expect both Frank and Nancy to regress all the way back to 50%. If their results were equal parts luck and skill, we would expect halfway regression: Frank should rise to around 25% (between 1% and 50%) and Nancy fall to around 75% (between 50% and 99%). If their results were entirely decided
...more
