What you have just described is similar to what happens when data is randomly generated. With the actual data, you would expect some players to perform better than average and other players to perform worse than average, just because that's the way things happen. In other words, the actual data is no different than the randomly generated data. There is no statistical significance behind those players whom you consider clutch. Furthermore, they can't repeat their 50% above the norm from one year to the next.