The Dangers of Rookies

日期:2013/03/07

One of the dangers of having a rookie on your team is that the rookies like to find regularities based on looking at every interval， every magnitude， every market， every combination thereof of x variables， and every time period. It's truly a search of implicitly hundreds of thousands of possibilities to come up with a regularity， fifteen on say 40 observations that has about a 5% chance of consistency with randomness assuming it was the only 1 selected. The problem is that they seem so good in isolation before you realize it was the fruit of a tremendous number of look backs， complexities， and multiple comparisons. I strive to tell them "Simplicity." Read Zellner. Another good thing to do is see all the biases from using cart or regressions trees of automatic interaction detector， and all the safeguards built into those methods， —- and of course they overfit， and multiple classify and are only recommended as preliminary by the authors. But …. but…. how destructive it is to receive one of these regularities during the middle of the day… especially when you have a position on the opposite side from the rookies. Proffer. What lessons can we derive from coaches that treat the rookies with grave skepticism like Woodson who calls Shumpert "Rook" and all the players that haze the rooks endlessly to prevent them from interfering with the natural order of things.