The Regression Fallacy and Baseball Salaries
The performance of baseball players is typically measured by accumulated totals of productive batting events. Assuming independence (roughly) among plate appearances implies that recognition is based on estimated success probabilities. Those of us working on the Baseball project know that the variance in those estimates are surprisingly large and typically the same order as the difference between players we consider "good" and those that are ordinary. Thus, the team that buys the star of the hour is almost assuredly overpaying. Is George Steinbrenner listening?