Tuesday, May 27, 2008

Warning: Does Not Imply Causation

Currently, the AL standings have a correlation coefficient (r-value) of .218 with my preseason projections. The NL is at .088. Those numbers don't say much about my prognosticating abilities, though I'm doing pretty well on these.

However, once we strip out the luck, things change. When comparing the preseason forecasts to the current third-order W-L records, we get .431 for the AL and .396 for the NL. Not perfect, but much better.

The difference? Third-order wins tell us which "breakout" teams are for real (Rays) or not (Astros, Marlins). You can forecast skills, but not luck.

No comments: