I recently posted this survey question:
Sometimes, information from Baseball Prospectus’ PECOTA player projection system is used on MLBTR. Should we continue to post this information?
Over 2,000 MLBTR readers responded; about 90% said yes, continue using it. The basic vibe of the comments was that more info can never hurt, even if we all acknowledge that a projection system isn’t gospel.
Many readers requested info explaining what PECOTA is and how it works. Wikipedia does a very nice job giving the basics. This is an oversimplification, but basically PECOTA projects a player’s future performance by looking at comparable players throughout baseball history.
How did PECOTA do trying to predict the 2007 season? It led all projection systems with a .451 correlation coefficient for pitchers. (A perfect set of predictions would have a coefficient of 1.0). PECOTA also had the highest correlation coefficient in predicting hitters for ’07, at .627. ZiPS, a free system from Baseball Think Factory, was close behind. As you can see, pitcher performance is very difficult to project.
