We’ve decided to evaluate the 2007 MLB player projections of 4 sources that would be available for CHALQ’s use.  The systems are:

1. Marcel
2. Bill James (pay)
3. ZIPS
4. Rotoworld (pay)

We’re trying to find out who projected closest to the actual 2007 results so we can choose a system for CHALQ.  To do this we are running an analysis similar to one Baseball Prospectus conducted in 2004. 

The metrics for evaluation we’ll be using are the Mean Error and Correlation Coefficient.  The minimums for inclusion are: Pitchers - 50 IP and Hitters - 250 PA (Plate Appearances).

The issue with incorporating minimums is that we’ll miss out on rookies and first year players that were called up at the end of the season.  These players projections are valuable to fantasy users.  For this reason we may include a combo of ZIPS and one other system in a weighted average projection.

On that note, after we’re able to rank all four projections systems we’ll attempt to beat all of them with a hybrid system by weighting each system together into one.  It would be nice to find an official CHALQ system using this method, but we’ll ultimately be offering whatever projections (including the hybrid) yields the best results.

I’ve got most of the results and am putting it into a tidy report.  I’ll post this in the next week or so.