Consensus forecasts, an averaging of forecasts from different sources, have done quite well in studies comparing forecast values with actual ones, so it makes sense to have a consensus of the projection sources RotoValue has for baseball. I’ve done an initial averaging of four sources:
- CAIRO – from S B of the Replacement Level Yankees Weblog.
- Marcel – the basic projection model devised by Tom Tango, coauthor of The Book. This year I’m using Marcel numbers generated by Jeff Zimmerman, using Jeff Sackmann’s Python code.
- Steamer/Razzball – developed by Jared Cross, Dash Davidson, and Peter Rosenbloom, and using playing time projections from Rudy Gamble of Razzball.com.
- RotoValue – my new and (I hope) improved projection algorithm.
This is a simple average of these sources – if a source does not project some stat (some sources don’t project batters faced for pitchers, for example), then the average does not include that source for that stat; but it will include a 0 in the stat so long as the source projects some players with positive values for it.