2010 FantasyScope projections in the works
Alright, our first stab at projecting players was admittedly flawed. With the 2009 regular season tucked away, we are already working hard at improving our system to bring you the most accurate projections possible. Here’s a sneak peak at some of the things we are working into thus year’s projections:
Weighted three-year averages for all players
One of the problems we suffered from last year was not being able to correctly predict a player’s talent progression or regression based on three-year total stats. Thus Chad Billingsley (among others) was not really given his due. This year we have moved to a weighted model which helps give the most emphasis to what a player has done lately, rather than 3 years ago. We use Cartwright weighting in our model, where each year counts double relative to the year before (4x, 2x, 1x). We are also tweaking our aging and playing time mechanisms to give a more realistic boost or decline in performance as those factors are adjusted.
Fully regressed projections by league and position
Last year we applied regression to only players of extremely small sample size, but it affected some things adversely, because the amount of weighting varied and some players were not regressed at all. Hitters will now be regressed by 220 PA according to their league and position (including bench players), and pitchers by 220 batters faced based on league and role as a starter or reliever. This should result in far more realistic expectations for players with small sample size as well as part-time players.
Improved correlative minor league projection models
Last year we used a linear average to project Major League performance in key rate stats. It did help us identify high-potential rookies like Brett Anderson, but it overestimated the performance of almost all of the projected players. This year, we have developed a correlation model which regresses each AAA and AA player against players who have previously made that same jump to the Majors, and it gives a far more conservative projection for each of these players. This Major League Equivalency will be averaged with any actual Major League data available to form the projection for each projected first-year player.
We are still in development and welcome any comments and suggestions you might have for us! Feel free to email at redsoxtalk at gmail dot com.
