Twenty teams have officially finished their seasons and are already considering available free agents, which means they also have to consider whom they can afford. This involves predicting the salaries they will have to pay to arbitration-eligible players. Rather than having their salaries determined by the highest bid, their salaries are set to be determined by an arbitration panel. Of course, very few players actually ever make it in front of that panel, since teams and agents spend considerable resources trying to resolve their salaries in advance.
Last year, MLBTR owner Tim Dierkes asked me if I thought I could put together a model that predicted arbitration salaries. I had studied free agent salaries, but I decided that I could probably do almost as well with arbitration salaries. It went better than expected: the model was within 10% of the actual salary for 55% of players who signed one-year deals, and was within $1MM for all but 4 of the 156 arbitration-eligible players.
Unlike free agents, whose salaries are determined by the highest bid among 30 teams with 30 different ways of predicting and valuing future performance, arbitration eligible players receive salaries based on the similarity between their past performance and the performances of other comparable players. A well-designed model can do a good job of sifting out which statistics are most important and predict salaries accordingly.
Last year’s model was strong, but there were still a number of players who were poorly projected. One category with which I struggled was breakout stars entering arbitration for the first time. Jordan Zimmermann received a salary of $2.3 million, above my projection of $1.8 million. Even though he only had 23 career starts in an injury-checkered past before going into 2011, his solid 3.18 ERA in 161 1/3 innings in 2011 seemed to matter more than his previous injuries. Another thing I learned in my projections for 2012 was that previous salary did not matter much for first-time eligible players. My biggest overestimates included projecting David Price at $7.8MM instead of his actual $4.35 million salary and Rick Porcello for $4.2MM instead of his actual $3.1MM, since I thought hefty Major League deals given to draft picks would give these players a leg up going into arbitration. This is not true, as I have since learned. I also missed big on some players who had strong rebound performances after being non-tendered the season before. The biggest miss was projecting Melky Cabrera’s 2012 salary; I only predicted a $4.4MM salary, instead of his $6MM earnings. It turns out that bouncing back after being non-tendered gives players like Cabrera a little extra room for raises, and such players are now projected for higher salaries in 2013.
I did a lot of work on improving pitcher projections for this year’s model. I originally included all pitchers into the same model, which gave them credit for wins, saves, and holds as they received them in each role. This was supposed to better incorporate swingmen and other pitchers with evolving roles, but now I have separate models for starters and relievers, which allows for more accuracy for everyone. In last year’s model, I ignored the importance of strikeouts for starters and had to introduce other measures to juice the salaries of elite starters. This year’s model incorporates elite starters much more smoothly. The starter/reliever distinction also gave me an opportunity to notice an important feature about arbitration -- declining marginal returns to individual statistics. It turns out that the gap in earnings is much larger between pitchers with 170 innings and 200 innings than between pitchers with 200 innings and 230 innings, and that a guy with 30 saves out-earns a guy with 20 saves far more than a guy with 40 saves out-earns a guy with 30 saves.
In the coming weeks, we will present the projections for all 30 Major League teams’ arbitration-eligible players. Last year’s projections had a good foundation, but we believe this year’s will be even better. These can help teams and fans alike as they try to anticipate trades, extensions and non-tenders and determine how much money is available for free agents.