Printer-friendly

February 21, 2008

Schrodinger's Bat

Flashing Leather Down on the Farm

by Dan Fox

"It is the peculiar and perpetual error of the human understanding to be more moved and excited by affirmatives than by negatives."
--Francis Bacon (1561-1626)

Earlier this week, both the sabermetric community and the topic of defense received a little mainstream press when Shane Jensen of the University of Pennsylvania discussed his Spatial Aggregate Fielding Evaluation system (SAFE) at a meeting of the American Association for the Advancement of Science in Boston. No doubt Yankees fans and some members of the mainstream media were once again offended, as The Captain's weaknesses were displayed. But seriously, when every fielding system in the free world comes to the same conclusion, one would think it would at least question the belief that Derek Jeter is a superior defender despite the cachet that comes with winning multiple Gold Gloves. Have no fear, though. As Bacon reminds us, come this April (March, actually!), rest assured that as soon as Jeter makes one of those jump throws from the hole at short, any lingering doubt caused by this week's press will melt away.

As discussed previously in this space, Jensen's system is but one of several built on detailed play-by-play data that includes fine-grained data on all balls put into play and their locations. SAFE happens to use data from Baseball Info Solutions (BIS), while other similar systems use data from STATS, Inc. Regular readers will also know that in recent weeks we've been discussing a fielding system based on play-by-play data that omits the detailed location data which I've dubbed Simple Fielding Runs, or SFR for short. While SFR is inherently not as accurate as SAFE or the metrics in that family, it does have the advantage of being able to be run on datasets that are necessarily not as complete. One of those datasets is the minor league data we receive here at Baseball Prospectus, and so in a previous column I ran the numbers for minor league infielders and provided the data in a spreadsheet. Today we'll complete that effort by running the numbers for minor league outfielders.

Leader Boards and More Leader Boards

Before we get to the numbers for the minor league outfielders, a couple of explanations are in order. First, the algorithm used to calculate outfield SFR for minor leaguers is identical to the one used for major leaguers. In other words, each outfielder is compared to all other outfielders at every park and position at which he played in 2007, taking into account the context of batter handedness and hit type. The difference, however, is that for the minor leagues I'm working with three years of data (2005-2007), so the park effects that are built into the system rely on just three years of data, whereas for major leaguers I've used five years (2003-2007). Obviously, when you're building a system such as this, based on probabilities and not on observation, the larger samples you have to compare to, the better off you'll be.