# The trouble with measuring outfield defense

As OPA! (my new Retrosheet-based fielding metric) continues to develop, I’ve run into a bit of a snag in the outfield.  I set out to look at something that I’ve sorta looked at before.  Ground balls obviously tell us a lot about infielders.  But what can they tell us about outfielders?  After all, once a ground ball has made it through the infield, it’s virtually guaranteed to be a single at least.  But here’s where the outfielder comes in.  The outfielder in this situation is charged with making sure that the ball stays just a single and not a double.  OPA! rewards the addition of out probability, and there’s no out probability to be had for the outfielder here, just extra bases to be taken away.  But, those are valuable bases.  Can we quantify that?
Again, for a little baseline work, I went back to 1993-1998, when Retrosheet had hit location data.  I was able to get an idea of roughly where ground balls that eventually went to the left fielder went.  Did they mostly go down the line?  Or were they shots by the third baseman.  Retrosheet gave me a rough idea.  With that in mind, I could then figure out how many in each zone went for an extra base hit and with a little math, I was able to get a good guess on what the expectancy for a single was for the average ground ball that made it through to left field.  I even controlled for pitcher and batter handedness.  Aren’t I a sweetie?
Now, of course, where the ball is rolling is going to make a difference.  A ground ball down the line has a good chance of getting into the corner for a double.  A grounder through the hole between third and short (and thus generally right at the left fielder) will probably be just a single.  Hit location data on recent year’s data would be invaluable here in figuring out how to properly credit or debit outfielders for the doubles they give up for being so slow.  But, we must go without.  We do know that a fielder with greater range will get to more of those balls and the likely result is that the runner will have to settle for a single.  Low marks on preventing XBH’s might also be a reflection of a weak armed fielder (he gets to the ball, but the runner knows his arm is a noodly appendage, so he takes second anyway.)
The best left fielder in 2007 at this particular skill of keeping grounders as singles, rather than letting them become doubles?  Would you believe Barry Bonds*???  Yes, Barry Lamar Bonds* was the best left fielder in baseball last year at cutting a ground ball single off and preventing extra bases.  And here, we have a problem of the smell test.  Bonds* was 42 and known to have creaky knees, and there was talk that if he were to come back to baseball this year, it would have to be as a DH because he’s not much of a left fielder any more.  On the other hand, this is a guy who used to have some wheels (he has 500+ career stolen bases), so even if that speed is fading, Bonds* might still have decent enough range for getting to balls in the outfield.
Then again, Bonds* also may have had some help to get to the top.  (No, not that kind of help.  The legal kind.)  Playing in left field in San Francisco last year, he had the benefit of playing behind Rich Aurilia (pretty good range) and Pedro Feliz (very very good range) at third and Omar Vizquel (range is over-rated, but he’s no slouch) at short.  Maybe the balls that would be hard to cut off in the gap had already been taken care of by those gentlemen and so Bonds didn’t have to do as much.  The problem with a Retrosheet based system is that it is blind to where the ball is and there’s just no way around that.  So, I present to you the concept of extra bases prevented (normed off the league average expectation), with the caveat that this one might not be as reliable or valid as a measure of range.  When I get to flyball range, we can talk.
Taking a look at the rest of the top 5 for preventing extra bases (minimum 50 ground balls fielded), we find

1. Barry Bonds*: 0.10 extra bases prevented per ball above the average fielder
2. Carl Crawford: 0.07
3. Eric Byrnes: 0.07
4. Craig Monroe: 0.07
5. Hideki Matsui: 0.07

The bottom five, for the morbidly curious:

1. Geoff Jenkins: -0.02
2. Luis Gonzalez: -0.04
4. Carlos Lee: -0.11
5. Moises Alou: -0.11

Left fielders aren’t known for their defensive prowess, but Crawford and Byrnes make sense as gentlemen who should be in the high range on this particular measure.  Craig Monroe has been rated highly on range measures in the past.  The bottom five are guys who are either old or terminally slow.  But Hideki Matsui is #5?  Well, I’ve made it clear that he doesn’t exactly play in front of a range-y shortstop.  A-Rod rates out as a below average 3B as well.  How did that happen?  Luck?  Some sort of park effect?  Oddly enough, Matsui does score well on RZR for last year (4th among qualifiers), although so does Geoff Jenkins.  This is just one isolated skill that I’m measuring vs. the more complete RZR, but it does provide some comfort that I might have some validity here.
One thing that I will eventually do is to test the stability of these stats over time to see whether they hold up year after year.  I’m curious to see whether this stat is like BABIP in that it is mostly luck from year to year or whether it is stable.  If there’s no stability, and this is all mostly luck, well then I suppose I’m chasing a whole lot of nothing and this skill isn’t really worth measuring.  As I mentioned before, this is a work in progress.