Print this Article
 E-mail this Article

Infield Defense — Back to Basics

Simplicity, simplicity, simplicity! I say, let your affairs be as two or three, and not a hundred or a thousand; instead of a million count half a dozen, and keep your accounts on your thumb-nail.
— Henry David Thoreau

In recent years, methods for evaluating defense have been springing up like mushrooms after a spring rain. We've got UZR and PMR and the plus/minus system, all based on play-by-play data. Then there are the zone rating-based systems, one from Chris Dial, another by Sean Smith and more stuff from folks working with the BIS zone rating data that we publish here at THT. And let's not forget those heroic attempts to evaluate fielding using traditional stats: Win Shares and FRAA.

The evaluation of defense is a fascinating subject, and I've read through much of the material that documents the above systems. And you know what? I still don't know which one works best or, more generally, how well any of them work. These systems are either a) very complicated or b) not fully explained (usually both), so it's very hard to wrap your head around them. I know, I've tried.

So, I've decided to go back to basics, and I'm starting with infield defense. I've got some precise hit location data for the 2005 and 2006 seasons that I obtained from mlb.com and I want to see if I can take a Thoreau-vian look at it, see if I can evaluate infield defense "on my thumb-nail."

Where do ground balls go?

This seems like a very simple question, but I've never seen anybody give the answer. For example, we are told that shortstop is the most important infield position (I'm excluding catcher from the infield), presumably because more balls get hit to the shortstop than to any other infield position. That's true, but how many more? And sabermetric wisdom has it that first base defense isn't that important, simply because first basemen field fewer balls than the other positions. Ok, but how many fewer? Do they field half as many as the third baseman? Eighty percent as many?

Then I just counted up the number of ground balls hit into each slice. Simple, right? Thoreau would be proud. The plot below shows the results for the roughly 100,000 ground balls hit during the 2005-2006 seasons (bunts have been excluded). The red and green curves show the results for right- and left-handed batters separately.

I don't know about you, but I find this plot fascinating (I need to get out more, I know). First of all, I wasn't expecting this shape to the ground ball distribution. I expected to see two humps around the shortstop and second base positions, but I am surprised to see that double-peak structure on the left side. Another curious thing is that, for a right-handed batter, it looks like almost as many balls are hit towards the third baseman as are at the shortstop, maybe even more.

I also didn't realize how few balls were actually hit up the middle. As you can see from the RHB and LHB curves, batters overwhelmingly tend to pull ground balls, which I knew. What I didn't know is that a right-handed batter will hit more balls at the second baseman then he will up the middle (see the peak in the red curve near 60). Same thing for a lefty hitting balls towards short.

Who fields the grounders?

O.K., we've seen where batters hit ground balls; now we'll have a look at how often they get turned into outs and who does the turning. The following graphic shows just that: the blue curve is just the distribution of all ground balls, it's the same as we saw in the previous plot. In red, we see the balls that were converted into outs (including plays scored as fielder's choice). The dotted lines show where each infield position recorded his outs.

Hmmm, there's lots of food for thought here. For example, you can see that there is quite a bit of overlap between the area covered by the third baseman and the area covered by the shortstop (and the same is true on the other side of the infield). We've all seen the third baseman cut in front of the shortstop on a slow chopper many times. This overlap region makes it difficult to assign credit to specific fielders for this kind of play.

...it's worth noting that Peralta didn't have the benefit of great support to his right and left; Boone lost almost as much ground in the field as he did at the plate (according to both scouts and metrics), while Ronnie Belliard's range at second left a lot to be desired. [emphasis added]

Shortstops field only 6% more balls than second baseman do, which doesn't seem like much, when you consider that shortstops are generally viewed as much more important than second baseman.

Conversely, despite having almost as many balls hit toward them, third baseman field far fewer ground balls than their shortstop colleagues. This makes sense since the third baseman is playing closer and has less time to react to hard-hit balls. It is called the Hot Corner, after all. First baseman field the fewest balls of all, 1) because fewer are hit their way and 2) because they often play out of position in order to hold a runner on first base.

Team Infield Defense

Since I don't really know how to divvy up credit (or blame) for individual fielders, how about if we look at team infield defense? Check out this graph:

To start off, focus on the blue curve in the upper plot. This is the mlb average fraction of balls turned into outs as a function of where the ball was hit. The dips in the blue curve show show the "holes" where most hits get through. It's also clear that balls hit towards the shortstop or the second baseman are easier to field than balls hit at the third or first baseman, i.e. the two peaks in the middle are higher than the two lateral peaks.

Okay, now look at the red curve in the upper plot. That is what the Cleveland Indians did in 2006. Tribe infielders performed below average wherever the red curve is below the blue curve, which is just about everywhere. It's evident from this plot that Peralta and Belliard showed quite poor range in 2006. Both were O.K. at balls hit at them, but they were both below average when asked to move off their set position. They were both particularly bad to their left.

The lower plot shows the number of plays made above or below average. As noted on the graphic, the Indians turned 90 fewer ground balls into outs than the average mlb team would have, given the same ground balls. That's an astonishingly bad performance, by far the worst of any team in 2006. Furthermore, since the overlap between SS and 2B is small and if we neglect the influence of pitcher fielding, we can divide the total into left and right side of the infield. We see that the Indians were equally bad on the two sides of second base.

I note that this plot is similar to the type of plot that David Pinto produces with his PMR system. David's plots, however, are based on his complex system and are shown separately for each player. In trying to stick to my Thoreau-vian principles, I haven't made a bunch of adjustments, nor have I tried to apportion credit to individual fielders.

The 2006 Baltimore Orioles also were terrible, according to my simple method. As you can see below, the O's were uniformly terrible on any ball hit anywhere near the center part of the diamond.

Both Miguel Tejada and Brian Roberts, who played the majority of the Orioles games at SS and 2B, seemed to play too far towards the hole, leaving the middle infield virtually uncovered. The O's corner infielders, Mora and (mostly) Millar, were pretty terrible also.

Why all this negativity?

Let's look at the graphs for a couple of teams with good infield defense.

Everybody knows that Adam Everett is the best-fielding shortstop in the game and my little pictures seem to bear that out.

Actually, the Houston infield looks pretty darn good across the board. I am surprised to see average play coming out of second base, which was mostly manned by Craig Biggio in 2006. Hmm, maybe this method is too simple. Both Ensberg and Berkman got to a lot of ground balls last year.

I hadn't heard much about the Rockies' infield defense, but by my measure it was excellent in 2006. The left side of Atkins and Barmes was especially good, but Helton and Carroll on the right side held up their share of the bargain.

This is not a system

I want to make clear that this is not a defensive evaluation system. It only takes into account one aspect of ground balls, namely, where they are hit. There are other variables that likely have an effect on the ability of a player to turn a ground ball into an out: speed of the batted ball is clearly one such variable. The number of outs and possible runners on the bases affect positioning and hence the out conversion rate. Other effects may also be important: the ballpark, handedness of batter or pitcher, etc. I've ignored all those things, because I want to see the "raw" data first, in its natural purity, before it gets gussied up with adjustments.

Maybe in a future article, I'll have a look at some possible adjustments, try to get a feel for how big they are and if they are really necessary. For now, I'll leave you with a list of team infield performance in 2006, based on my simple method.

Infield Plays Made Above Average in 2006
Team      Total     Left      Right
HOU        44        33        11
COL        43        35         8
DET        40         8        32
SDN        30        21         9
SLN        29         9        20
SFN        29        25         3
NYN        28        14        14
PHI        23         7        16
TOR        18       -10        28
FLO        15        10         5
MIL         5        13        -7
ARI         5        14        -9
CHA         2        18       -16
ATL         1        -1         1
BOS         0         8        -8
OAK         0        -7         7
SEA        -3        -5         2
CHN        -8       -10         2
LAN        -8         6       -14
MIN       -11        12       -22
TEX       -12       -12         0
KCA       -13       -29        16
NYA       -14       -12        -2
ANA       -15       -23         8
PIT       -16       -14        -2
WAS       -22       -10       -13
CIN       -25       -22        -3
TBA       -38       -10       -28
BAL       -40       -26       -14
CLE       -90       -47       -43