THE BOOK--Playing The Percentages In Baseball

Monday, September 19, 2011

Is where you play on the field conducive to better hitting performance?

I’ve ALSO been meaning to do this study for a long time.  All data from 1993-2010.  I also limit the data to players aged 25-29.  I ensure that a player is equally weighted in the two pools being compared (and that weight is the lesser of the plate appearances in the two fielding positions being compared).

(Imagine there’s three or seven paragraphs of a yapfest going on here.  I’m not going to do it.)

Ok, now, thanks for your patience in wading through all that.  Now onto the data.

Unlike the batting lineup, where I have enough data to make all possible combinations permutations (72… stay in school kids), I don’t have this luxury on the fielding side.  On the lineup side, the least matching pair was cleanup hitter and last hitter, where I had “only” 3740 plate appearances (while I had 94,024 matching PA for sixth and seventh hitter).

On the fielding side?  I have only 54 combinations for the nine fielding positions (excluding pitcher, including DH), with a minimum of 2000 plate appearances.

The biggest gap, by far, involved the DH.  Remember, I’m only looking at players aged 25-29.  I look at how they did at DH and at each of the other fielding positions.  The same players involved, and for each pair of positions, the player is equally weighted (proportionate to the lesser of his plate appearances at the two positions).  It is quite a striking difference:

Across the board, a player hits worse at DH than at each of the other fielding positions.  (You’ll note I didn’t have enough DH/SS in there.) The average difference was 16 points of wOBA.  Each individual position shows enough statistical significance on its own to jump out.  Taken together, and it’s overwhelming.

16 points of wOBA is 9 runs over the course of a season.

Now, I did not try to distinguish between “lifers” at DH, and just those who are transient.  It’s possible the reason some players were DHing was because they had some injury that was hobbling them on the field, but was hoped to be masked while hitting.  Could be many reasons, really.

***

How about the other positions?  If we look at just LF, CF, RF (that is, players who played LF and CF, or CF and RF, or LF and RF), and we see almost no difference in performance.  As an example, of the 45,030 matching PA for players who played both LF and RF, they batted .344 while playing in left and .345 while playing in right.  At 0.3 standard deviations, it’s a difference not even worth pointing out.  (Which I did, which is weird.)

How about in the infield (2B, SS, 3B)?  In that case, there was one HUGE standout: .320 while playing 3B, but .331 while playing 2B (3.5 standard deviations apart).  3B to SS also had a bit of difference: .321 at 3B, .326 at SS (only 1.0 standard deviations).  It’s possible that playing 3B takes a bit of toll on the hitting side.

How about infield/outfield (the above positions)?  When playing 2B, players hit .320, but playing in the outfield was .329.  That difference was 2.0 standard deviations, so it’s possible that it’s a bit easier to hit while playing in the outfield.  Then again, playing 3B, players hit .334, and .336 in the outfield.  We had just figured that playing at 3B is harder than playing at 2B, when it comes to hitting.  But now, we see that 3B/OF is fairly close, but 2B/OF is far apart.

If we just look at the IF/OF comparison as a group, we get: .323 in the infield, .328 in the outfield.  That gap is 1.9 standard deviations.  So, I think we can say that it’s easier to hit while in the outfield, but it’s not that big a difference.

Now, how about 1B?  Surprisingly, it’s harder to hit at 1B!  It was .347 at 1B and .352 at the other six positions I noted above. 

***

Finally, catcher.  Unfortunately, there’s not many players that play catcher and some other position, other than 1B and DH, at the ages of 25-29.  But, of those that did, oh boy!  Get ready for this.  If you played catcher and 1B, you hit .334 while catching and .351 while at 1B.  That’s a 18 (rounded) point difference and is 2.0 standard deviations from the mean.  And if you played catcher and DH, you hit .341 at catcher and .321 at DH, which is 2.2 SD from the mean, going the other way.

In short, the data is showing this:
.348 at 1B
.330 at C
.310 at DH

Yowza.

Now, before we get too far here, it’s worth pointing out we have tons of 1B/DH comparison, without needing to go through the catcher.  Those guys hit .364 at 1B and .349 at DH, a 14 (rounded) point difference (and 3.7 standard deviations).

Let’s go back to catcher.  If I relax the standard, and only look for 1000 matching PA, I also get 3B, LF, and RF.  Including 1B, then this is the comparison:
.323 at catcher
.342 at 1B, 3B, LF, RF

That 19 point difference is 3.3 standard deviations from the mean.  So, we can safely say that it’s harder to hit as catcher than hit at any other position.  Is it really a 19 point difference?  Well, the only thing we can say is that we are almost positive it’s greater than 0.  This is where Bayes would help.  If we had catchers tell us that their body can’t really handle it as easily when behind the plate than when they play the field, then we would use a better prior.  Heck, our prior could have been even 30 points, and seeing only 19 would understate the difference.  The problem is that we have no way of quantifying a good prior.

I’m inclined to give at least a 10 point difference as the true difference, which is 6 runs of impact.

***

Anyway, lots more work to be done.  But, without question, we need to apply a huge bonus to guys who DH, and to guys who play catcher.