Chronicles of the Lads

PMR, LEFTFIELDERS
David Pinto has begun posting his probabilistic model of range defensive numbers. As in years past, here is my attempt to convert those figures, which are presented in outs above and below average, to runs. My method of doing this, cribbed from Chris Dial, is explained here, and data from past years can be found along the sidebar.

I've got some technical things to say on this, but I'll toss those down at the end.

David has begun with leftfielders, so here we go. Every player is presented with an actual total in addition to a rate total (based on a typical number of opportunities for each position, which should be roughly equivalent to around 150 games).

Player            Runs Above Average          
                 Total   Per 270 Opp         
Melky Cabrera   14.9236    20.2431
Matt Diaz       10.7749    19.3897
Reed Johnson    10.4507    24.2351
Dave Roberts    10.3676    12.3570
Matt Murton      8.4553     9.9332
Brandon Fahey    7.9482    23.4690
Emil Brown       6.7177    11.7079
Ryan Langerhans  6.0609    11.0043
Alfonso Soriano  5.8780     4.9762
Andre Ethier     5.3542     8.7318
Jason Bay        5.0965     4.4407
So Taguchi       4.7972    15.9453
Jason Michaels   4.4812     5.8000
F. Catalanotto   4.0073     8.0040
Cliff Floyd      3.5085     6.5885
David DeJesus    3.3090     6.6663
Juan Rivera      2.9099     6.4137
Garret Anderson  1.5464     2.1959
Marcus Thames    0.8896     3.4846
Barry Bonds      0.7898     1.1401
Kevin Mench      0.7566     2.5828
Jay Payton       0.3824     0.8711
Luke Scott      -0.1580    -0.5253
Jeff Conine     -0.3492    -1.0663
Brad Wilkerson  -0.4074    -0.7885
Luis Gonzalez   -1.1640    -1.2209
Carl Crawford   -1.7792    -1.5795
Nick Swisher    -2.3362    -3.6502
Josh Willingham -2.6106    -3.3703
Preston Wilson  -3.6582    -6.1578
Craig Monroe    -3.6997    -5.7926
Chris Duncan    -4.6808   -17.6436
Matt Holliday   -7.0669    -6.6832
Adam Dunn       -7.1002    -6.6670
Bobby Kielty    -7.1750   -21.8577
Pat Burrell     -7.8152    -9.8419
Scott Podsednik -8.8128    -9.3093
Raul Ibanez    -11.3403    -9.7005
Carlos Lee     -13.8844   -15.3828
Manny Ramirez  -22.4145   -29.9660

Thanks to Chris Dial, who has done work on converting zone rating to runs, we can see how some players came out in PMR vis-a-vis zone rating. This is alphabetical by first name, and I've rounded off the PMR Runs to integers for convenience of comparison:

Player          PMR-R    ZR-R    Diff
Adam Dunn        -7      -12       5
Alfonso Soriano   6        5       1
Andre Ethier      5        2       3
Barry Bonds       1        0       1
Brad Wilkerson    0       -5       5
Carl Crawford    -2        8      10
Carlos Lee      -14       -3      11
Cliff Floyd       4        1       3
Craig Monroe     -4        1       5
Dave Roberts     10       14       4
Emil Brown        7        9       2
F. Catalanotto    4       -2       6
Garret Anderson   2       -1       3
Jason Bay         5       -6      11
Jason Michaels    4        3       1
Josh Willingham  -3      -15      12
Luis Gonzalez    -1        2       3
Manny Ramirez   -22      -32      10
Matt Diaz        11        6       5
Matt Murton       8        6       2
Matt Holliday    -7        3      10
Melky Cabrera    15       -6      21
Nick Swisher     -2        4       6
Pat Burrell      -8        0       8
Preston Wilson   -4      -20      16
Raul Ibanez     -11        4      15
Reed Johnson     10        5       5
Ryan Langerhans   6       11       5
Scott Podsednik  -9        7      16

Though both systems see many players similarly, there are some pretty big discrapencies, as well. PMR takes park into account, where ZR does not, so that may be one source of difference. Of course, they also are derived from different data sources and tracking systems, PMR from Baseball Info Solutions and ZR from STATS, LLC.

The correlation between the two lists is .519, if that has meaning for you.

I don't know eough about the players for whom there is a big difference to really speak to why that might be, but there's the data.

***

TECHNICAL NOTE: There is one difference between my conversions this year and in past years. What PMR does is, for every batted ball in the world, gauge the probability of any one fielder making that play. So, for a ball hit to X place, the shortstop might make an out 75% of the time. If any shortstop has 100 balls hit there, he's expected to make 75 outs. If our guy makes 85, he's +10 plays (which I then convert into runs), if he's 65, he's -10. I call the 75 outs "predicted outs" here.

Now, that hypothetical 75% figure was, in years past, a multi-year average. As a result, if you look at any given year at a position, you find some weird things. The shortstops in 2004 (for whom Pinto provided data), for instance, averaged around -16 runs against predicted in a full season. When doing these conversions, I recentered it so that the -16 would be average.

I didn't have to do that this year. David, because of his data (read about it here) is using 2006 data only to derive these numbers, which makes my job a lot easier.

If you add up all the leftfielders here, you'll see that they came out to around +13 runs above average. That means that all the other leftfielders in baseball added up to -13. Before, I used to have to guess and just kind of assume the population we had for PMR represented the population at large. That may have been a bad assumption, but at least now we don't have to guess.

As for another technical note, last year David had an "original model" and a "smoothed visitor" model that dealt with parks slightly differently. I don't know which model he's currently employing.

Labels: David Pinto, fielding, probabilistic model of range