I recently downloaded the MYSQL version of the retrosheet database, and am ready to look at the objective probabilistic model of range (PMR). It’s objective since it does not include the subjective vector and hard hit parameters that reporters enter in both the STATS and BIS data. In other words, the models for objective PMR just use batter handedness, pitcher handedness, batted ball type, and park to build the model. I’m using data from 2005-2011 for visiting teams only for model. Note that this means that 2011 fielding has no influence on the models.
As with the normal PMR model, I’m measuring the defensive efficiency record (DER) of the team against the expected DER. The following table shows the results for the 30 teams.
Team | In Play | Actual Outs | Predicted Outs | Actual DER | Predicted DER | Index |
TBA | 4157 | 3007 | 2917.7 | 0.723 | 0.702 | 103.1 |
ANA | 4413 | 3106 | 3030.1 | 0.704 | 0.687 | 102.5 |
CLE | 4462 | 3083 | 3024.5 | 0.691 | 0.678 | 101.9 |
CIN | 4053 | 2867 | 2820.5 | 0.707 | 0.696 | 101.7 |
BOS | 4167 | 2926 | 2884.9 | 0.702 | 0.692 | 101.4 |
WAS | 4259 | 2967 | 2930.7 | 0.697 | 0.688 | 101.2 |
ATL | 4054 | 2825 | 2792.6 | 0.697 | 0.689 | 101.2 |
TEX | 4161 | 2923 | 2891.1 | 0.702 | 0.695 | 101.1 |
ARI | 4075 | 2867 | 2838.0 | 0.704 | 0.696 | 101.0 |
SEA | 4298 | 3000 | 2972.1 | 0.698 | 0.692 | 100.9 |
SLN | 4162 | 2855 | 2833.0 | 0.686 | 0.681 | 100.8 |
TOR | 4324 | 3002 | 2981.6 | 0.694 | 0.690 | 100.7 |
PHI | 4016 | 2828 | 2808.9 | 0.704 | 0.699 | 100.7 |
MIL | 3955 | 2732 | 2712.7 | 0.691 | 0.686 | 100.7 |
LAN | 3735 | 2619 | 2601.6 | 0.701 | 0.697 | 100.7 |
FLO | 4140 | 2862 | 2842.5 | 0.691 | 0.687 | 100.7 |
SDN | 3921 | 2764 | 2748.8 | 0.705 | 0.701 | 100.6 |
NYA | 4277 | 2936 | 2922.6 | 0.686 | 0.683 | 100.5 |
DET | 4232 | 2932 | 2922.7 | 0.693 | 0.691 | 100.3 |
CHA | 4337 | 2978 | 2971.6 | 0.687 | 0.685 | 100.2 |
CHN | 3985 | 2706 | 2699.4 | 0.679 | 0.677 | 100.2 |
SFN | 1992 | 1389 | 1390.7 | 0.697 | 0.698 | 99.9 |
HOU | 3966 | 2694 | 2698.4 | 0.679 | 0.680 | 99.8 |
PIT | 2118 | 1447 | 1450.1 | 0.683 | 0.685 | 99.8 |
COL | 4022 | 2760 | 2765.1 | 0.686 | 0.688 | 99.8 |
KCA | 4426 | 3045 | 3059.1 | 0.688 | 0.691 | 99.5 |
MIN | 4491 | 3042 | 3066.5 | 0.677 | 0.683 | 99.2 |
OAK | 4224 | 2917 | 2942.4 | 0.691 | 0.697 | 99.1 |
BAL | 4416 | 3018 | 3060.8 | 0.683 | 0.693 | 98.6 |
NYN | 4286 | 2926 | 2993.4 | 0.683 | 0.698 | 97.7 |
If you look at the ranking of the teams using UZR/150, there is some agreement between the two systems. Four of the top five teams in objective PMR also ranke in the top five in UZR/150. There is a glaring difference, however, and that’s the Cleveland Indians. UZR ranks the Indians as the 2nd poorest fielding team in the majors, while objective PMR ranks them second. Where are the differences?
Position | UZR/150 Rank | Objective PMR Rank |
Pitcher | N/A | 2 |
Catcher | N/A | 6 |
First base | 28 | 13 |
Second base | 30 | 25 |
Third base | 5 | 1 |
Shortstop | 28 | 28 |
Leftfield | 21 | 20 |
Centerfield | 27 | 23 |
Rightfield | 12 | 1 |
As you can see, UZR does not rank ranges for pitchers and catchers, and PMR rated both those positions well for the Indians. Apart from that, most positions are in decent agreement. Both systems rank the shortstops poorly and the third basemen well. The two biggest discrepancies occur at first base and rightfield. I’m not arguing that one is right and one is wrong. It could be that a bias exists in that direction on the field. Maybe it’s the camera angle, so the balls to that side of the field look easier to field than they actually are. On the other hand, the vectors and batted ball velocities may add important information to the model that knocks the Indians down.
It could also be both those effects are in place, and the Indians real fielding value lies somewhere between the two.
The big differences make for the most interesting comparisons. I’d love to be able to dig deeper into this to see what the distribution of fly balls and ground balls is for the Indians, to see which model is doing a better job of creating a model.
I’ll continue this series looking at teams and fielders by position.
The information used here was obtained free of charge from and is copyrighted by Retrosheet. Interested parties may contact Retrosheet at 20 Sunset Rd., Newark, DE 19711.
Not to make you go too remedial, do you have a PMR primer? I think i know what DER is, but I’m not sure what to make of “actual” and “predicted” values therefor. Thanks!
Scooter » Yes, I have one somewhere. If you search far enough back in the Probabilistic Model of Range category, you’ll find it.
I think I also did a video a number of years ago. I’ll try to post that.
Great stuff. Are you going to show the data for all players and teams at some point?
Plank » Yes, at least for the regulars at each position.
I realize this is for last year and it begs a question about the Orioles. The Orioles have, for some time, based their future on drafting pitchers.
We’ve seen the lack of development of these pitchers (notably Tillman) and even the regression (notably Matusz). Yet by building a team that lets lots of extra balls get through, the Orioles are undermining these young arms by making them work much harder than necessary.
In fact the big difference for the Rays from 2007 to 2008 is that they went from bottom of the DER standings to the top. That helped their pitchers and was the “secret” of their success. If the Orioles want to follow that model they really need to fix the defense. (And now that I see how good they were last year at short, I shudder to think what the the other positions were!)
David » You have a point, but remember, the Rays fixed the pitching before they fixed the defense. It was clear looking at the 2007 Rays that their pitchers were not living up to their FIP. Once the pitching staff is in place, then the Orioles can figure out how to plug the defensive holes.