February 09, 2005
Probabilistic Model of Range, Leftfielders
Without further ado, the range rankings for leftfielders.
Probabilistic Model of Range, Leftfielders 2004, 1000 balls in play.
Player | InPlay | Actual Outs | Predicted Outs | DER | Predicted DER | Difference |
Aaron Guiel | 1012 | 83 | 73.18 | 0.082 | 0.072 | 0.00971 |
Carl Crawford | 3170 | 273 | 246.74 | 0.086 | 0.078 | 0.00828 |
Charles W Thomas | 1710 | 133 | 119.60 | 0.078 | 0.070 | 0.00783 |
Reed Johnson | 1412 | 98 | 92.77 | 0.069 | 0.066 | 0.00371 |
David Dellucci | 2077 | 152 | 144.33 | 0.073 | 0.069 | 0.00369 |
Terrmel Sledge | 1767 | 133 | 126.73 | 0.075 | 0.072 | 0.00355 |
Kevin Mench | 1120 | 73 | 69.64 | 0.065 | 0.062 | 0.00300 |
Carlos Lee | 3902 | 283 | 273.04 | 0.073 | 0.070 | 0.00255 |
Eric Byrnes | 2676 | 172 | 166.39 | 0.064 | 0.062 | 0.00209 |
Craig Monroe | 1369 | 102 | 99.95 | 0.075 | 0.073 | 0.00150 |
Jayson Werth | 1592 | 115 | 112.70 | 0.072 | 0.071 | 0.00144 |
Hideki Matsui | 4326 | 303 | 302.47 | 0.070 | 0.070 | 0.00012 |
Brad Wilkerson | 1300 | 94 | 94.05 | 0.072 | 0.072 | -0.00004 |
Jay Bay | 2930 | 206 | 206.55 | 0.070 | 0.070 | -0.00019 |
Dave Roberts | 1298 | 87 | 87.57 | 0.067 | 0.067 | -0.00044 |
Barry Bonds | 3498 | 214 | 215.76 | 0.061 | 0.062 | -0.00050 |
Moises Alou | 3746 | 239 | 241.33 | 0.064 | 0.064 | -0.00062 |
Jose Guillen | 3464 | 264 | 266.40 | 0.076 | 0.077 | -0.00069 |
Luis Gonzalez | 2748 | 162 | 163.92 | 0.059 | 0.060 | -0.00070 |
Dee Brown | 1413 | 93 | 94.28 | 0.066 | 0.067 | -0.00091 |
Ray Lankford | 1175 | 82 | 83.16 | 0.070 | 0.071 | -0.00099 |
Cliff Floyd | 2759 | 164 | 167.47 | 0.059 | 0.061 | -0.00126 |
Raul Ibanez | 2920 | 227 | 230.68 | 0.078 | 0.079 | -0.00126 |
Larry Bigbie | 2793 | 214 | 217.89 | 0.077 | 0.078 | -0.00139 |
Manny Ramirez | 3293 | 198 | 204.49 | 0.060 | 0.062 | -0.00197 |
Miguel Cabrera | 1464 | 92 | 95.01 | 0.063 | 0.065 | -0.00206 |
Jeff Conine | 2139 | 175 | 180.21 | 0.082 | 0.084 | -0.00244 |
Geoff Jenkins | 4131 | 261 | 273.14 | 0.063 | 0.066 | -0.00294 |
Lew Ford | 2074 | 149 | 156.19 | 0.072 | 0.075 | -0.00347 |
Matt Lawton | 3291 | 231 | 242.50 | 0.070 | 0.074 | -0.00350 |
Adam Dunn | 4196 | 250 | 266.18 | 0.060 | 0.063 | -0.00386 |
Bobby Kielty | 1105 | 71 | 75.62 | 0.064 | 0.068 | -0.00418 |
Pat Burrell | 3261 | 216 | 231.22 | 0.066 | 0.071 | -0.00467 |
Rondell White | 1917 | 126 | 136.17 | 0.066 | 0.071 | -0.00530 |
Lance Berkman | 1741 | 93 | 102.23 | 0.053 | 0.059 | -0.00530 |
Eric Young | 1019 | 58 | 63.55 | 0.057 | 0.062 | -0.00545 |
Craig Biggio | 1902 | 116 | 126.52 | 0.061 | 0.067 | -0.00553 |
Matt T Holliday | 2963 | 176 | 193.05 | 0.059 | 0.065 | -0.00575 |
Eli Marrero | 1157 | 80 | 87.90 | 0.069 | 0.076 | -0.00682 |
Shannon Stewart | 1937 | 103 | 119.94 | 0.053 | 0.062 | -0.00874 |
Ryan Klesko | 2237 | 134 | 155.40 | 0.060 | 0.069 | -0.00957 |
Our eyes often deceive us when it comes to defense, but not in the case of Ryan Klesko. He's just as brutal as he looks. It's also interesting to note that some of the older players (Bonds, Alou, Luis Gonzalez) are right where they should be, getting the balls at the expected rate. Sig Mejdal wrote the injury prediction section of The Bill James Handbook 2005 and one thing he's looking at is how aging and experience effect skills in the game. For example, he's found that the physical process of aging hurts HR production, but experience helps HR production. I'm wondering if that is going on here; these left fielders have so much experience in the outfield that they can make up for their old legs with positioning.
Posted by David Pinto at
10:17 AM
|
Defense
|
TrackBack (0)
As a Tigers fan, I find it interesting that Rondell White is so low on your list, yet he had the highest zone rating of all left fielders (.925). Would this suggest that the zones used for ZR are too small, for what average LFs should be actually responsible for?
A big thanks to David for making excellent use of very granular data, and, more importantly, for sharing the results with us!
Ideally, we would all quit our jobs, and spend 14 hours a day poring through this, and we'd have the answers we are all searching for in 2 months.
Pretty high rate for Byrne, which conflicts with his reputation. Think Beane already knew this about Charles Thomas, though?
I would bet that Beane's definitely knew, and that his scouting reports also would have said the same thing.
It interests me that people would think Thomas was a surprise-- anyone who watched Braves games surely knew he'd be way up at the top-- there were folks who actually claimed that Andruw was the worst of the Atlanta outfielders, tho' I wasn't one of them. I also had no idea Byrne had a bad rep-- he looked quite good the couple of times I saw him... overall this matches pretty closely the ranking I'd have given them off the top of my head leaving out a few people I've never seen like Johnson and Mensch...
Once again, the problem of comparing a fielder to himself shows up. There is no way Ramirez or Bonds is as good a fielder as a guy like Ichiro. But because they're being compare to themselves, they end up looking good. In fact, here is a simple exercise I woud love for you to undertake: show the numbers for various positions played fielders that are known to be very good/bad play, i.e. left field in Fenway Park or RF in Safeco. You would expect the numbers to be within +/- 5 runs (or in other words, to average out). Instead, I'll bet you'll find that LF in Fenway is extremely positive and RF in Safeco is very negative. Another way to check this would be to post the home/road splits for these types of players. I bet Ramirez is much worse on the road than at home and that Ichiro is much better on the road.
***
BTW, since my first explanation of why you need to remove all home players when finding the baseline against which to compare sucked, let me take another crack at it:
Let's say Seattle has two RF, Ichiro Suzuki and Manny Ramirez, who split time evenly but only play at home. By a freak of nature, they have the same number of BIP in each slice, and the same exact number of L/R on the mound, etc. Essentially, they should have the exact same expected DER.
However, when you are measuring Ichiro's expected DER, and remove him, you'll find an xDER of, say .67 because 2/3 of the xDER comes from opposing fielders (.70) and 1/3 (because Ramirez only plays half the games) comes from Ramirez (.60). But when you go to find Ramirez's xDER, you get .72 because 2/3 comes from opposing fielders (again, .70) and 1/3 comes from Ichiro! (.76). The difference in xDER is huge, even though we know that the two should have the same exact xDER. But if you remove all home fielders, both end up with an xDER of .70. Easy as pie, and completely correct. You MUST remove all home fielders before finding xDER.
So what you are saying David is that part of Charles Thomas surplus outs over prediction shows up because the prediction includes 29 games of indifferent-and-injured Chipper, while at the same time he's being predicted against his own 70 games which cuts him back some? That would skew almost every result both ways-- I guess the outcome wouldn't be random, but it would be pretty fuzzy...
Even if you take out home players, won't there still be some distortion from the unbalanced schedule? If Tree Stump Jones is playing more visiting games in Yankee Stadium for Baltimore than Deerfoot Smith does for Detroit won't that have the same effect to a lesser degree?
That's a statistical risk you have to take (that the away fielders won't average out), but one that is much smaller and will skew the numbers much less than taking out only the home fielder you're measuring. Just read my example, it simply makes sense that Ramirez and Ichiro should have the same xDER and if they don't, the method is incorrect. Thus, taking out only the player you're measuring is incorrect. And we can already see that keeping the fielder in the sample really skews xDER. The only way that remains is to remove ALL home fielders when measuring xDER. I'm quite certain of this.