November 07, 2007
Probabilistic Model of Range, Shortstops
A number of people are suggesting new ways to construct the models, but before I try those methods I'd like to present the model used last year for the nine fielding positions, starting with shortstops. I am including something new, however, the full team at the position.
Team Shortstop PMR, 2007, Visit Smooth Distance Model, 2007 data only
Team | In Play | Actual Outs | Predicted Outs | DER | Predicted DER | Ratio |
Rockies | 4599 | 657 | 602.67 | 0.143 | 0.131 | 109.01 |
Twins | 4384 | 556 | 523.57 | 0.127 | 0.119 | 106.19 |
Dodgers | 4310 | 556 | 526.50 | 0.129 | 0.122 | 105.60 |
Royals | 4528 | 543 | 514.33 | 0.120 | 0.114 | 105.57 |
Blue Jays | 4349 | 567 | 544.69 | 0.130 | 0.125 | 104.09 |
Phillies | 4505 | 531 | 516.45 | 0.118 | 0.115 | 102.82 |
Indians | 4548 | 571 | 558.76 | 0.126 | 0.123 | 102.19 |
Pirates | 4608 | 588 | 575.51 | 0.128 | 0.125 | 102.17 |
Red Sox | 4226 | 500 | 492.12 | 0.118 | 0.116 | 101.60 |
Giants | 4467 | 592 | 584.51 | 0.133 | 0.131 | 101.28 |
Diamondbacks | 4351 | 493 | 488.99 | 0.113 | 0.112 | 100.82 |
Brewers | 4392 | 501 | 497.76 | 0.114 | 0.113 | 100.65 |
Angels | 4325 | 502 | 498.77 | 0.116 | 0.115 | 100.65 |
Marlins | 4491 | 508 | 506.53 | 0.113 | 0.113 | 100.29 |
Mariners | 4535 | 515 | 514.50 | 0.114 | 0.113 | 100.10 |
Orioles | 4403 | 505 | 506.89 | 0.115 | 0.115 | 99.63 |
Astros | 4530 | 561 | 563.85 | 0.124 | 0.124 | 99.49 |
Braves | 4404 | 516 | 520.04 | 0.117 | 0.118 | 99.22 |
Cardinals | 4587 | 539 | 544.84 | 0.118 | 0.119 | 98.93 |
Reds | 4533 | 496 | 502.70 | 0.109 | 0.111 | 98.67 |
Athletics | 4499 | 531 | 538.40 | 0.118 | 0.120 | 98.62 |
Padres | 4476 | 536 | 544.49 | 0.120 | 0.122 | 98.44 |
Mets | 4362 | 506 | 518.72 | 0.116 | 0.119 | 97.55 |
Cubs | 4177 | 481 | 495.42 | 0.115 | 0.119 | 97.09 |
White Sox | 4545 | 563 | 580.23 | 0.124 | 0.128 | 97.03 |
Tigers | 4486 | 517 | 536.95 | 0.115 | 0.120 | 96.28 |
Rangers | 4518 | 531 | 556.38 | 0.118 | 0.123 | 95.44 |
Devil Rays | 4378 | 441 | 466.20 | 0.101 | 0.106 | 94.59 |
Nationals | 4591 | 532 | 566.26 | 0.116 | 0.123 | 93.95 |
Yankees | 4511 | 478 | 516.85 | 0.106 | 0.115 | 92.48 |
The above table will give you an idea of how the regular shortstop fit in the team context. You might imagine that Troy Tulowitzki was very good and Derek Jeter very bad:
Individual Shortstop PMR, 2007, Visit Smooth Distance Model, 2007 data only (1000 balls in play)
Player | In Play | Actual Outs | Predicted Outs | DER | Predicted DER | Ratio |
Troy Tulowitzki | 4294 | 615 | 564.54 | 0.143 | 0.131 | 108.94 |
Tony F Pena | 4010 | 480 | 449.44 | 0.120 | 0.112 | 106.80 |
Rafael Furcal | 3574 | 473 | 445.28 | 0.132 | 0.125 | 106.23 |
John McDonald | 2389 | 311 | 294.27 | 0.130 | 0.123 | 105.69 |
Jason Bartlett | 3631 | 466 | 443.58 | 0.128 | 0.122 | 105.05 |
Jimmy Rollins | 4447 | 528 | 511.62 | 0.119 | 0.115 | 103.20 |
Jack Wilson | 3657 | 470 | 457.15 | 0.129 | 0.125 | 102.81 |
Yunel Escobar | 1116 | 135 | 131.47 | 0.121 | 0.118 | 102.69 |
Jhonny Peralta | 4206 | 512 | 502.37 | 0.122 | 0.119 | 101.92 |
Omar Vizquel | 3739 | 504 | 497.76 | 0.135 | 0.133 | 101.25 |
Julio Lugo | 3592 | 431 | 426.14 | 0.120 | 0.119 | 101.14 |
Adam Everett | 1631 | 217 | 214.61 | 0.133 | 0.132 | 101.12 |
Orlando Cabrera | 3997 | 462 | 456.91 | 0.116 | 0.114 | 101.11 |
Alex Gonzalez | 2728 | 306 | 306.06 | 0.112 | 0.112 | 99.98 |
J.J. Hardy | 3873 | 442 | 442.35 | 0.114 | 0.114 | 99.92 |
Cesar Izturis | 1904 | 216 | 216.36 | 0.113 | 0.114 | 99.83 |
Bobby Crosby | 2524 | 313 | 313.77 | 0.124 | 0.124 | 99.75 |
Stephen Drew | 3877 | 434 | 435.25 | 0.112 | 0.112 | 99.71 |
Hanley Ramirez | 4054 | 460 | 462.96 | 0.113 | 0.114 | 99.36 |
Ryan Theriot | 2494 | 301 | 303.06 | 0.121 | 0.122 | 99.32 |
Khalil Greene | 4206 | 504 | 507.64 | 0.120 | 0.121 | 99.28 |
Mark Loretta | 1537 | 177 | 178.28 | 0.115 | 0.116 | 99.28 |
Yuniesky Betancourt | 4103 | 464 | 467.60 | 0.113 | 0.114 | 99.23 |
Edgar Renteria | 3067 | 361 | 365.13 | 0.118 | 0.119 | 98.87 |
Eric Bruntlett | 1075 | 131 | 132.81 | 0.122 | 0.124 | 98.63 |
Royce Clayton | 1538 | 200 | 202.77 | 0.130 | 0.132 | 98.63 |
Marco Scutaro | 1064 | 122 | 124.14 | 0.115 | 0.117 | 98.28 |
Juan Uribe | 4113 | 513 | 524.43 | 0.125 | 0.128 | 97.82 |
Jose Reyes | 4295 | 500 | 511.97 | 0.116 | 0.119 | 97.66 |
David Eckstein | 3002 | 349 | 357.57 | 0.116 | 0.119 | 97.60 |
Miguel Tejada | 3317 | 363 | 373.46 | 0.109 | 0.113 | 97.20 |
Jeff Keppinger | 1209 | 130 | 135.67 | 0.108 | 0.112 | 95.82 |
Carlos Guillen | 3361 | 389 | 408.05 | 0.116 | 0.121 | 95.33 |
Felipe Lopez | 2949 | 359 | 377.76 | 0.122 | 0.128 | 95.03 |
Michael Young | 4083 | 476 | 504.85 | 0.117 | 0.124 | 94.29 |
Josh Wilson | 1340 | 141 | 151.37 | 0.105 | 0.113 | 93.15 |
Brendan Harris | 2336 | 234 | 253.12 | 0.100 | 0.108 | 92.45 |
Derek Jeter | 4117 | 421 | 461.63 | 0.102 | 0.112 | 91.20 |
Cristian Guzman | 1189 | 117 | 130.96 | 0.098 | 0.110 | 89.34 |
Troy really blew the competition away in terms of PMR, and Tony Pena did his best to make up for his poor hitting. And while New York enjoys two fine offensive shortstops, neither exactly sparkles with the glove. You can also see why the Tigers are moving Carlos Guillen to first. Michael Young may not be far behind him.
The Reyes/Hanley Ramirez comparison is very interesting. Conventional wisdom on the web is that HanRam is a butcher with the glove and Reyes is pretty good; you find that HanRam is ok and Reyes is the butcher. It will be interesting to see what other systems (e.g. UZR) say about that.
Two other guys ranked surprisingly high: Jimmy Rollins and Jhonny Peralta. Interesting.
Rollins was highly regarded by The Fans (as was Tulo and MacDonald). No surprise on that front for me.
Is it entirely coincidental that only three SS had it easier than Tulowitzki, and only one had it harder than Guzman (and only three had it harder than Jeter)? I have some trouble believing that there isn't still a not-insignificant dependency between the degree of difficulty in BIP and these rankings - e.g. the harder it is to play the position, the less likely it is that a SS will do well.
Oh, and I very much like the "actual/expected" column at the end. Now we can say that a player will get 10% more outs than average, etc.
And since David shows both columns, one can easily do in his head actual minus expected.
Given the choice, this is the best one. Dividing in one's head is a bit harder.
David, do these ratings include flyouts caught by shortstops?
Mike, yes, they do include flyouts caught by shortstops.
There is a wide disparity between PMR's evaluation of Hanley Ramirez and Jose Reyes and the evaluations of the various zone rating methods. Rally's combined zone rating using both BIS and Stats data (http://home.comcast.net/~briankaat/statsite.html) has Ramirez at -15 and Reyes at +13. Fielding Bible (http://www.billjamesonline.net/fieldingbible/2007-plus-minus-leaders.asp) has Ramirez at -37 and Reyes at +13.
Part of the difference in the systems probably results from the inferior performance of opposing shortstops in Marlins' games compared with in Mets' games. Whether this was due to luck, park differences, offence differences or what, I do not know. If I were to guess, I would say that the Fielding Bible's rating of Ramirez is probably too harsh (Ramirez was -6 in 2006 according to the Fielding Bible, and I really doubt that he fell off the table in 2007).
I'd like to see the results in the other model (not using visiting players). I am guessing that they will be fairly close to Rally's.
So according to my calculations, Tulo was a +42.5 runs, if I use a .85 run value on the outs over expected. Am I doing this right?
"Part of the difference in the systems probably results from the inferior performance of opposing shortstops in Marlins' games compared with in Mets' games. Whether this was due to luck, park differences, offence differences or what, I do not know."
Mike: I can't find team BABIP for different ball types (anyone know where that's available?), but accourding to B-Ref FL hitters were .325 BABIP against GB pitchers, compared to .301 for NYM hitters. FL hitters were also 4 years younger than the Mets' on average (27 vs. 31), so likely they were faster running to first. So offense differences could be part of the answer.
I'm pretty surprised by Yuniesky Betancourt's rating. From watching him play I would have guessed he would rate much higher. Anything weird going on with his numbers?
David, thanks for the clarification. I believe Plus/Minus includes flyouts as well, but they use a ball-hogging adjustment (described below) that results in virtually zero _net_ flyout plays per infielder. (The highest persistent score in the 2003-05 Fielding Bible numbers was about +5 flyouts per season by Orlando Hudson at second.)
The ball hogging adjustment works as follows. If the infield/short outfield fly ball has a .95 probability of being caught by _somebody_, then the guy who catches it gets only +.05 credit. If the league average probabilities are .45 that the SS catches it, .30 that the second baseman catches it, and .20 that the CF catches it, and _nobody_ catches it, then each guy gets docked by the probability that his position would catch it. So everybody let the fly ball drop, then the SS gets docked -.45.
Guy,
That makes a lot of sense. I don't think of the Mets as a slow team (they did lead the league in stolen bases), but they probably are overall...BABIP home/road splits by batted ball type would definitely clarify the issue.
"So according to my calculations, Tulo was a +42.5 runs, if I use a .85 run value on the outs over expected. Am I doing this right?"
For shortstops, a .75 run value works better, as pretty much every play they make or save is a single.
Look for Chris Dial's work in the Dialed in blog on BTF for exact run values by position, or just remember these rules as they are close enough:
.75 for middle infield
.80 for corners
.85 for outfielders