November 14, 2006
Probabilistic Model of Range, 2006
Update: Please ignore this post. See here.
It's time to start the yearly look at defense using the Probabilistic Model of Range. My explanation of this model is here. I've settled on the smoothed visiting model this season. As always, let's start off with the teams.
Probabilistic Model of Range, 2006. Model Includes Parks, Smoothed Visiting Team Fielding
Team | InPlay | Actual Outs | Predicted Outs | DER | Predicted DER | Difference |
Blue Jays | 4326 | 2994 | 2914.85 | 0.692 | 0.674 | 0.01830 |
Royals | 4618 | 3120 | 3050.93 | 0.676 | 0.661 | 0.01496 |
Yankees | 4472 | 3103 | 3040.92 | 0.694 | 0.680 | 0.01388 |
Mets | 4310 | 3028 | 2976.80 | 0.703 | 0.691 | 0.01188 |
Cardinals | 4448 | 3096 | 3044.72 | 0.696 | 0.685 | 0.01153 |
Mariners | 4431 | 3054 | 3003.92 | 0.689 | 0.678 | 0.01130 |
Tigers | 4439 | 3112 | 3062.36 | 0.701 | 0.690 | 0.01118 |
Padres | 4386 | 3116 | 3073.11 | 0.710 | 0.701 | 0.00978 |
Giants | 4422 | 3098 | 3055.20 | 0.701 | 0.691 | 0.00968 |
Astros | 4342 | 3039 | 3002.64 | 0.700 | 0.692 | 0.00837 |
Cubs | 4152 | 2903 | 2874.47 | 0.699 | 0.692 | 0.00687 |
Red Sox | 4463 | 3028 | 3000.30 | 0.678 | 0.672 | 0.00621 |
Brewers | 4300 | 2950 | 2923.48 | 0.686 | 0.680 | 0.00617 |
Twins | 4328 | 2967 | 2943.16 | 0.686 | 0.680 | 0.00551 |
Angels | 4301 | 2970 | 2947.46 | 0.691 | 0.685 | 0.00524 |
Diamondbacks | 4462 | 3049 | 3025.84 | 0.683 | 0.678 | 0.00519 |
White Sox | 4528 | 3138 | 3116.50 | 0.693 | 0.688 | 0.00475 |
Braves | 4490 | 3078 | 3061.18 | 0.686 | 0.682 | 0.00375 |
Rockies | 4590 | 3129 | 3119.94 | 0.682 | 0.680 | 0.00197 |
Rangers | 4542 | 3084 | 3084.42 | 0.679 | 0.679 | -0.00009 |
Phillies | 4438 | 3021 | 3026.91 | 0.681 | 0.682 | -0.00133 |
Orioles | 4435 | 3013 | 3022.25 | 0.679 | 0.681 | -0.00209 |
Devil Rays | 4545 | 3048 | 3058.68 | 0.671 | 0.673 | -0.00235 |
Athletics | 4530 | 3120 | 3134.39 | 0.689 | 0.692 | -0.00318 |
Dodgers | 4536 | 3084 | 3099.77 | 0.680 | 0.683 | -0.00348 |
Marlins | 4339 | 2971 | 2988.96 | 0.685 | 0.689 | -0.00414 |
Indians | 4594 | 3099 | 3119.50 | 0.675 | 0.679 | -0.00446 |
Reds | 4527 | 3081 | 3107.77 | 0.681 | 0.686 | -0.00591 |
Pirates | 4448 | 2997 | 3037.04 | 0.674 | 0.683 | -0.00900 |
Nationals | 4594 | 3173 | 3220.53 | 0.691 | 0.701 | -0.01035 |
I've been told that Ricciardi isn't overly concerned with defense, but the Blue Jays led the pack in 2006, turning 90 more balls into outs than expected. And that with the loss of Hudson and Troy Glaus at third base.
It looks like having Johnny Damon and Melky Cabrera in the outfield helped. And despite A-Rod error prone ways at third, the Yankees did a very good job of turning balls into outs in 2006. And although the Royals defense came in for much criticism this year, it looks like it was much more the fault of the pitching staff. They Royas fielders did a fine job with what they were given.
A big surprise on the downside was the Cleveland Indians. The Tribe ranked near the top in 2005 as a team, but fell to near the bottom in 2006 with pretty much the same team. They gave up nearly three games worth of outs more than expected.
You can also see just how much the Red Sox defense helped them. Only Kansas City's DER was predicted to be lower.
The top three National League teams each won a division. The Dodgers and Athletics each made the post-season giving away more outs that expected.
There will be a lot more to come in the following days and weeks.
Update: I neglected to mention that the model is based on the 2002-2006 seasons.
I don't mean to question the integrity of your results, but some of the team rankings just don't smell right.
For instance, were the A's really second worst fielding teams in the AL this year? That doesn't sound right, and from what I saw watching nearly every game, it's not something I would agree with.
But of course, eyes are easy to trick. Good work, in any case.
David -
What year(s) are you using as the model for predicted? If I remember correctly in the past you used a couple seasons, not just the current. Is that the case this year?
Something I'd been noting all season is the impact (IMO) of Jack Wilson coming to camp with "20 extra pounds" this year. By every metric, their team defense is down this year, this being no exception. They're at -0.025 from last year, which isn't quite as dramatic as the Indians' -0.031, but still pretty hefty. And they did shake things up a bit more, personnel-wise, but the middle infield remained the same.
Billfer, I used 2002-2006 to build the model.
Pretty interesting, David. The Dewan Plus/Minus Team Results are going to be published in the Hardball Times Annual, and the results don't track nearly as well as I thought they would (R squared of only .37).
The two models have radically different answers for the Yankees and Red Sox (over 100 outs), for instance. It's fascinating (and disconcerting) that two relatively similar models, based on the exact same data, could come up with very different sets of answers.
The simple team model we use at THT, based only on batted ball types, is a bit closer to Dewan's projections, but not a lot. In the extreme cases, like the Sox and Yanks, it's in between the two extremes of PMR and Plus/Minus.
Any thoughts?
David,
I strongly recommend that you throw out 2002 in building your model. I understand the desire to maximize your sample size, but as you know 2002 was the first year that BIS collected any data, and the numbers from that year just aren't as high-quality as subsequent seasons. I threw out 2002 data in the batted ball work I did for the THT Annual, and I would recommend you do the same.
That said, I am very excited to see your results as you post them.
Christmas came early this year.
Out of curiosity, what time frame does BIS use for their norms? Perhaps the difference in time frames accounts for the differences between BIS' +/- and David's PMR.
So am I to understand that based on the Red Sox personnel, they were expected to be the second to worst defensive team in the league? and that they were actually only the 5th worst team in turning predicted outs into actual outs? Can this be split easily between infield and outfield? by position?
Tim, no, that's not the right interpretation. Based on the balls in play, the Red Sox should have turned fewer balls into outs than they actually did. In other words, the balls in play against them were difficult to field. This might be the fault of the pitchers, and to a certain extent the ballpark. Fenway, with the Green Monster, lowers the ability of balls in play to be turned into outs. The Red Sox defense played very well.
I believe Dewan's system only uses one year's data for the baseline. They looked at using multiple years, but reportedly didn't find that it made much of a difference.
Also, they don't adjust for ballpark (which would account for some of the Red Sox difference, anyway). But they do weight their plus/minus system by "degree of difficulty" of the batted ball, which I don't believe David does (at least not at this level).