Inside DERby Dave Studeman
January 26, 2006
David Pinto started publishing his fielding stats this week (here's a link to his second base rankings). David obviously doesn't have a background in marketing; he calls his stat the Probabilistic Model of Range (or PMR). But it's way cool anyway because...
1. It's based on batted-ball types, speed and detailed "vectors" (sort of like zones) for every batted ball, and
2. It's free, though David asks for (and deserves) donations for his great blog.
He's in the process of posting his rankings for each position, so be sure to check his blog regularly.
Most advanced fielding systems, like UZR and David Gassko's Range, compare a fielder to his peers that same year. But PMR compares each fielder against his peers over four years, which leads to some interesting conclusions. For instance, Pinto's findings suggest that fielders did a better job in 2005 than in 2004. Unfortunately, there appear to be a few data consistency issues, making that conclusion less than certain.
Still, I thought it might be fun to look at this phenomenon a little bit as it applies to individual pitchers. So I pulled the batted-ball results of all pitchers who faced at least 500 batters in each of the last three years, just to see what happened to their batted balls. The highest Defense Efficiency Ratio (DER—percent of balls successfully fielded for outs) I found was Al Leiter's in 2004 (.760) and the lowest was Sidney Ponson's in 2005 (.650). A difference of .180 is a wide swing, but a fairly typical one. Here's a link to all qualified pitchers in 2005, top to bottom, in DER.
Batted-ball types play a role here. Not surprisingly, pitchers who give up more line drives have lower DERs and those who have more infield flies have higher DERs. But the thing is, at least for this bunch of pitchers (2003-2005), batted-ball types weren't that important. The R-squared between batted-ball types and DER was only .10. That means that only 10% of the difference in pitcher DER could be explained by the types of batted balls they allowed. The other 90%? Well...
What really matters is what happens to a batted ball after it's hit, regardless of whether it's a ground ball, fly ball or line drive. And there can be some big differences between pitchers in a given year. Take the simple ground ball. In general, ground balls are fielded for outs about 75% of the time. But, as they say, "results vary by pitcher and year." Take a look at the pitchers who had the highest ground ball out percentages in the last three years:
Year Last First Team Grnd Out % 2003 Ortiz Russ Braves 83.5% 2004 Lima Jose Dodgers 82.6% 2003 Lewis Colby Rangers 82.0% 2005 Martinez Pedro Mets 81.9% 2003 Zambrano Victor Devil Ray 81.7% 2003 Day Steven Expos 81.4% 2005 Zito Barry Athletics 81.3% 2005 Carpenter Chris Cardinals 81.1% 2005 Pettitte Andy Astros 81.1% 2005 Carrasco Daniel Royals 80.8%
Russ Ortiz in 2003? Jose Lima in 2004? Now you know why those guys looked so good for such a short time. More of their ground balls were outs. Conversely, here's a list of pitchers whose ground balls were fielded for outs the fewest times in a given year:
Year Last First Team Grnd Out % 2004 Santos Victor Brewers 63.1% 2004 Prior Mark Cubs 65.8% 2003 Leiter Al Mets 66.5% 2005 Ponson Sidney Orioles 67.2% 2004 Bedard Erik Orioles 67.3% 2004 Acevedo Jose Reds 67.3% 2005 Capuano Christopher Brewers 68.6% 2005 Kazmir Scott Devil Ray 68.6% 2004 Lohse Kyle Twins 68.6% 2003 Dempster Ryan Reds 68.8%
There was a 20-point difference in groundball outs between Russ Ortiz in 2003 and Victor Santos in 2004. That's, how do you say, a lot. The number two guy from the bottom is Mark Prior in 2004, whose ground balls have consistently been fielded for outs at lower rates each of the last four years. The number three guy from the bottom is Al Leiter in 2003, one year prior to his .760 DER in 2004 (when his Grnd Out% was 77.3%, over 10 points higher).
Systems like PMR help us understand how much of the variance between pitchers is due to their fielders. There is also some evidence that a few pitchers legitimately impact the fieldability of their ground balls (like Prior or Jon Garland). But most variances are just random, as in Leiter's and Santos's cases. Take a look at Santos's batted-ball table (defined here), particularly the difference in net run value per ground ball between 2004 and 2005.
|Net Runs per Ball||% of Batted Balls||%/OF||%/PA||Total Net Runs|
As you can see, the run value of his ground balls improved (went down) nearly a 10th of a run in 2005, because more of them were fielded for outs.
Next I looked at how many "non-ground balls" were fielded for outs. I combined outfield flies and line drives because there are some questions about the consistent categorization of those (as you can see in some of the PMR comments). Let's call all outfield flies and line drives "airballs" just for fun. Here is a list of those pitchers whose air balls were fielded for outs most often (not including home runs in the denominator).
Year Last First Team Air Out % 2003 Franklin Ryan Mariners 72.7% 2004 Gobble Jimmy Royals 71.5% 2005 Elarton Scott Indians 70.6% 2004 Cabrera Daniel Orioles 70.4% 2005 Blanton Joseph Athletics 70.1% 2004 Santana Johan Twins 69.5% 2004 Milton Eric Phillies 69.5% 2003 Zito Barry Athletics 69.1% 2005 Lee Cliff Indians 69.1% 2005 Contreras Jose White Sox 68.9%
Ideally, I would park-adjust these figures because ballparks can have an impact on flyball out percentages. But this isn't really a "research" article, and I just don't feel like it. (Start your angry e-mails now!) I'd rather just move on to the list of pitchers whose outfield flies were caught least often. Check out Derek Lowe's 2004:
Year Last First Team Air Out % 2004 Lowe Derek Red Sox 44.9% 2003 Brown Kevin Dodgers 48.7% 2004 Webb Brandon Diamondbacks 49.0% 2003 Cook Aaron Rockies 49.2% 2003 Lewis Colby Rangers 50.9% 2005 Webb Brandon Diamondbacks 51.2% 2005 Carrasco Daniel Royals 51.2% 2003 Westbrook Jake Indians 51.5% 2003 Rusch Glendon Brewers 51.6% 2003 Day Steven Expos 52.4%
There is almost a 30-point swing from the highest to the lowest airball out percentages. In fact, fewer than half of Lowe's air balls were caught for outs! Now, some of this variance is definitely caused by giving up more line drives, so let's look at Lowe's line drives and outfield flies in his batted-ball table:
|Net Runs per Ball||% of Batted Balls||%/OF||%/PA||Total Net Runs|
Lowe's line drive rate was a low 17% in 2004, so line drives didn't cause his low out percentage. Basically, his outfield flies were caught for outs just 61% of the time. No matter how you cut it, Lowe was robbed by his fielders or the Green Monster OR he just gave up too many hard-hit balls or the balls landed in bad zones. I didn't watch him enough to know, but I'll bet a lot of Red Sox fans have their own opinions.
That's a little bit of "what's inside DER." I simply wanted to see how much differences in pitcher DER can be explained by batted-ball types. The answer is "not a lot"—you will usually need to examine the out rates of each type of batted ball for each pitcher to get a feel for his DER.
After I published the "How Much is a Batted Ball Worth" article in the Hardball Times Annual, someone pointed out to me that I could have used better run values. And when I saw someone else write in his blog that "Studes shows that a ground ball is a negative event," I realized that he was right.
The run values I used in the article (and in the pitcher tables included in this article) are relative run values (I call them net run values), by which I mean they are relative to the average plate appearance. So the average plate appearance is worth zero runs in this context, because all the other values are expressed as a difference from the average. In other words, a ground ball is worth less than an average plate appearance, but it's not a negative event. A team that hits all ground balls will still score a run or two.
To help clear up the confusion, let me post the following table with two sets of run values. The first is the set I published in the Annual, based on "net run values," where the overall average is zero. The second column is based on absolute runs scored in a game, where the overall average is around 4.6. As you can see, a ground ball does have a positive value in the second column.
Net Runs Runs Strikeouts -0.29 -0.11 Walks 0.30 0.30 HBP 0.34 0.34 Infield Fly -0.24 -0.07 Outfield Fly 0.03 0.17 Line Drive 0.36 0.40 Groundball -0.10 0.03 Bunts -0.10 0.04
I'm sure this is confusing, and I apologize about that. The basic difference between the two columns is the relative value given to an out. For a better understanding of what I'm talking about, you might want to read Tangotiger's excellent article How are Runs Really Created? (Part Two).
References and Resources
Actually, Tangotiger's three-part series on how runs are really created is one of the best and most important baseball stats articles of the last few years. If you haven't read (or even if you have), I encourage you to read it now:
David Cameron has an excellent overview of all the various defensive rating systems being published and quoted these days.
Our baseball stats are compiled by Baseball Info Solutions. If you're interested in acquiring great baseball stats, contact them.
Dave was called a "national treasure" by Rob Neyer. Seriously. Comments about this article can be sent to him through the miracle of e-mail.