Episode 20: April 15, 2011
There have been many attempts over the years to quantify defense. Despite claims that current stats describe 60% of fielding, it’s my opinion that little progress has been made. However, a new technology is coming which will soon change everything.
Please note: this is not an exact transcription of the episode.
In last week’s issue of Bloomberg Businessweek magazine there’s an article about baseball fielding statistics. I’ve read a lot of articles on fielding statistics and most of them are a lot of nonsense, but this one is different because this is the first time I’ve seen analysis of fielder positioning data gathered by cameras placed above the field. This is the beginning of “real” fielding stats.
I’m Alex Reisner...
Back in Episode #15 (Why Baseball Has Statistics) I alluded to the fact that I don't think we can have a full set of meaningful fielding statistics without more data on fielder positioning and movement. This week I'd like to talk about the existing fielding stats, where I think they fall short, and what fielding stats might look like in a few years when we have fielder positioning data from overhead cameras.
Let’s start with what’s wrong with the existing stats.
Errors are very subjective. Every year visiting teams are charged with more errors per inning than home teams, which reflects the bias of the official scorers. Did you know that? The Wikipedia article on Official Scorers says that the bias was significantly reduced after 1980 when writers were no longer allowed to be official scorers, but I’ve run my own set of numbers and the bias doesn’t seem to have changed much. The away teams still get charged with more errors than the home team 9 out of every 10 years.
Anyway, even without scorer bias there are fundamental problems with errors, the most obvious being that you can only get an error if you touch the ball. So a fast fielder who gets to a lot of balls has chances to make errors on plays that a slow fielder doesn’t. Bill James’ solution to this in 1976 was a stat called Range Factor, which is basically the number of balls fielded per inning. It’s actually putouts plus assists per inning but the idea is to quantify the player’s range. Range factor is easy to calculate from data which is readily-available going back to the 1800s, but it’s very position-specific, and it doesn’t really work for first basemen who get a lot of putouts on ground ball outs, or pitchers and catchers, who get an assist and a putout for every strikeout. It’s also somewhat skewed by the kind of pitchers on the team. Sid Fernandez, who pitched for the Mets in the late 80s and early 90s, was a fly ball and strikeout pitcher. So those Mets infielders probably had slightly lower range factors than they should have. Any fielder who plays for a team with multiple strikeout pitchers is somewhat penalized when it comes to range factor.
There are other stats you can get from box score data like Defensive Efficiency Record, which is the percentage of batted balls that are converted into outs, and Pete Palmer’s Fielding Runs and Fielding Wins, but these stats still don’t provide much detail about the plays.
So in the late 80s, organizations like STATS, Inc and Project Scoresheet started collecting detailed information on where batted balls were hit. Basically they divided the field up into “zones”. There was a chart they gave out to data collectors who sat in the press box and decided which zone each ball was hit to. Each fielder was responsible for a few zones and would get credit for an “opportunity” every time a ball was hit to one of their zones. If they fielded the ball they would get credit for fielding it, and stats like Defensive Average and Zone Rating were calculated just like batting average: the number of balls fielded divided by the number of opportunities.
Later John Dewan started a company called Baseball Info Solutions which collected similar data but used computers to do it in a much more detailed way. Instead of sitting in the press boxes, the data collectors watched video and entered the exact location of each ball, how hard they thought it was hit, the trajectory, where it bounced, etc.
This gave Dewan the data he needed to invent a new stat called plus/minus. It’s not a great name, but what it tells you is probably pretty useful. Basically it works like this: let’s say there’s a medium speed ground ball hit right up the middle. Dewan knows, from his data, what percentage of shortstops field that ball successfully. Let’s say it’s 50%. If a fielder makes the play he gets a +1, if he misses it he gets a -1. Those numbers aren’t accurate, this is just for the sake of explanation. So now, let’s say there’s a medium speed ball hit directly at the shortstop. Dewan knows that the ball is fielded successfully 99% of the time so if the shortstop fields it cleanly he might get +.03 or something small like that, but if he misses it he might get a -2. Again, those numbers aren’t right, but the point is that you get busted for missing an easy play and you get a lot of credit for making a hard play, and at the end of the season you have a number which is above zero, which means you’re better than the average player at your position, or it’s below zero, which means you’re below average.
Dewan first published these numbers in a book called The Fielding Bible. A few years later he published Volume 2 and translated his various defensive metrics into a number of runs saved. These books created a big stir because the stats made sense and they had some surprising results. The best shortstop turned out to be Adam Everett, who a lot of people hadn’t even heard of. Derek Jeter turned out to be one of the worst shortstops, which was surprising only to Yankee fans. Albert Pujols turned out to be an excellent fielder, and so did Jack Wilson and John McDonald. Michael Young and Nate McLouth didn’t look so good. A lot of Gold Glove selections started to look pretty shaky. People who believed in stats were really into this stuff and people who weren’t thought it was a bunch of crap.
Now here’s my problems with it.
First, I think the idea that you can make each player responsible for certain areas of the field is a little flawed. Players change position depending on the situation. What happens if the shortstop is playing to the right of second base because David Ortiz is batting? Defensive positioning can make certain parts of the field harder or easier to cover than they usually would be.
Second, whether the fielder makes a clean play to me seems separate from whether he got to the ball, but these stats are based on “successful fielding of the ball,” which combines the two things. What if a ball is hit right at the third baseman but takes a crazy bounce? Or what if it hits the base? Or if a runner nearly interferes with it? I think defenders of these stats would say that stuff happens rarely enough that it doesn’t matter, but to me it doesn’t negate the fact that getting to the ball is a different skill than catching it and should be measured separately. What about an outfielder who makes a lot of spectacular diving catches that would be routine plays if he was faster at reading the ball as it comes off the bat? Or if he took a better route to where the ball was going to land? He’s good at catching, and he probably looks like a great fielder, but you’d rather have someone who shows how easy those plays can be.
Third, not only do these stats conflate the separate skills of getting to the ball and catching it, they ignore throwing entirely. And there are a lot of aspects to throwing: it’s not just how hard or accurately a player throws, it’s how quickly they get rid of the ball, and perhaps most importantly whether they throw to the right base. Last year I saw Mets second baseman Luis Castillo, with a runner on first and one out, field what should have been an inning ending double play ground ball. It was hit pretty hard...right at him; he fielded it quickly and cleanly, and then decided to throw to first base. But then he realized he was supposed to throw to second so he turned around but it was too late to get the runner at second, and it was also too late to get the runner at first. Castillo cost the Mets two outs on a play most little leaguers would have made. If he had succeeded in throwing the runner out at first he wouldn’t have been penalized at all, by any existing defensive stats, because he fielded the ball and made an out. It doesn’t matter that it was the wrong out or that the Mets should have been out of the inning. The same goes for outfielders throwing to the wrong base or missing the cutoff man and allowing a baserunner to advance. These mental errors can be just as costly as physical errors.
There’s a John Dewan quote that is reprinted frequently where he claims that current defensive stats explain about 60% of fielding. But I don’t see how they’re even close to 60%. I’d say they’re around 25% at best. It’s not that the data is inaccurate or that the stats aren’t well designed, it’s that we still don’t have the data we really need and we’re still not looking at the things we need to be looking at. Right now I do still believe that a coach can evaluate a player’s defensive value better than the stats. And in the case of Luis Castillo, a third grader is probably sufficient.
Anyway, this is where the Bloomberg article comes in. Last year when I speculated about cameras mounted above the field to track player positions, little did I know it was already happening. A company called Sportvision had already installed cameras at AT&T Park in San Francisco, and they’ve added them at four more parks this year, and hope to have them in all parks by next year.
The data they collect should change everything, but it will take a while to figure out how to handle it. It’s a lot of information and there’s a lot that can be done with it. For example:
* We can look at fielder positioning, which probably says more about the coaches than the fielders but coaching stats would be pretty cool, right?
* We can look at fielder reaction time and also correctness. So: not only did the outfielder get a good jump on the ball, did he initially go in the right direction, or was he just moving because he knew the ball was going to be somewhere in the outfield? Also for infielders: did they know immediately that they were going to have to dive? And did they dive in the right direction at the right time?
* We can also look at foot speed when fielders have to take more than a few steps.
* And then there’s the separate skill of catching and holding onto the ball.
* If there’s a play to be made we also want to know about the quickness and sure-handedness of transferring and throwing the ball, arm strength and accuracy, and of course throwing to the right base or hitting the cutoff man.
Anyway, hopefully I’ve substantiated why I think the existing stats only quantify around 25% of fielding. There’s a lot to look forward to with this new data which should start to be availble next year. I don’t know whether MLB is going to make the data available for free or if it will be affordable or even available at all, but my guess is that we’ll start seeing some new stats pretty quickly.
This episode is already too long but I do want to talk about one more thing because it’s timely and it won’t take long.
I have a lot of respect for Joe Posnanski and on the big issues I also find that I rarely disagree with him. However: Manny Ramirez is retired and a lot of people are trying to sum up his career and decide whether he should be in the Hall of Fame, and in this, Posnanski wrote that, even though he admits the term "genius" is over-used, Manny Ramirez was a hitting genius. He said that Albert Pujols is the best hitter in the game, but there's no mystery to that, that Pujols isn’t a genius because he works harder than anyone else. He says Manny was HOF-caliber even though he seemed to not even care. I completely disagree with this. Manny is not a genius.
Now Joe admits that Manny must have worked hard, that he couldn’t have been so good without working really hard, but that he hid that part of himself from the world. So, what, if you don’t let people see you practicing then you’re a genius? I’m not saying I know what defines a genius but I think any reasonable definition has to have more to do with who you are than with what you choose to hide.
And anyway, to me, Manny was not that mysterious. He was just the Dennis Rodman of baseball. Even if you don't follow basketball you probably know Dennis Rodman from his various publicity stunts, but Dennis Rodman was one of the best rebounders of all time. Watching him rebound you could see how easy it was for him. He moved in this crazy way...he knew exactly where the ball was going and how to get there before anyone else. But watching him play offense was almost comical. He didn't really want the ball. His shot was awkward. He was so good at defense because he didn't care about offense, so he spent his time practicing defense, and this is what Manny Ramirez did. Don’t be fooled by him showing up late to spring training--during the season he was one of the first players at the ballpark every day, and he spent his time swinging the bat because that's what he liked. He was defiant and stubborn enough that he decided to play just one part of the game, and he had enough natural talent that teams let him do it. At least until they got tired of it.
So anyway, Manny Ramirez: genius? I don’t think so. Hall of Famer? Unfortunately, probably not since he apparently tested positive for performance-enhancing drugs twice at a time when nobody in their right mind would still be taking them. He does have the credentials, but the steroids may kill his chances. But that’s a discussion for another week.
You’ve been listening to Game of Chance...
All content on this web site and in podcasts copyright © 2010-19 Alex Reisner.