As everyone knows
, defense is the hardest thing to measure in baseball. I've been thinking about it, and I think the problem boils down to this: When the majority of major re-thinking was done on how to measure baseball performance (Bill James in the 80s, everyone else in the 90s), the tools at their disposal were not sufficient to properly analyze defensive performance. I believe that may be changing.
In the book Moneyball
, there is a passage on a consulting team (I'm gong by memory here, so I don't remember the chapter or the consultants) who described a particular way they had in measuring every offensive player's offensive ability. Rather than look at the hits, singles, etc., they broke down every at bat and measured what an average ball hit with the trajectory, speed, location, etc., would have done in a typical at bat. They then assigned a value of this kind of result independent of the actual result of the play
. This is critical, because it removed the random effects of superior or inferior defenders, wind, park factors, etc., etc. which can affect any individual at bat. By breaking down hitting into the simplest components possible, they were better able to quantify ability, and therefore identify bargains.
I believe that a similar thing can be done with regards to defense. The only problem lies in a lack of a database with the proper information. Great strides have been made in the area of defense with the development of Zone Rating
. However, ZR does not, in my opinion, take sufficient variables into effect to be as good as it could be. ZR does look at where a ball is hit, but not how hard it is hit.
If a ball is hit in Zone J
250 feet and results in a single, what does that say about the defender? Not enough to draw any conclusions. Was it a hard liner which the defender cut off, preventing a possible double? Was it a lazy fly ball the defender couldn't get to because he's too slow? ZR seems to consider that these things will even out over the course of a season. Basically, it ignores the question.
I believe that the simple act of timing the flight of the ball from the bat to the point of contact (either the glove or the ground) would give an easy way of measuring how hard it was hit. I'm no physicist, but the speed of the ball can easily be computed from this information. With a sufficient database for comparison, similar hits in similar areas at similar speeds can be measured for expected result. You can then compare the expected result against the actual results for a given player in that, or indeed, any situation.
For example, Aaron Rowand's play on the triple in last night's game
. If we had a database on all balls which hit that zone in the amount of time it did, we could easily tell how often, expressed as a percentage, such a ball results in an out, single, double, triple or home run. The compare that result expressed in terms of how it affects win percentage versus the actual result. The difference in win share can be cumulative, and fielding win shares can then be used to evaluate defensive performance.
After the basic question of catching, baserunning is easy. How often does a fielder hold a runner at first to one base on a ball hit to Zone B 220 feet which takes 1.2 seconds to hit the ground? How did this particular fielder do? What is the expected win share of the play versus the actual win share of the play?
In a similar way, infielders can be judged. Because everything happens so fast in the infield, I don't think that timing the "bat to contact" would work, so the groupings would have to be more general, with the location of the first contact with glove and ground measured, then the infield contact would have to be classified as hard or soft, liner, grounder, roller, or pop-up. Measure the difference between the expected win share of an average infielder on a similar play, and compare to the actual results.
First basemen can also be given a special separate win share for their abilities. If a ball is thrown in the dirt, what is the average result? How often does an average first baseman successfully turn that play into an out? Into a one-base error? A two-base error? The difference in the win share for those results can be measured and compared, too.
The reason no one has compiled a database with this analysis is no one has fully realized the value of it. In a similar way, in the past no one fully realized the value of tracking numbers like caught stealing, batting average in various pitch counts, and the value of a walk vs. a hit. But today all such things are measured and quite commonly accepted as valuable tools to understanding the game and the contributions of individual players. The data can be measured and compiled, and would bring fruitful insights to our understanding and enjoyment of the game.
If the statistical community were to make the effort to track all defensive plays in a way similar to the one I've outlined above, I believe it would revolutionize the way we see defense. At last there would be a way to compare the true value a player brings to the defensive table against his offensive contributions. It would be tangible, testable, and repeatable. In short, scientific.