People in the financial business often study things they call derivatives. I’ve mentioned it in passing here before, but I really didn’t delve too much into it. In the financial market, derivatives are pieces of the whole. You often cannot manipulate the entire whole, but you can change it if you know which parts to look for.
In this case, if a team wants to improve their fielding overall, they must find the players that are a drag on their fielding and replace them with better fielding. This is where things get dicey. In any good study there is a control group and an experimental group. The control group is something we all must agree on and then we compare the experimental group(s) to that control group. So, today we are looking at team fielding.
The control group for today is something called defense efficiency rating. It is simply the percentage of balls that are put in play (minus home runs) that get converted into outs. Some analysts don’t like DER because it is non-specific in terms of what types of balls (fly balls, line drives, groundballs) are included. This is true. Yet, it gives us an unfettered look at a team’s overall performance. Sure, there is luck and random distribution involved. However, those numbers correspond very well to pitching performance.
So, what we do from here is what I would call an ordinal study. Ordinal data is a simple look at rankings. This team is better than that team and so forth. Interval data would obviously be preferable, but I like to sleep, eat, and have quality family time. So, with ordinal data we are simply comparing team’s rankings in DER (the control group) with their rankings in the various defensive metrics commonly used. For our purposes, we are utilizing fielding percentage, fielding bible runs, baseball prospectus runs, fangraphs runs, and baseball-reference runs. Finally, we will also look at what happens when we average the four sabermetric models together.
Before we throw in the chart we should talk about our goals. The idea is that if a team wants to improve their fielding, they ultimately want to see their DER improve. However, DER cannot be manipulated directly. So, you want to find the metric that best correlates with DER and manipulate that on an individual basis. For instance, the Orioles can improve greatly by substituting just about any third baseman in the known universe for Mark Reynolds. Therefore, their DER would naturally improve even if there is no such individual DER ranking for players.
When you look at the bottom you see the strength of correlation in the difference column. The closer the number is to zero, the stronger the correlation. As we can see, the fielding bible data is clearly stronger than the other four alternatives. I would add that Baseball Information Systems is releasing it’s third volume of the Fielding Bible this winter. If you buy any book on fielding let that one be it.
Baseball Information Systems has replaced nearly all of the human factor and replaced it with computers for this volume. In previous volumes, they relied on play by play data to classify the type of ball (ground ball, fly ball, line drive) put in play. That classification used to depend on the call from the play by play announcer or official scorer. Now, cameras automatically classify it for them, so there will be a set standard in all 30 parks. This will only serve to make their data that much more accurate.
What this means is that when you want to improve your team’s DER, you know that either Fielding Bible data alone or a combination of the metrics will give you the best chance to accurately make the moves necessary to do so. Furthermore, since all four metrics are expressed in runs, it will give you an idea of how much bang you are getting for your buck. Even if you wanted to use fielding percentage, it wouldn’t tell you to what degree you would be helping or hurting your ball club.