Since I missed out on last week’s Saber-Terms series, I figured I’d give out a double dose of explanations this week. Since I talked yesterday about Carl Pavano and his value to the 2004 Florida Marlins, I figured it would be nice to start a discussion on defense independent pitching statistics, more commonly referred to as DIPS. And so we begin another Saber-Terms chapter!
Why do we need DIPS?
Most casual fans start analyzing pitchers with win-loss record and ERA. I think everyone can attest to the awfulness of using W-L as a measure of pitcher performance. So much of W-L is dependent on teammates, both offensively and defensively, that it is almost a hopeless endeavor to attribute it all to a pitcher.
The same cannot necessarily be said about ERA. After all, runs allowed are runs allowed, and the statistic uses “earned” runs, which is supposed to distinguish fault between pitcher and defense. Well, never mind the fact that earned runs are awkward themselves, as shown in the linked piece by Maniac friend Jack Moore on his latest FanGraphs piece. Beyond that lies the fact that quality of defense is much more intertwined with pitching than simply in terms of errors.
Let’s look at a clear-cut Marlins example. Take Ricky Nolasco. Ricky has a tendency to give up his share of fly balls. Let’s say he gives up two fly balls to right field of identical type, from their trajectory off the bat to the bat speed and the expected landing area (for this example, let’s put the landing area as right on the chalk of the right field line in Landshark Stadium). Now, for the first fly ball, Jeremy Hermida is manning the right corner. He chugs his frame over there, only after initially misreading it. Despite his athletic looks, he isn’t much a burner, and the ball slices away from him as he haplessly waves a glove hand at it. It lands for a double.
For the other fly ball, we have the gifted Brett Carroll on duty. He reads the ball perfectly off the crack of the bat and begins his straight line to the ball. As the ball cuts towards the line, Carroll is already beginning to camp in the expected landing area. Instead of hitting the line, it lands safely into Brett’s glove for an out.
This is an extreme, but believable example. Obviously we spanned the entire spectrum of right field defense by using Hermida (among the worst) and Carroll (among the best). However, in either case, the batted ball was theoretically the same. How come in one case, Nolasco was charged with a double, while in the other he got tagged with an out? Because defense, beyond errors, matters.
DIPS theory in a nutshell
Without claiming any originality (sabermetrician Voros McCracken did the starting work on DIPS), here is how I would explain it.
Pitchers have little control over the results of their batted balls.
By “results,” I mean the result of single/double/triple/out. Of course, pitchers clearly have some control; after all, pitchers do have a fair bit of control over the type of batted ball, and that itself confers some range of likely results to a pitcher’s balls in play. But the responsibility of all aspects of batted balls is so intertwined between pitchers, hitters, and fielders that any control would be extremely difficult to discern.
McCracken initially determined that pitchers had essentially no control over batting average on balls in play (BABIP), pointing out that team defense had a major effect on a pitcher’s batted ball results. The discovery led to the development of DIPS ERA, one of the initial ways to separate defense from pitching. Nowadays, simpler methods are being used in the saber-mainstream.
From the initial attempts of DIPS ERA came many different ways to separate defense from pitching. Here I will detail a few.
Linear component run estimators
One method commonly used is to take defense independent components generally considered within the control of the pitcher and determine linear weights for each component. From these components, a run value for the pitcher’s contributions can be determined. Two such linear estimators are FIP (Fielding Independent Pitching), originated by Tom Tango, and tRA, originated by Graham MacAree. These both assign linear weights to strikeouts, walks, and home runs (tRA assigns both run and out values), while tRA adds batted ball types into the equation. These are then scaled up to runs per nine innings (FIP is in the ERA scale, while tRA is in the RA scale). Since this method is linear, it does not take into account context that a pitcher may put himself into, so it does not do as good a job of reflecting run scoring for pitchers as it does for hitters.
Using prorated team defensive runs
Sean “Rally” Smith uses this method for determining pitcher WAR. It is based on using team defensive runs saved, using a metric like TotalZone or UZR, prorated over the pitcher’s percentage of team balls in play allowed. This takes into account sequencing and context, but because it is prorated based on team level defense, it is not particularly accurate to the pitcher himself. A pitcher could have been more or less hurt by the specific defense behind him than a generic prorated team defense value would provide.
Pitcher Zone Runs
This method, originally discussed by MGL, measures a defensive statistic based on the balls in play allowed by a pitcher rather than the balls in play involving a specific player. In this case, it is like UZR for pitchers rather than defensive players. Defense behind a pitcher is measured, then the defensive run support is removed from a pitcher’s total runs allowed. This is, in my opinion, the best method currently out there for determining defense independent pitching production. However, MGL warns that this also invariably removes some BABIP skill, since correlation tests for pitchers switching teams has yielded some small correlation; theoretically, behind a different defense than the previous one, we would expect pitchers to show no correlation in defensive support.