DIPS

Analyzing pitching is complicated for many reasons, but one of the most important issues is that we judge pitchers on their ability to prevent runs, yet preventing runs is a team effort among the pitcher and his defenders. It’s ultimately a team game, but when you want to evaluate individual players you want to isolate their individual contributions.

For many years, the primary method of judging pitchers was based on the number of earned runs they allowed. While this seems like a sensible strategy on its face — subtracting out runs that occurred due to errors — team defense is still a huge factor in ERA. Bad defense isn’t just about errors, it’s about not getting to the ball in the first place as well. Pitchers play generally in front of one defensive unit all year so pitchers with good defenders will allow fewer runs than a pitcher of the same quality who throws in front of bad defenders.

So far, this is all very straightforward. We want to measure pitchers based on their contributions, not the contributions of the other players on the field. Actually doing so is quite difficult, but we can approach the problem in a number of ways. The overarching idea however is that we want a class of measurements that are independent of defense, or as we sometimes say Defense Independent Pitching Statistics (DIPS).

Over the years, DIPS as an idea has been confusingly blended with specific DIPS-based metrics like Fielding Independent Pitching (FIP). There are some parts of pitching which are independent of one’s defense and some that are dependent on one’s defense. This is undeniable. Certain metrics attempt to measure pitching using only those events which are independent of defense, but a metric is built in the DIPS tradition if it attempts to separate out what pitchers can control in some fashion.

DIPS theory took off about 15-20 years ago when research by Voros McCracken was published demonstrating how little control pitchers have over outcomes on balls in play. In more typical parlance, pitchers have limited control over their BABIP-allowed. As a result of this finding, a family of metrics was created that estimated a pitcher’s performance based on those defense-independent outcomes (walks, hit batters, strikeouts, home runs). The most famous of these is FIP, but you probably have also seen Expected Fielding Independent Pitching (xFIP) quite often, which replaces home runs with fly ball rate.

Because FIP and xFIP (along with tERA and SIERA) were the most popular first generation DIPS metrics, it became standard to equate the DIPS concept with these particular implementations. In other words, it is common to suggest that DIPS theory states that pitchers can’t control their BABIP at all. This is obvious not true, but these things tend to get oversimplified as they gain popularity and some of the nuance was lost. FIP treats all balls in play as equal, but DIPS doesn’t require that. DIPS simply requires that you do something to try to strip out the effect of one’s defense. 

With the passage of time, we’ve gained new data and improved the tools we use to analyze that data. This means that while in 2001, we didn’t really have a good way to determine how much control a pitcher was exercising over their BABIP, we have a better idea how to sort that out today. In 2001, we realized that it was better to assign zero responsibility than 100% responsibility, but we didn’t have the tools to get closer to the objective truth. Today, we have better tools. You may have read research in recent years that claims to “prove DIPS wrong” by showing pitchers control their BABIP to some degree. In reality, what that research shows is that we now have the data to construct better metrics than the ones that use the 0% assumption. We knew 0% was wrong in 2001, we just knew it was better than 100%.

DIPS is about building pitching statistics that isolate pitcher performance. The first generation metrics that used DIPS often utilized the assumption that pitchers couldn’t control outcomes on balls in play at all, but newer metrics, such as Deserved Run Average (DRA), have started to leverage more advanced processes that attack the problem in a different way. But the concept holds across the board that we are aiming for metrics that strip out the quality of the defense. That’s what DIPS is all about. DIPS is not a theory about pitcher value that states pitchers can’t control balls in play. DIPS is a conceptual approach to pitcher evaluation that states we should attempt to measure pitchers independent of their teammates.

-Neil Weinberg

Links for Further Reading:

Voros McCracken’s Groundbreaking DIPS Theory – Baseball Prospectus

How Can we Tell if a Pitcher is Any Good? – The Book Blog

Why DIPS Does What It Does – Beyond the Boxscore

Confessions of a DIPS Apostate – Hardball Times