First, I’d like to give an obligatory hat tip to the San Francisco Giants for winning the World Series against the Texas Rangers, 4-1. Despite my inner-feelings to not root for you (due to my allegiance to the A’s), that was one of the best pitching performances of post-season history, probably since the 2001 Arizona Diamondbacks. Despite losing, Texas has a lot to be proud of. They continued to play their type of baseball day in and day out.
The subject for tonight’s post is a metric not many casual baseball fans know of: batting average on balls put in play (or from hereon, BABIP). It essentially answers the question, out of all the balls a player hits that are field-able by the defense, what percentage of balls will fall for a hit? Note, this is different from a regular batting average, which includes strikeouts and home runs.
Baseball statisticians love this metric because, for obvious reasons, pitchers are not always in control of the amount of hits they allow in a game. There’s just too many factors that can affect the outcome of a hit: Hard line drives are caught by diving center fielders, a bloop single can fall between defenders, ground balls can barely get past the glove of an infielder. When these ‘are you serious?’-hits are allowed, we kind of assume tough luck has graced the pitcher. And when we see excellent defensive plays, we think the pitcher is lucky and fortunate to have player X in the outfield. How many times have you seen this happen in baseball games? Too often.
So how can we judge a pitcher’s value, ultimately, without factoring in these good and bad luck tendencies? After all, the goal is to analyze a player’s contribution to winning games. Well, many journalists (and ultimately, fantasy players forecasting player performance) like to point to a player’s BABIP to see how luck has played its part in their season so far. For pitchers, a low BABIP is deemed lucky, and unsustainable (random s/o 2 Linda) whereas a high BABIP is viewed as ‘tough luck’. The opposite is thought to be true for hitters. If you read any literature on BABIP, they will tell you a pitcher’s BABIP usually rests at 0.300. Anything above or below this number is thought to regress toward this mean over the course of the season.
But why should this assumption to be true? As any person should do, know your facts and assumptions before diving into any analytical work. Thus, I wanted to check the assumption of why people thought .300 was a good number to select. For reference, I used the past 9 year’s of data (2002-2010), where all starting pitchers with more than 150 innings pitched were included. The total sample size was 779 pitcher. Below is a histogram, with its density plot embedded of their BABIP:
Evidently, the assumption fits the past 9 years of data. Even more surprisingly, it fits a bell curve shape very well (aka the normal distribution)! From this sample, we also see a standard deviation of about 0.02 points. Thus historically over the past 9 years, we can make a 95% confidence interval of 0.259 to 0.336.
But it still leaves me questioning the usage of BABIP. Pitchers have different skill sets that they try to bring to the field. Some pitchers are able to induce ground balls more often, while others are better at getting harmless fly balls. What if this affects what their seasonal BABIP is? Are there other metrics that can indicate what a pitcher’s true BABIP should be? From this post, I think you now know which project topic I chose. Hopefully I’ll be able to answer some of these questions in my research over the next few weeks.