One could imagine, if you really tried, that if there was a Fans’ Bill of Rights for baseball, the First Amendment would read something similar to this: The right of the fans to question the umpire’s judgment, eyesight, or sanity shall not be infringed. Similar to the acquired nuance and complexity attached to the actual First Amendment, this pretend Fans’ First Amendment has become more complex since the days when you simply believed the word of the beat reporter. Nowadays, we have freely available video sources, multiple camera angles, and high definition. For the unlikely fellow calling balls and strikes behind the plate, we have Pitchf/x.
What is Pitchf/x?
It is a tracking system that is installed in every Major League stadium, producing pitch data since 2006. With two devoted cameras, the system can measure pitch velocity, movement, spin, release point, and final location. Simply put, it can follow a ball through its entire path in three dimensions while noting spin and velocity. A pitch is defined by its movement, so this enables you to differentiate between different pitches. Equations can be utilized to designate a pitch as a two seam fastball, a curve, a knuckleball, or whatever. Of course, no system is perfect. Sometimes you may see sliders and curves blend. You may have four seam, two seam, and sinkers blend together. Remember, the formula does not know what grip or throwing motion is used. It simply knows why happens when the ball leaves the hand.
A great place to find this information on your own is Brooks Baseball (www.brooksbaseball.net).
How can we apply this grand technology?
Pitchf/x enables us to complain more accurately and determine whether we truly have anything to complain about. This season, we (Lou and myself) were both looking at umpire call accuracy and whether it was benefitting one team or another. My personal interest lies mainly in the Orioles having to face the Latin God of Framing, Jose Molina, whose pitch framing techniques were famously touted as being worth 50 runs saved by Baseball Prospectus’ Max Marchi last October. There was also some interest in whether our more wild pitchers (e.g., Chris Tillman, Jakes Arrieta) were getting squeezed. Lou’s interested piqued due to the frustration he felt when it seemed like the Orioles were getting the short end of the strikezone, while other teams such as the Yankees and Red Sox were getting far more calls to their benefit.
This season, we both (separately) decided to look more closely at the data. I had a simple system in place looking at missed calls and associating run expectancies to these while Lou constructed a much more complex database to more accurately look at whether a pitcher was truly miscalled. Over the course of the 2013 season, we will be using this collection of data to answer several questions, including but not limited to:
- Are the Orioles having more pitches called in their favor or against them over the course of a game, a series, and the season?
- What are the sum effects of incorrectly called pitches on the Orioles?
- Is there a significant difference in called borderline pitches for the Orioles and their opponents?
- Why are any differences in called pitches occurring?
- Which games and umpires had the most correctly called strikezone?
All data used for this analysis is courtesy of Brooks Baseball and the PITCHf/x tool. We will try to provide the data used in an uploaded new spreadsheet to Google Docs for review. Google Docs does not handle equations very well, so some of the methods may not be directly available through that feature. Below, we will go a bit more into how we will use the data. It might help if you bring up a table of the data to help understand.
For every pitch, PITCHf/x reports a horizontal and a vertical location. All vertical measurements are based in feet above the ground and are in the column ‘pz’. For example, a value of 1.89 in the ‘pz’ column is equivalent to a pitch that is 1.89’ or 22.68” above the ground as it crosses home plate. All horizontal measurements are based in feet to the left (negative numbers) or right (positive numbers) of the center of the plate, as seen by the umpire. As such, negative numbers would be inside to a RHB and outside to a LHB while positive numbers would be outside to a RHB and inside to a LHB.
For the purposes of this analysis, two different strikezones were defined. The Real Strikezone is defined by the following boundaries: bottom – 1.625’, top – variable, based on batter’s height and reported by PITCHf/x, but generally from about 3.2’ to 3.8’, sides – ±0.85’, which accounts for the width of the plate. This strikezone is what could be called the strikezone as suggested in the rule book. The second strikezone is based more on how umpires actually apply the strike zone. The Typical Strikezone is the zone that PITCHf/x shows is normally called for a LHB or RHB. The top of the zone remains the same as for the Rzone. The bottom drops to 1.5’. The most significant difference is the width, which changes based on the handedness of the batter. For RHB, the zone is ±1.1’. For a LHB, the zone is from -1.25’ to 0.85’.
I have also defined borderline pitches in two ways. The first is based upon the Rzone: bottom – 1.4’ to 1.85’, top – 0.15’ inside or 0.15’ outside of PITCHf/x’s reported top of the zone, left – -0.7’ to -1.05’, right – 0.7’ to 1.05’. This represents a pitch that is right on the edge of the rule book’s strikezone. The second definition is based upon the Tzone. By looking at pitches that are within the Tzone, but outside of the Rzone, we can get an idea of how “tight” or “loose” the umpire is calling the strikezone that night, as well as how adept a team’s pitchers are at pitching to the edges.
There are a few variables which this methodology cannot account for. While a significant difference in the called strike rate on borderline pitches may indicate that one team benefited more than another, no motive or intentional bias can be assigned to the umpire. Additionally, PITCHf/x does not track a catcher’s glove, which means that there is no way of telling by how much a pitcher missed his target. If the pitcher hits the corner, but missed his target by a foot, the umpire may be more inclined to call a ball. In the future, I hope to run analysis on what types of pitches (breaking balls, offspeed pitches, or fastballs) are more likely to be called one way or another.
Converting Pitch Accuracy to Runs Expected
Do you understand linear weights? Maybe not? Well, linear weights are perhaps one of the more dominant types of metrics that have been taking over MLB front offices for the past decade. This is not to say that scenario descriptive statistics are not important or that qualitative analysis is dead…this is simply a type of quantitative approach that has shown value, so teams use it. So, yes, many teams look at their own versions of WAR and what not. If you don’t, then you are behind the times or are being employed by the Phillies.
The basic genesis of linear weights came from early work by Ferdinand Cole Lane and some major tweaking from George Lindsay to devise run expectancy tables that we all know and rarely directed use these days (http://www.tangotiger.net/re24.html). The way you read those charts is by looking at which bases hold runners and how many outs are present (as well as how you wish to define your run environment…Is 2013 more like the Offensive Era or more like the previous Modern Era or the Raised Mount Era?). The resulting number informs you how many runs are typically scored in that scenario. It is an environment neutral assessment with no consideration of whose pitching, fielding, running, or hitting. So, it has contextual issues, but it does give a baseline to work from to determine what should be expected.
Similarly, we can use the same approach to determine the value of events as opposed to scenarios. This is wonderfully shown in this Beyond the Box Score post. From there, we see that a walk is worth 0.70 runs in 2010 with a strikeout being worth nothing. If you convert it to a pitch-by-pitch approach, then you wind up with each ball being worth 0.175 runs and each strike being worth 0 runs. However, there is a problem with that perspective because it assumes all counts are the same. We know this is not true. Performance can be vastly different based on the count. Pitchers with throw differently based on the number of balls and strikes on them and batters will be less inclined to let an umpire call a pitch if he is sitting on two strikes, so we need to take those into account.
Dan Turkenkopf did take those into account in this article on Stealing First. What we find is that counts are not distributed evenly and, along with performance difference, we can see that there are major differences that need to be taken into consideration. And…this is what we will do.
Tonight though, we will publish an article that will show the impact of umpires on play calling for the Orioles with the Pitchf/x portion fully constructed. We are still putting together the count by count formulas. Using weighted averages based on Turkenkopf’s work, we will be using 0.14 runs as a place holder until we can be more specific.