Data Driven Decisions

With copious amounts of data being utilized in today's game, how can it actually be used to improve a team's performance on the field? In this article Sam Bornstein answers that exact question while adding in some effective implementation strategies he's seen used throughout his career.

“The trend of today’s game is to collect as much data as possible to make informed decisions. The difficult part about this is understanding exactly what that data means and then being able to translate it into something useful. This revolution in data driven baseball isn’t about changing how the game is played. It’s about using the available technology to increase your chances at success. In other words, we want to measure instead of guess.”

This quote came from Jake’s very first Simple Sabermetrics video over a year ago, and is still as prevalent as ever. Hello everyone, my name is Sam Bornstein and I’m very excited to be joining the blog as a contributor! I’m currently a student manager and the Lead Data Analyst for the University of Iowa Baseball program. At the same time I am working towards my Master’s of Science in Business Analytics. Like Jake and fellow contributor, Adam Schuck, I am a disciple of Desi Druschel and have worked under some first-class coaches that have allowed us to assist in the process of data-driven decisions.

The most recent baseball analytics revolution started in 2015 when Major League Baseball introduced Trackman at all thirty teams’ stadiums. Since then we have seen both the minor league system and several college programs follow suit by adding Trackman as their in-game pitch-tracking technology. Off the field we have also seen a rapid increase in the use of Rapsodo’s pitching unit, which collects information on pitch metrics such as velocity, spin, movement, and more. These pitch-tracking devices allow players and coaches to more precisely assess an athlete’s performance, while also giving us a vast amount of data to dive into. Those who are able to create systems for applying the data have begun to reap rich rewards across all levels of baseball.

While some may view the introduction of these technologies as bad for the game, that is certainly not the case. These technologies give us a quantifiable method to make data-driven decisions. Using technology to aid in the player development process is a lot like using a calculator on a math exam - without it you may be able to get to the right answer, but with it you can be sure you're making the best decisions possible to get to that answer quicker. This example is directly applicable to today’s game as well. The coaches who have been in the game forever have an immense amount of valuable experience. However, as more and more technology is introduced we are able to rely more on data to make our decisions than previous experiences.

Let’s consider a scenario involving the future development of a pitcher. Pitcher A’s arsenal includes a four-seam fastball, two-seam fastball, slider, and a changeup. Lately his pitching coach has noticed that the pitcher is running into two problems: the four-seam and two-seam appear to be moving similarly and the slider isn’t performing well.

That pitching coach is at a crossroads for what needs to be done. While experience is still extremely important in this game, the old-school way of relying on what’s worked in the past and the trial-and-error system is long gone due to the amount of information we have available to help us make optimal decisions. In modern day baseball, data-driven decisions are essential to the development of the athlete.

So, let’s ask a few questions about what’s next for the pitcher. Should one of the two fastballs be ditched, or is it worthwhile to adjust one of them? Does the slider need a different movement profile, or does it need to be thrown at a different velocity or spin rate? Both of these questions focus internally on fixing what is currently in the arsenal, but maybe it’s an external issue that is causing one or both of the problems.

The arsenal could be lacking a pitch that is different from the rest. Perhaps he should add a curveball, but should it replace the slider or simply just be an addition? Is this an issue of how he is sequencing his pitches, the rate at which he is using each pitch, where the pitcher is standing on the mound, or maybe something related to his mechanics or off-the-field strength training? None of these questions are effectively answered without allowing data to factor into the decision-making process.

Now, I just listed a bunch of specific questions that would take hundreds, maybe thousands of words to answer in-depth. Though that isn’t the main focus of this article, I would like to provide a few data-driven methods to attack these questions. It is assumed that one has access to unique sets of data that encompass a large population and if you are fortunate enough to have access to that, great! If not, there are publicly available datasets for those that are not affiliated with an organization, and I strongly recommend you check those out.

When it is time to redesign a pitch -

whether that’s its velocity, spin, movement, etc. - there are two ways to utilize data to help you make the correct development decisions. The first route is to sift through the pitcher’s past data and pinpoint the exact windows that the pitch performed well according to your desired key performance indicators. The second route is matching certain qualities in your pitcher’s pitches or entire arsenal to the rest of the population, with the goal of understanding how they succeed or how they fail.

After a period of collecting data on our pitchers last year, the Iowa analytics team analyzed the marginal relationships between the pitch characteristics and key performance indicators such as swing and miss percentage and ground ball percentage. Given a sufficient sample size against quality competition we can, for example, determine the potential gain of increasing the horizontal movement of a slider. If the pitcher induces more swing and misses when the pitch has 12” of horizontal break compared to 6” of horizontal break, we have an easy decision at our hands. If you do not have a sufficient sample size or the pitch has never performed well, then this likely isn’t the best route to go down. This is when you have to think outside the box and find comparable pitchers or pitches within the rest of your dataset.

Most pitchers throw some combination of a fastball, curveball, slider/cutter, and a changeup although there are several variations of each pitch type. Some factors that cause the diversification of pitch shapes include velocity, spin, movement, release point, grips, wrist pronation/supination, mechanical limitations, and more. In baseball terms, you’ll often hear labels such as a rising fastball, turbo sinker, 12-6 curveball, sweeping slider, etc. (I’d strongly suggest reading this article). Fortunately, these labels are quantifiable!

Each fastball is like a snowflake, but maybe we can compare snowflakes that look similar to each other. If we hypothetically consider that there are nine or ten types of fastballs out there (per pitcher handedness), we can start to analyze a pitcher against his competition. Let’s say that a Type 1 fastball is thrown between 90-92 miles per hour around 2200 revolutions per minute with 15” of vertical break and 10” of horizontal break, which should come from a pitcher throwing near a ¾ arm slot. If our pitcher in consideration throws this Type 1 fastball and there are ten pitchers in our population who throw the same type, we can use this information to our advantage in a couple ways.

First, this effectively multiplies the size of our sample by ten. This allows us to analyze our newly increased sample size as if it were Pitcher A’s data, though I would warn that this method is not free of risk. At the granular level, comparing fastballs to fastballs should only be done when considering the entire arsenal. Does Pitcher B have quality secondary offerings that compliment the heater? If so, that could be an underlying reason his Type 1 fastball is successful, among other reasons. Also, some pitchers have mechanical limitations that restrict them from matching characteristics in other pitchers. For example, those that have natural cut on their fastball often have difficulty making the necessary wrist and finger adjustments to increase spin efficiency. If that pitcher wants to move from one fastball type to another - one with higher spin efficiency - that option may not be entirely feasible. Though some pitchers are able to make those adjustments, it cannot be assumed that those abilities are transferable from player to player. As is the case with every other aspect of baseball research, it is best to dig as deep as possible to truly understand the scope of potential answers.

Second, we can study how other pitchers in the same group use the pitch for game-planning and pitch development purposes. Sometimes searching for the right pitch traits branches off into several other directions to develop a pitch or an entire arsenal. How often do they throw the pitch, what other pitches are in the arsenal, how do they sequence this pitch with the others, and what locations does the pitch succeed? If there is a common pattern among the ten pitchers then maybe that’s something to implement for your pitcher. On the other hand, if we find that the rest of the pitches in the comparison group do not perform well, then that is a good sign that it’s time to redesign the fastball into a different type.

Both of these approaches are your first steps in utilizing data to aid in your decision-making. Once you determine the optimal pitch characteristics, mechanical changes, or game strategy, the next step is to sit down with your trusted individuals and develop an actionable plan moving forward. Of course, these developmental changes are easier said than done. The focus of this article up to this point was to get you to recognize these issues with data, analyze why they have occurred, identify potential fixes to the problem, and pinpoint desirable traits. What goes on during that “war room” meeting to develop a plan to make these adjustments is a whole other article in itself, and I am sure that one of the writers for this blog will cover that at some point in the near future.

A common theme you may have noticed in this article is that data simplifies the decision-making process by minimizing the risk and increasing the reward. While that’s a bold statement, it holds true and is a driving factor in this new era of baseball technology. These methods sound a bit better than the eye-test, don’t they? Let’s skip the trial-and-error experiments and rely on facts and advanced information. Remember, this revolution isn’t about changing how the game is played, it’s about making informed decisions to increase your chances at success.

Data Supports Decisions.


Thanks for reading! Please feel free to reach out to me with questions. You can message me on Twitter, @Sam__Bornstein.

Special thanks to Jake for allowing me to contribute to the Simple Sabermetrics blog. I’m excited to continue moving the needle forward and give back to the baseball community. All the data that this technology collects is what attracted me to this sport and is what is pushing me to create a career out of it. This article is meant to be an introduction to my area of interest and what I will be contributing to this blog. In my future posts I plan on diving deeper into the process of collecting, analyzing, and interpreting data, attacking baseball problems from a technical standpoint, and sharing certain experiences I’ve had in my years in college baseball working for a data-driven program.

Back to blog