Leaders from the Red Sox Player Development group and Professional, Amateur, and International Scouting staff will host a panel and answer questions from the audience.
Covering baseball is hard work, but someone needs to watch sports for a living. Our panel of media insiders will answer questions about covering baseball, the state of the game, and share insights about how analytics influence their work.
There has been much discussion in the past year or so about the surge in home runs in MLB starting about midway through the 2015 season. Much has been written, with the focus on three possible reasons for the surge: a "juiced ball"; reduced air drag; and a change in swing plane.
A juiced ball refers to one in which the so-called coefficient of restitution (COR, or bounciness) of the ball has been elevated, whether deliberately or accidentally. An elevated COR will result in higher exit speeds off the bat, leading to longer fly ball distances and consequently more home runs. Last year at this very event, I gave a talk showing evidence for higher exit speeds but casting some doubt on the juiced ball theory. Others have subsequently written disputing my skepticism. I will not discuss that topic again this year.
A new approach has been discussed focusing on the drag properties of the baseball. As the ball flies through the air, it collides with air molecules, slowing it down and reducing fly ball distances. The drag depends, among other things, on the size and surface roughness of the ball, the latter affected by the height of the seams. In particular, a smaller ball and one with lower seams will experience less air drag, resulting in longer distances. In recent articles, it has been suggested that this is responsible, in part, for the surge in home runs. In the first half of my talk, I will present a new analysis, never before shown in public, that directly addresses this issue. So, what has this analysis found? To find out, you'll have to come to the talk!
Finally, it has been suggested that batters are trying to hit more home runs by adjusting their swing plane to obtain simultaneously a high exit speed and a launch angle optimized for fly ball distance. In the second half of my talk, I will discuss the two aspects that the batter controls, the swing plane and ball-bat offset, and how these determine the "squareness" of the collision as well as the exit speed and launch angle. I will show how the optimum swing plane and offset involves compromises, which depend on what feature of the batted ball the batter is trying to optimize. Finally, I will show how Statcast data can be used to infer information about a batter's swing plane, allowing one to investigate directly whether the extent to which batters are "swinging for the fences", resulting in a surge in home runs.
We develop a pitcher similarity measure that is based on the speed and movement of every pitch. This tool can be used to address a range of application areas. The similarity measure allows the direct comparison of pitchers across various contexts including MLB, MiLB, amateur, and foreign leagues which can improve predictions for pitcher performance across environments. The identification of similar pitchers increases the sample sizes that can be used to forecast the outcome of batter/pitcher matchups and supports regression using more appropriate population statistics by projection models. The similarity measure can also be used to quantify the relationship between pitch distributions and pitcher performance. In addition, we can use the new tool to monitor pitchers over time and to develop improved models for the health risk and aging characteristics associated with different pitcher classes.
The Earth Mover's Distance (EMD) is used by the similarity measure to compare the pitch distributions thrown against right-handed and left-handed batters by each pitcher. A whitening transform is used by the EMD to account for the variances and correlation structure of the speed and movement variables. We demonstrate the similarity measure for several tasks including the identification of similar and dissimilar pitchers, the identification of unique pitchers, the quantification of year-to-year pitcher stability, and the quantification of pitcher variation with batter handedness and the count. We also show how the new measure can be used to improve pitcher projections. Properties of the similarity tool are visualized using non-metric multidimensional scaling.
Dr. Chris Geary, Chief of Sports Medicine at Tufts Medical Center, will review a variety of baseball-related issues in Sports Medicine that have occured during the 2017 season.
Zack Scott (Red Sox Vice President of Baseball Research & Development) joins Brian Bannister (Red Sox Assistant Pitching Coach, Director of Pitching Analytics, Saberseminar Junkie) for a Q&A / panel discussion about the use of analytics in building a major league team.
Organizations are beginning to think critically about the ways in which they develop, train, and coach young players. Are the strategies that teams use supported by sound theory and science, or are certain methods holdovers from "the past?" Fernando Perez will discuss his experience developing as a player, describe his vision for a player development system, and take questions from the audience.
Hitting instructors and pitching coaches see things other than physics when they look at super slow motion. However, there may be some overlap between the observations of a physicist and those of the experts.
Baseball pitching injuries are increasing at an alarming rate across multiple competition levels. An increased emphasis on enhancing throwing velocity has become popular in baseball. Weighted ball throwing programs are commonly implemented in an attempt to improve pitching velocity. While early research has shown efficacy in improving velocity, little is known about the safety of these programs on the shoulder and elbow. The purpose of this study is to examine the effects of training with weighted baseballs on pitching velocity, passive range of motion (ROM), muscle strength, and elbow torque. By analyzing these differences, it can be determined if a weighted ball routine improves pitching performance, as well as show the acute effects and stresses, if any, on the thrower’s arm.
Tom Tippett will give a brief presentation followed by a Q&A session.
The objective of the research will be to understand how teams can optimize the chances of inducing a double play ball.
In order to answer this question, several modeling activities will be needed and are sequentially related:
1) Model what type of batted ball is ideal for turning a double play. Some batted balls will not turn into double plays for a number of reasons (botched feed, bad throw to first base, bad hop, etc.), but in general we should be able to identify the characteristics of a batted ball that has a higher chance of turning into a double play. Candidate features include batted ball launch angle, exit velocity, horizontal spray (e.g., within some range at 3B, SS, 2B, 1B), and batter handedness at a minimum. The goal would be to develop a model that can be applied to any batted ball, resulting in a likelihood that the batted ball could turn into a double play.
2) After we have a way to identify high potential double play balls we need to model what kinds of pitches are more likely to lead to those types of batted balls. This could end up being 4 models — one for each type of platoon situation (RHP v LHH, RHP v RHH, etc.). Rather than make batter and pitcher handedness features of a single model it may be easier to have four separate models. In either case, the model may include the following features: pitch velocity, vertical and horizontal movement, spin rate, vertical and horizontal location, count, etc. Pitch type may also be included, but it may also be that the individual components that are used to define pitch type may work as well or better. The model may also include terms for individual batters and pitchers.
3) Once the modeling is done the question is what pitch, in what situation, will give the team in the field the highest chances of inducing a double play ball. This is the optimization step. A model/tool will be developed that will allow a user to input a pitcher and batter (and maybe count, etc.) and see what combination of attributes (pitch type, location, etc.) gives the highest chances of inducing a double play ball on contact.
With the release of Baseball Prospectus’ new pitch tunneling data, a variety of new questions can be asked about what makes up a pitcher’s success. In this presentation, we will look at tunnels and sequencing from their practical applications, specifically Dallas Keuchel’s 2015, 2016, and available 2017 data.
Additionally, in order to better explore the impact of pitch selection on tunnels, and those tunnels on a given pitcher’s results, we use Long's ""Pitch Calculator"", where it is possible to calculate the full range of ‘tunnel stats,’ including the average diff at tunnel. For example, if in 2016 Dallas Keuchel threw curveballs instead of sliders, this would changed his average diff at tunnel by .5%, which we can postulate would lead to slightly better results. This pitch calculator, along with our research on Keuchel’s pitch availability and selection will help demonstrate the value of pitch selection, pitch sequencing, and tunneling in a Major League situation.
There is an old adage among those who scout baseball players that “There’s no such thing as a relief prospect”. It is hard to argue this idea as bullpens around the game of baseball of full of failed former MLB starters (Andrew Miller, Wade Davis, and Zach Britton) as well as guys who never developed into elite starters in the minors despite their teams’ best efforts (Dellin Betances). A brief look at the top 100 prospect lists will rarely reveal a player who is listed as a reliever as teams will give elite arms every chance to develop into a starter. However, in recent seasons the building of elite bullpens and stockpiling power arms to close games has become a much more common occurrence with teams like the Cleveland Indians and Kansas City Royals riding elite bullpens through the playoffs. In their nature relievers are extremely unpredictable. They have wild fluctuations in performance form year to year save for a few of the best in the game and as a result it can be difficult to really project how they will perform in the upcoming season or even to find a guy who will succeed when used in a higher leverage role.
In this paper I use leverage index and RE24 to determine the per inning efficiency of different relievers adjusted for the situation context. There are very few metrics in the game that fully comprehend the impact a relief pitcher can have on the game as they rely heavily on the context of game situation they are put into and in some more traditional metrics like FIP and ERA misjudge some pitchers. For example using a FIP based WAR Zach Britton was not the most valuable reliever in the game. Using this per inning efficiency metric, which can be modified to fit into a WAR style scale to compare pitchers, I will then look at the underlying statistics and traits most significant for the success of relievers. Using these important traits I have developed a multiple linear regression equation to predict how a relief pitcher should have performed in a given season. The next step in using this predictive equation is to look at its predictive power. Every season there are a number of players who move into high leverage roles that they did not occupy the year before. Using their numbers from the season prior I will be able to predict how I would expect these players to perform in their high leverage roles. I am able to compare these expected values to the values achieved in the next season in order to show that by using this model we can significantly improve upon our ability to predict how well certain pitchers will perform in high leverage roles.
How do top writers and analysts dream up creative ideas for their next project? How do project timelines and deadlines influence this process? Three of the top writers and analysts in baseball share their intuitions about how they design stories. They will also take questions from the audience.
What began as a project to help the Sonoma Stompers find a woman backup catcher for the 2017 season became a project to find what about softball catching is transferrable to baseball—particularly framing. Jen Ramos, the former Assistant General Manager for the Sonoma Stompers, along with Shawn Brody and Ronnie Socash, both college students and baseball writers, began analyzing pitch framing data in softball and reaching out to NCAA Division 1 softball catchers to see what can and cannot transfer to professional baseball. After consulting with Christopher Long and utilizing his public github repository for NCAA softball, we believe that gender integration in baseball can begin with the transferability of pitch framing and catching methods.
We invited numerous softball catchers to Spring Training with the Sonoma Stompers, though schedules and prior commitments prevented them from joining for the 10 day spring. However, this did not prevent us from doing on-field work, as we worked with an elite softball catcher and framer at her alma mater, University of Florida.
We believe that, regardless of this individual project’s success, it is the start of tapping into an underrated market of catchers. We believe that this project is also a way to lay the groundwork for more gender integration in the sport of baseball, whether it be at the high school level, college level, or professional and affiliated levels. Though we might not find the first woman to play Minor League or Major League Baseball, our hope and goal is to examine the game and analytics in such a way that other analytics and scouting teams will build upon our research.
An obvious factor that baseball analysts and executives consider when assessing a player is his age. Players are evaluated based on their current ability and production, but teams also consider age to better predict to what extent players' skills might improve or decline in the future. In doing so, teams ascertain a stronger estimation of the present value of players. A specific example of the applicability of age is when teams want to hypothesize how much longer older players near the end of their careers can provide offensive value. The answer to this question is important for both signing an older player to a new contract and thinking about future team needs. One way in which we can group players in terms of aging is by position. We are specifically curious about the rate at which catchers age compared to players at other defensive positions. Catchers play their position by squatting down behind the plate, thereby exerting a physical strain on their lower bodies. Outfielders, as an alternative, play their positions by standing upright and doing much more running. By watching and studying baseball we see a difference in both the amount and the type of physical exertion between positions. As all players age, such player tools as speed certainly declines. Our research examines to what extent age impacts the production of players at the plate, with a particular focus on catchers. We have considered the following questions for our research: How do the offensive aging curves for different positions compare to those of catchers? Do some positions experience a more gradual drop off over time? Do some have a sharper decline? Which positions have similar aging curves?
Professional and amateur pitchers know very little about advanced statistics and trackman data. Though modern analytics have tremendous explanatory and scouting implications, they currently underserve the player. Drawing on professional baseball experience, this presentation will shed light onto how professional players view cause and effect, their comprehension of advanced metrics, how they explain good and bad performances, and how their abilities can be improved with an understanding of sabermetrics and Statcast data.
In addition to inadequacies in the average player’s understanding, the analyst also requires greater context to analyze data correctly. Many analysts render incorrect conclusions simply because they are not privy to the advanced thought and technique used by even low-level players. By increasing both the player and analysts’ level of understanding, we can unify the two groups, helping both to increase the power of their conclusions and job performance.
As professional sports teams strive to be proactive with the health and wellness of their players, the issue of mental wellness has come to the forefront in the last couple of years. Teams like the Chicago Cubs, Boston Red Sox, and Washington Nationals have started mental skills departments. Through these programs, players are encouraged to use both eastern and western techniques to improve their mental well-being. Further, as players such as Dontrelle Willis, Zach Greinke, and most recently David Fresse talk publically about their struggles with depression and anxiety, the stigma that you cannot be successful and suffer from a mental illness is being broken. With each step teams and players are realizing that many players and coaches are dealing with mental illness and that drugs and drinking, lack of sleep, and being "injury prone" are signals of a much larger problem. While Major League Baseball organizations are developing programs at every level, the sad truth is that most athletes will never sign with a professional team. While one hopes that MLB players speaking out about their mental illnesses will trickle down to remove the stigma at the high school and college level, the message will not be received by the athletes unless the infrastructure is in place. This project examines the awareness of college baseball and softball about mental illness, the signs, and the resources available to the athletes. This work is valuable as a survey commissioned by the NCAA in 2017 purportedly about mental health instead focused on the athletes use of drugs and alcohol. By conducting an independent study of college athletes at all levels of competition this project shows that the athletes are aware that mental illness is a real problem for other athletes, but that they downplay the importance of seeking help for themselves. This presentation will examine the (1) athlete’s conceptions of mental illness, (2) their willingness to help other players as well as themselves, (3) the perceived stigma of having a mental illness, and (4) their ability to recognize the warning signs of mental illness in themselves or a teammate. Finally, the presentation will examine what needs to be done for college players to act on the messages being sent being taken by major league teams and players to increase mental health awareness and remove the stigma of seeking help.
"For many years, private industry, public education, and the military have all used the discipline of instructional design to improve the training of knowledge, skills, and attitudes among learners of all types. By taking a systematic approach to the design of training interventions, it becomes possible for organizations to de-mystify player development and create measurable events that improve performance, and can be carried across an entire organization.
(While there is no substitute for the decades of experience that coaches and field staff bring to baseball, the role of an instructional designer is to make this expertise translate to observable successes and become repeatable over large learning groups.)
This presentation briefly covers the basic underpinnings behind the discipline of instructional design–including the systemic nature of learning, learner-focused outcomes, and measurable definitions of success–before providing examples of ways in which field staff and front offices alike can use these tools to improve ballplayer performance.
As a first semester freshman at the University of Illinois, Urbana-Champaign, I had no prior experience with baseball analytics. Evidently, neither did the university baseball program. As I became more interested in data analytics within baseball, I sought help from Dr. Alan Nathan who introduced me to the assistant coaches. We began a dialogue and I volunteered to send the coaches data-driven reports on teams before each series, highlighting players of interest and stand-out statistics. My first official project was to transcribe game-by-game pitching data and send the coaches short analyses about pitch selection, accuracy, average velocity, etc. This turned into my next two projects: developing a website as a hub for data including D1 splits and the pitching data, and designing and building an Android app to record the pitching data in real time and analyze it the day of the game. Over the summer, I will be working with incoming freshmen players on collecting their swing data via SwingTracker—hardware Dr. Nathan suggested to the baseball program. My aim is to integrate the SwingTracker data, along with the pitching data from the Android app, into the website so there is one centralized web-based location for all necessary baseball data.
My talk will outline the challenges of working with a program with no prior analytics experience, and ways students can get involved with their team. I am only two semesters into my project, but I am scheduled to study independently under Dr. Nathan next semester and travel with the team as Student Manager of Analytics in the 2018 season.
With no hard salary cap in baseball, teams are free to construct their rosters however they want. Deep-pocketed owners will compete each offseason for the best available players, while smaller-market teams will wait until the new year to scoop up the leftovers. Because of this, players are not always paid exclusively on their own ability, rather the circumstances that surround each individual case. The goal of my model is to value each free agent’s worth based solely on performance. In this age of data analytics taking over Major League front offices, teams are always looking for competitive advantages. With my model, teams can more quantitatively value free agents and know when their price tag exceeds the value they’ll bring. To do this, I compiled ten years of free agent data, and plotted performance versus annual salary. The performance metric I used was a combination of sabermetrics with a large emphasis on WAR. Also, I weighted the performance of each player for the three seasons prior to their free agency. After accounting for inflation, this regression gave me the model I use today, where I plug in position, age, and performance and get a salary figure. "