Strengthening Sabermetrics with Big Data: Finesse Pitchers vs. Flamethrowers
As the season changes from the verdant greens of summer to the saturated hues of autumn, baseball fans throughout the world gear up for the spectacle that is the MLB Fall Classic. From the seven-hundred and fifty of the world’s preeminent baseball players competing in the timeless and time-less game to be part of the final twenty-five, baseball has been a cornucopia of romantic modern mythology. However, akin to the erroneous notion of Golden Age Thinking, traditionalists have tried to deny the tide of progress by relying on outdated concepts.
From the 1980s, ‘Bill James and mathematics cut straight through’ the hazy traditional understandings of the sport. The old guards may still decry the permeation of sabermetrics—applying statistical analysis to baseball—but as with the rest of the world, technological advancements have modernized baseball with deeper and clearer comprehension. Accordingly, as the baseball industry has grown exponentially, the economics behind advanced analytics have followed suit. From PITCHf/x to Statcast, researchers have utilized technology to augment the multi-billion dollar industry.
One such traditional assumption about the intrinsic nature of baseball is that a pitcher with a varied assortment of pitches puts the opposing hitters at a disadvantage because of the uncertainty of pitch selections. However, research conducted by Professor Woo-Sung Jung from the Department of Industrial and Management Engineering and the Department of Physics cut straight through the conjecture and successfully demonstrated that a pitcher who mainly relies on a blistering fastball is more effective than a finesse pitcher who relies on an assortment of breaking balls. This achievement was published as a highlight paper in New Physics: Sae Mulli, the oldest Korean pure scientific journal.
Professor Jung introduced an indicator of pitching uncertainty through normalized mutual information (NMI) of not only pitch type but also pitch zone information that has been neglected in previous studies. He then calculated the pitching uncertainty of MLB starting pitchers from the last ten years and evaluated its correlation with an advanced sabermetric indicator of pitching performance called Fielding Independent Pitching (FIP). In contradiction to the industry-wide assumption, the research demonstrated that NMI and FIP are uncorrelated throughout the last ten seasons. In other words, flamethrowers with predictable repertoires are more effective than finesse pitchers.
This discovery has paradigm shifting potential as it eviscerates traditional surmise and introduces new avenues of analytics for the furtherance of baseball. Professor Jung expressed his excitement to employ more big data to add new conditions, such as innings, ball counts, and the catchers framing ability, and to extend the approach to other leagues to broaden the lucid understanding of the most romantic sport.