College of Science and Liberal Arts Associate Dean for Undergraduate Studies and Mathematics Professor Dr. Bruce Bukiet is most well known not for his doctoral work in detonation theory, or his application of math to biological systems, but for a unique baseball predictions model that has seen him spotlighted by the New York Times and CNN. Starting in the late 1980s, Bukiet began to tinker with different models after reading an academic paper on modeling tennis outcomes.
“During my undergraduate years, I took a class called Operations Research,” said Bukiet, “and on an exam, we had a question about optimization in blackjack. I started wondering — could I apply this to baseball?” His current model, developed to fulfill this need, is more accurate than 80% of industry-standard predictors.
The concept of the model is based on the finite number of possibilities of baseball. “Baseball is limited to 25 situations,” he explained. “The batter could not make it to any base, or he could make it to first base, second base, third base, or a home run. And when he comes up to bat, there might be no one on any base, one person on any base, or [multiple people].”
Summed up, there are eight base-runner situations, along with three “out” conditions. In baseball, a batter gets three chances to hit a pitch before they are out; a player could come up to bat with the team having zero outs, one out, or two outs. Eight base-runner conditions multiplied by three out possibilities makes 24.
“We call the third-out situation the 25th,” remarked Bukiet. “In Markov processes, it’s called the absorbing state — essentially, it marks the end,” he added. The novelty of Bukiet’s model is that within any Markov process, the prelude to the situation does not make a difference. “You don’t care how the people on different bases get there, or that a team is only winning by one run — what matters is the transition probability, which is the probability that the next batter up will get a single, double, and so on.”
Based on each player’s past performance, Bukiet predicts how likely they are to get each hit, walk, or be out, as well as the probability of any person already on base moving on to the other bases. However, no matter what happens with the player — whether they hit a ball or not, and no matter how far along the diamond they get — the successive situation is always limited to one of the aforementioned 24 situations.
The model consists mainly of matrix multiplication as well as keeping track of the number of runs scored in each situation for each team — with every team playing 162 games in a season, and 2,430 taking place overall, there is a considerable amount of data to consider. “Once I get the runs distribution of each team,” said Bukiet, “I compare each team’s, and then decide who is more likely to win.”
Before modern computing, the strategy was primarily to take the overall average of the team, without considering each individual player. This was because with nine players at a time, the number of possibilities ballooned to 9!, or 362,880 ways just to order them, without even considering other nuances. However, by computing the runs distribution probabilities of any particular lineup, the model is able to determine the lineup that should produce the most possible runs.
“Using a ‘simple’ brute force approach, this would take millions of years,” commented Bukiet, “and using the Markov process perspective turns it into something that only takes 1.5 seconds. All the overlaps that would take huge numbers of operations become manageable all of a sudden, because it’s just 25 things that become another one of the 25.”
For this year, Bukiet predicts the outcomes listed below: the Atlanta Braves are expected to win the National League East, while the St. Louis Cardinals are expected to become the champions of the National League Central. The San Diego Padres and the LA Dodgers are slated to tie in the National League West. The New York Mets and the Philadelphia Phillies should also make the playoffs.
The model also predicts the Toronto Blue Jays to claim victory in the American League East. The Cleveland Guardians are expected to win in the American League Central, and the Houston Astros are favored to win the American League West. The New York Yankees, the Tampa Bay Rays and the Los Angeles Angels should also make the playoffs. To learn more about Bukiet’s work, visit egrandslam.com and keep track of the 2023 baseball season!
