| FINALISTS

​

TEAM 97: Alex Frederick, Jack Kalsched, Pranav Rajaram, Andrew Zaletski (UC San Diego)
The Man in the Middle

Cutoff plays are a critical but under-analyzed component of public baseball analysis. In this project, we develop a framework to assess cutoff play decision-making using anonymized MiLB player and ball tracking data. First, we construct a logistic regression model to estimate baserunner safe probabilities for extra base advancement paths. Next, we combine these probabilities with RE24 run expectancies to compute expected run values for each possible cutoff action, identifying the optimal cutoff choice for every play. Using these labels, we train a Random Forest classifier with features such as runner and fielder distances, arm strength, and sprint speed to predict optimal actions. Our findings reveal that cutoff men overwhelmingly favor aggressive throws when our model recommends holding on to the ball, leading to notable run expectancy losses. Finally, we present an interactive dashboard that provides team- and player-level cutoff decision-making insights that can supplement scouting and coaching. This project offers a scalable approach to cutoff play evaluation that can be used by teams to improve their defensive strategy in crucial situations.
 

TEAM 158: Timothy Jackson Millar (University of Canberra, Canberra, Australia)
The Cognitive Clock Score (CCS) Framework: Measuring Player Cognition during the Pre-Acquisition Phase

Traditional defensive metrics in baseball excel at quantifying physical outcomes but fail to capture the subconscious, split-second mental processing that precedes them. This paper introduces the Cognitive Clock Score (CCS), a metric framework designed to objectively measure a player's cognitive performance during the pre-acquisition stage of a defensive play: the critical process from bat-on-ball contact to securing the ball in the glove. The data is displayed in a dynamic, lightweight web app available at https://cognitiveclock.com.
Leveraging SMT’s tracking data, the CCS deconstructs plays into five core cognitive components: Reaction Time (RT), Path Efficiency (PE), Movement Decisiveness (MD), Adaptability (AD), and Anticipation (AN). These metrics are calculated and contextualised via a data pipeline that normalises performance against league positional benchmarks. In addition to creating a 20-80 score for each factor, multi-layered Cognitive Archetypes are generated for  Players. This framework translates complex quantitative scores into intuitive narratives, identifying players with archetypes such as "Thinkers," "Reactors," or "Hesitant Processors" and detailing specific trade-offs such as "Explosive but Inefficient" or “Chaotic Creator”. The result is a powerful tool that complements traditional eye tests, providing an empirical foundation for player evaluation, targeted development, and strategic team construction.

TEAM 199: Ty Albao (UC San Diego)
It’s BRI, Bro: Evaluating BaseRunning Intelligence
 

In my analysis, I aim to obtain a way of evaluating a player’s and team’s baserunning choices in critical and fast-paced situations. I first get two probabilities. One is the breakeven probability of a runner advancing that tells the runner what the needed probability of being safe is to make it worth attempting to advance (calculated using a run expectancy matrix), and the other is the actual probability of a baserunner being safe on a given play (calculated using the model I built). I then compare these two probabilities to see if the runner made the right decision, to stay or advance. This adds another dimension for baserunning critique as it enables analysts to observe aggressiveness and good decision making, regardless if the runner was safe or not.
 

TEAM 208: Tom Kim (University of Toronto), Bennett Moore (University of Guelph)
The Intent to Give it Your All

Modern player evaluation concentrates on talent metrics such as maximum pitch velocity and sprint speed due to their predictive capacities. However, the expression of talent depends on deliberate athlete intent. Despite widespread acknowledgment, the concept of “hustle” remains relegated to conventional wisdom rather than quantitative analysis. In this project, we develop quantitative metrics, Hustle+ and Safe Probability Added (SPA), to objectively measure hustle and its influence in game outcomes. We define these metrics through statistical estimation of athletes’ true capacities and their effort-driven expression. True capacity is estimated using softened maximum-a-posteriori estimation combined with Z-score-based variance reduction, while the expression of this capacity is modeled based on established sport science literature. Gradient-boosting decision tree ensembles are employed to simulate outcomes. Results indicate a linear relationship between hustle and performance. Furthermore, significant discrepancies in hustle are identified among teams, with consequential impacts on seasonal outcomes.