Browsing by Subject "Play calling"
Now showing 1 - 1 of 1
- Results Per Page
- Sort Options
Item Statistically driven decision making in football through the use of reinforcement learning, random utility models, and parametric modeling(2022-12-01) Biro, Preston; Walker, Stephen G., 1945-; Calder, Catherine; Murray, Jared S; Fink, JoshDecision making under conditions of uncertainty is inherently a difficult task. The use of data can alleviate this difficulty by informing the decision maker of past results, but often raw data can still mislead if not properly put into context. Additionally, long-term optimal behavior does not always align with short-term needs, and thus even a well-tailored algorithm can provide undesirable results. The game of football provides a unique avenue for application due to the structure of the game and increasing levels of data availability. Through the use of reinforcement learning and random utility models, statistically optimal decisions can be identified under a variety of utility mindsets. Parametric models properly representing the data generating process can also provide insight on the underlying information. Using football play-by-play data from the NFL and college level (specifically for the Presbyterian College Fall 2021 team), a system of algorithms is designed to assist with the task of play calling. These algorithms rely on the uncovering of the underlying utility of states in the game, which themselves can provide additional information on how to increase efficiency.