Poster Abstracts

(Presenting authors are with stars*)

The Causal Effect of a Timeout at Stopping an Opposing Run in the NBA

Connor Gibbs*, Ryan Elmore, and Bailey Fosdick
Colorado State University

n the summer of 2017, the NBA reduced the total number of timeouts, among other rule changes, to regulate the flow of the game. With these rule changes, it becomes increasingly important for coaches to effectively manage their timeouts. Understanding the utility of a timeout under various game scenarios, e.g. during an opposing team’s run, is of the utmost importance. There are two schools of thought when the opposition is on a run: (1) call a timeout and allow your team to rest and regroup, or (2) save a timeout and hope your team can make corrections on the fly. This presentation investigates the credence of these tenets using the Rubin causal model framework to quantify the causal effect of a timeout in the presence of an opposing team’s run and provides coaches with the needed analytic justification for better in-game decision making.

A Bayesian Joint Model of Mark and Intensity of Marked Spatial Point Processes with Application to Basketball Shot Chart

Jieying Jiao*, Guanyu Hu, and Jun Yan
University of Connecticut

The success rate of a basketball shot may be higher at locations in the court where a player makes more shots. In a marked spatial point process model, this means that the marks are dependent on the intensity of the process. We develop a Bayesian joint model of the mark and the intensity of marked spatial point processes, where the intensity is incorporated in the model of the mark as a covariate. Further, we allow variable selection through the spike-slab prior. Inferences are developed with a Markov chain Monte Carlo algorithm to sample from the posterior distribution. Two Bayesian model comparison criteria, the modified Deviance Information Criterion and the modified Logarithm of the Pseudo-Marginal Likelihood, are developed to assess the fitness of different joint models. The empirical performance of the proposed methods are examined in extensive simulation studies. We apply the proposed methodology to the 2017–2018 regular season shot data of four professional basketball players in the NBA to analyze the spatial structure of shot selection and field goal percentage. The results suggest that the field goal percentages of these players are significantly positively dependent on their shot intensities, and that different players have different predictors for their field goal percentages.

Catching Trout

Natalie Maurice, Danielle Sebring*, Jacquelyn Valenzuela, and Ronald Yurko
California State University, Fullerton

In baseball analytics, researchers are interested in identifying the top players in the league. In today's game, Mike Trout is arguably the best player in Major League Baseball. In our research, we first assess Mike Trout's skills and determine his strengths. We then identify one set of players with similar strengths using box score data and another using statcast data. Finally, we combine these datasets and create a final model to distinguish which players are truly "Trout-like".

Determining Player Valuation Through Team Impact

Proby Shandilya
Saratoga High School, California

Every year, the typical NBA player salary increases. With the continuous growth of the NBA salary cap and the flurry of max contracts being given to basketball players, it is hard to evaluate if a player is deserving of the compensation which they are receiving. However, a team’s outcome can be heavily reliant upon their allocation of salary and how much they pay players: overpaying players does not always translate to on-court success. This research aims to utilize regression analysis of statistics indicative of team impact in order to determine a player’s value for a certain team. The underlying theme behind this research is the assertion that one player is not worth the same financial value for every team; each team has different needs and a player’s value to the team should be representative of how well they fit the team and their needs. Take the Los Angeles Lakers and the New York Knicks, for example. If a good shooter who is a high-impact player on the scoring side of the ball is on the free agent market, his value would likely be higher for the Knicks than the Lakers (because LA has 3 great scorers while New York has 0). This project looks to find an accurate way to measure player value relative to the impact they can make on their team.

Musings in NFL Analytics

Brian Bader

How did the NFL’s extra point rule change affect point after touchdown success? The math says teams should always go for two. Recently, teams have been going for two more often, but should that trend continue? How hard is it to model the distribution of NFL scores? I discuss why direct parametric models fail, as well as the pros and cons of alternative approaches.

Predicting the Outcomes of Soccer Games

Jack Schooley* and Jun Yan
University of Connecticut

The outcomes of soccer games can be massively important to soccer fans and bettors alike. To predict the outcomes of these games, we consider the English Premier League (EPL) dataset from Kaggle, which has data from the 2006/2007 through the 2017/2018 seasons. The main predictive model used was an ordinal logistic regression model, and its predictive ability was compared against simple logistic regression models for each of the three possible outcomes of soccer matches. Variable selection for all models was done using separate lasso fits, which selected the features for each model using shrinkage. Finally, the resulting probabilities generated from the ordinal logistic model were compared against the implicit probabilities of the published Bet365 odds for each game prior to kickoff. The results of this comparison can help people make more informed bets.

Bayesian hierarchical models applied to fantasy games

Lucas da Cunha Godoy* and Marcos Oliveira Prates University of Connecticut

Sports fantasy games are becoming increasingly popular. Currently, these games, created in the first half of the twentieth century in the context of baseball, have boomed with the popularization of the internet. In Brazil, Cartola FC is the most famous game of its kind. Launched in 2005, today it has more than 5 million registered users.'s famous product has attracted many sponsors, making it a very profitable product. The game is based on the 1st division of Brasileirão, and all users start with 100 units of play money called cartoleta. The main idea is for users to build a team composed of real players from different real teams. Every player has a price, so the user has a limitation to build his/her team. The price of these players varies according to the performance at each league game. The performance of each team is measured based on the scouts of the players, i.e. goal kicks, defenses, right passes and so on. Each player's action adds or subtracts a quantity to its final score. The score of each team is composed by the sum of the scores of its players. We collected such data for all players throughout the 2018 championship. We propose a Bayesian hierarchical model to predict the player score using INLA. The main goal, is to create the best team based on model predictive score and constraint to the amount of cartoleta available at each round. Such approach is general and not specific for Cartola, and, therefore, it can be easily extendend to any other fantasy games.