Jim Poserina’s scientific contributions

This chapter applies the six models—game scores, team statistics, logistic probability, team ratings, logit spread, and logit points—to the 2014–15 National Basketball Association season. The models are evaluated for their accuracy in determining a favorite, the favorite’s probability of winning, and the margin of victory using in-sample data. Additional analyses examine how much out-sample data is required before the models can accurately predict the outcomes of future events.

This chapter provides an overview of the use of probability and statistics in sports modeling applications. The chapter includes an overview of the important mathematics required for probability and statistics modeling and a review of essential probability distribution functions required for model construction and parameter estimation. The chapter also includes an introduction to different sampling techniques that can be used to test the accuracy of sports prediction models and to correct for data limitation issues which are often present in sports modeling problems. This is primarily due to limited data observations and/or not having games across all pairs of teams.

Previous chapters looked at various sports using season-level data: win–loss records and statistical sums and averages for players and teams. This chapter introduces the reader to ways in which freely available baseball play-by-play data can be used to perform more intricate analyses. To illustrate this technique, run creation and win probability added are demonstrated. Individual batting events are traced to determine how often they result in a run being scored or an run batted in being credited, using a data set comprising 76 seasons and 10.8 million individual plays. Individual game situations (inning, outs, baserunners, score differential) are aggregated to determine how often a team in a given situation went on to win or lose; that data is then correlated with plays and players to determine which have added the most to their team's probability of winning.

In this chapter we provide an overview of mathematical probability models that can be used to rank teams, predict the winner of a game or match, and estimate the winning margin and the total number of points likely to be scored.

This chapter applies the six models—game scores, team statistics, logistic probability, team ratings, logit spread, and logit points—to the 2015 Major League Baseball season. The models are evaluated for their accuracy in determining a favorite, the favorite’s probability of winning, and the margin of victory using in-sample data. Additional analyses examine how much out-sample data is required before the models can accurately predict the outcomes of future events.

This chapter provides overview of six sports modeling techniques that will be applied to different sports including football, basketball, hockey, soccer, and baseball. The goal is to provide readers with a step-by-step set of instructions to be able to formulate the model, estimate the parameters, and predict winners and probability of winning.

This chapter applies the sports models from previous chapters to fantasy sports competitions. We show how these models can be used to predict the number of points that a player will score against an opponent and we show how these models can be used by gamers looking to draft or pick a fantasy sports team for a season or weekly contest. These models and approaches can be applied to any type of fantasy sports scoring system as well as used to compute the probability of achieving at least a specified number of points. These models can also be used by professional managers and/or coaches to help determine the best team to field against an opposing team. The five models applied to fantasy sports in this chapter include (1) game points model, (2) team statistics model, (3) logistic probability model, (4) logit spread model, and (5) logit points model.

This chapters applies the six models—game scores, team statistics, logistic probability, team ratings, logit spread, and logit points—to the 2015 Major League Soccer season. The models are evaluated for their accuracy in determining a favorite, the favorite’s probability of winning, and the margin of victory using in-sample data. Additional analyses examine how much out-sample data is required before the models can accurately predict the outcomes of future events.

This chapters applies the six models—game scores, team statistics, logistic probability, team ratings, logit spread, and logit points—to the 2015 National Football League season. The models are evaluated for their accuracy in determining a favorite, the favorite’s probability of winning, and the margin of victory using in-sample data. Additional analyses examine how much out-sample data is required before the models can accurately predict the outcomes of future events.

This chapters applies the six models—game scores, team statistics, logistic probability, team ratings, logit spread, and logit points—to the 2014–15 National Hockey League season. The models are evaluated for their accuracy in determining a favorite, the favorite’s probability of winning, and the margin of victory using in-sample data. Additional analyses examine how much out-sample data is required before the models can accurately predict the outcomes of future events.