Baseball Win Predictor - Todd Wilson

For the last decade the second-hottest topic in Major League Baseball (behind PEDs) has been the ability of advanced statistics to measure player performance compared to more traditional methods or statistics. For an introduction to the debate I recommend the 2011 film, Moneyball. Last year I decided to see for myself whether advanced statistics were better at predicting a team’s win total in a 162-game season. I gathered 91 different team statistics for every team in every 162-game season between 1973 and 2012. These 91 statistics included both traditional (batting average, home runs, ERA) and advanced (WAR, OPS+, RAR) metrics. I then created a number of regression models based on different groupings of these statistics. I found that traditional statistics were just as predictive of team wins as advanced statistics. You can find the complete findings of my research here: http://www.dodgertodd.com/1/post/2013/09/baseball-statistics.html

The next step of my research was always to provide a way for a MLB franchise’s front office to evaluate how an individual player contributes to a team’s expected win total. I imagined a tool that would allow an executive to manipulate a team’s stat totals and show win predictions from each of the best performing regression models.

What I have provided, the Baseball Stat Win Predictor, is the first step towards a realization of that front office tool. The Win Predictor allows a user to import the statistical data used to generate the models, select a year and a team, and change the team’s statistics to see the effect on predicted win totals. The Win Predictor also allows a user to start with a blank predictor worksheet to enter values for seasons after 2012.

The Win Predictor in its current form is not my ideal front office solution. However it is an excellent way to become familiar with the ability of statistics to predict winning.