Oakland A's Case Study Regression Model

Executive Summary.  Mark Nobel’s value to the Oakland A’s franchise is as a winning member of the Oakland A’s pitching staff and a major contributor to the most recent 1980’s winning season.  I do not assess that Mark Nobel has crossed into what would be considered a franchise player, and would recommend against compensating him for anything outside his on-the-field performance.  In this report, I will provide an example model for predicting ticket sales for an Oakland A’s home game, using the New York Yankees as the variable opponent.  Next, I will provide counter arguments, with analysis, to Mr. Nobel’s agents claims that ticket sales increased based on Mr. Nobel being the starting pitcher.  Finally, I will provide a recommendation on the type of contract the Oakland A’s should negotiate with Mr. Noble.  

Predicting Ticket Sales.  In order to understand the best predictors for Oakland A’s home game ticket sales, I built a model using the regression function in MS Excel.  The data and information I used to inform the model was based on input from Mr. Steward Roddey.  The initial thought process was that the day of the week a game is played and the opponent are two of the biggest factors in ticket sales.  In order to validate my hypothesis, I ran regressions with the dependent variable of ticket sales and independent variables that included all the predictors suggested by Mr. Roddey (see case study).   What I found was that the best predictor of Oakland A’s home ticket sales is a model using the following variables (See Figure 1).  

• Opponent
• Double Header
• Days of the Week (Monday – Sunday)
• Precipitation (Yes or No)
• Time of the Game (Day or Night)
• Promotions (Yes or No)

The model I created was based on predicting an Oakland A’s home game against the New York Yankees.  The model is significant with an F value less than .05, an R Squared value of 87% and an adjusted R Squared value of 84%.  The only variables with a P value above .05 are Promotions and Time of Game.  I left both variables in the model because I think they add value to the model; otherwise, I discarded variables with a  p value above .05.  For example, the time of game is important for an A’s versus Yankees Monday night game since a night game generally sells more tickets.  The Time of Game coefficient takes that into account by subtracting 1412 tickets when predicting ticket sales for a day game.  Since this model will only predict ticket sales for a Yankees vs. A’s home game, I suggest creating a separate model for each team the A’s play holding constant all the other variables (DH, Days of the Week, Precipitation, Time of Game, Promotions).  The reason why it is important to include the team the A’s are playing is because this is highly correlated to the number of tickets sold and without including the opponent in the prediction model, I do not think the model will provide as accurate a prediction.  I validated the correlation of ticket sales to opponents by running a separate regression (See Figure 2).  The model was significant and the P values of the opponents were all below .05, indicating they are all meaningful variables in the model.  Finally, I can only base the accuracy of the model on the data from the 1980 season.  Some of the Yankees tickets sales seem like outliers; however, to validate the model previous seasons’ ticket sales data should be included in additional models to validate the pattern of an increase in ticket sales when the Yankees visit Oakland.