Our first case study was adapted from the INFORMS Transactions on Education article “Case–Converting Point Spreads into Probabilities: A Case Study for Teaching Business Analytics” by Eric Huggins, Matt Bailey and Ivan Guardiola. I will post their case study for you to review, but your requirements are listed here. Please submit your Excel spreadsheet with the answers to each of the questions. Suspense: Thursday, 17 September at the beginning of class (5 pm).1. Using the spreadsheet (NFL_Scores.xlsx) provided with data from the NFL 2013-16 seasons:a. Determine for each point spread what proportion of games are won by the favorite and what proportion are won by the underdog. This will be most easily by using a Pivot Table.b. In Super Bowl LIII, the Patriots were 2.5-point favorites over the Rams. What was the probability that the Patriots won? What was the probability that the Rams won?c. What is the probability that a team favored by a touchdown will win? What about a two-touchdown favorite?d. Create a scatter plot of the point spreads versus proportions using Excel. Fit a linear regression line to the data. Interpret the equation of the regression line and explain how satisfied you are with your model? Where does the model work well and where does it seem to fit poorly?e. Create a second scatter plot looking at only the data points that had both successful and unsuccessful predictions (Truncate the data of all the 100% entries). Fit a linear regression line to your new data set. Interpret the equation of the regression line and explain how satisfied you are with your model? 2. Your analysis in part one consisted of 34 point spreads and their associated probabilities from your model. In reality, there were 1,086 data points that were not evenly distributed—for example, the high point spreads of 26.5, 19.5, and 17 were rare, each occurring only once over four years, whereas 3-point games occurred 131 times! Repeat part d with all 1,086 data points, thus weighting the more common points spreads more heavily. (You will most likely want to use a VLOOKUP)3. Raw data is rarely pretty and requires some work to convert into a usable format. The four years of data in the Excel I provided were precleaned and formatted for your analysis. The raw data is available from www.goldsheet.com and will require some data cleansing to make the data match the data format in your spreadsheet. Copy the data from the “Historic Logs and Ratings” for the 2017 NFL season from the website, paste it into Excel, and clean and format it so that it looks exactly like the data in your worksheets for 2013–2016.a. Determine for each point spread what proportion of games are won by the favorite and what proportion are won by the underdog. Again, his will be most easily by using a Pivot Table.b. Create a scatter plot of the point spreads versus proportions using Excel. Fit a linear regression line to the data. Interpret the equation of the regression line. How does your linear model compare with your results from your first model?ReferencesEric Huggins, Matt Bailey, Ivan Guardiola (2020) Case–Converting Point Spreads into Probabilities: A Case Study for Teaching Business Analytics. INFORMS Transactions on Education
0 comments