In one of my previous posts, I gave a lengthy explanation of the relationship between the Vegas spread in football and the probability that the favored team would ultimately win the game. Utilizing spread data that I collected over about an 8 year period, and assuming that the actual margin of victory would adopt a Normal / Gaussian distribution centered around the spread. I found that it was fairly simple to calculate the spread vs. victory curve once you know (based on a lot of data) that the standard deviation of the deviation of the actual result from the spread is around 14-15 points.
I was also curious about whether a similar analysis could be made based on college basketball data, but this was something that I never looked into previously. Fortunately, I realized a few days ago that my go-to web sight for spread data (a site called "Prediction Tracker") has an mostly complete archive of spread data for all sorts of sports, including college basketball. I quickly downloaded some CVS files and within a couple of hours, I set up a simple database of college basketball spread data fro 2004 to 2017 including over 47,000 games. As one would expect, you can perform the exact same analysis and learn that in basketball, the key standard deviation value is almost exactly 10 points and this value appears to hold steady (unlike football) as the spread gets bigger. As a result, I was once again able to plot the spread vs. victory curve, which is shown here, along with the raw probably data from the database:
As one would expect, due to the tighter standard deviation the probability of victory rises quicker than it does in football such that a 7-point spread means a 75% chance of victory, the odds go over 90% at 13 points, and cross 99% at a spread of 23.5. In tabular form, the data looks like this:
I was also curious about whether a similar analysis could be made based on college basketball data, but this was something that I never looked into previously. Fortunately, I realized a few days ago that my go-to web sight for spread data (a site called "Prediction Tracker") has an mostly complete archive of spread data for all sorts of sports, including college basketball. I quickly downloaded some CVS files and within a couple of hours, I set up a simple database of college basketball spread data fro 2004 to 2017 including over 47,000 games. As one would expect, you can perform the exact same analysis and learn that in basketball, the key standard deviation value is almost exactly 10 points and this value appears to hold steady (unlike football) as the spread gets bigger. As a result, I was once again able to plot the spread vs. victory curve, which is shown here, along with the raw probably data from the database:
As one would expect, due to the tighter standard deviation the probability of victory rises quicker than it does in football such that a 7-point spread means a 75% chance of victory, the odds go over 90% at 13 points, and cross 99% at a spread of 23.5. In tabular form, the data looks like this:
Anyway, I thought it was interesting. Enjoy!
Comments
Post a Comment