BIG BLUE FANS FOR
01 ANALYSIS OF COLLEGE BASKETBALL
When I started in the 90s, one of the first issues I had to confront was how to value home court advantage (HCA). Admittedly, I did not take a scientific approach but arbitrarily decided to use 6 points. While I realized that the HCA is not a constant for any team, much less from team to team, I adopted the simple view that it was a constant. I did experiment with other values from time to time, such as 5 and 7 points, but the 6 point HCA seemed to produce more reliable comparisons to actual outcomes than either 5 or 7 points, so I maintained that value for many years. Many fans who were aware of my use of a 6 point HCA for Rupp Arena expressed a view that 6 points was too low for the advantage that Rupp Arena provides. Nevertheless, I could not find any justification for exploring any variations beyond the 5 and 7 point values mentioned above. As a side note, these same fans also believed the 6 point value was too high when the Cats played away from Rupp Arena against most of their opponents.
THE POMEROY INFORMATION:
During the summer of 2017, Ken Pomeroy has published information about his views and measurements of the home court advantage. On August 5, 2017, he also tweeted two graphs tracking this factor over a span of many seasons (1999 through 2017). Generally, Pomeroy's data indicate the home court advantage is not nearly as large as most of us believe it is, and it has been shrinking.
The first graph shows the average HCA per Pomeroy for each of these seasons, and it is clear that this data shows a trend of a declining value of a home court.
I wish his data extended further back in time, ideally another 20 to 30 seasons to span college basketball over a time period to determine how high the HCA may have been in early seasons, and whether it peaked after some period of annual increases. Second, I am interested in how the future will play out, because I seriously doubt that this linear trend can continue forever because at some point, the home court would become either meaningless or even a liability, neither of which are feasible IMO. So, will this trend level out at some value, such as 2 points, or will this trend hit a local low and then begin to rise again. Have there been rule changes or changes in the way the game is played that explain changes in the HCA?
The second graph shows the home court winning percentage over the same span of seasons.
This appears to be consistent with the first graph because as the HCA decreases, the percentage of games won by the home team should decline, and it has. In 2017, the home court winning percentage dropped to a low of only 59.0%
All of Pomeroy's analysis is limited to conference play in these seasons because that provided the greatest uniformity of competition. That is why the value of home court winning percentage in the low 60s is not the same as the overall home court winning percentages reported by the schools which averages closer to 70% than 60%. However, I think to measure HCA, Pomeroy's decision to limit his analysis to conference games is a good decision. In today's major college basketball world, most major teams play most of their significant non-conference games at neutral locations, and load up on a host of cupcakes for their non-conference home schedule.
MY HCA ANALYSIS
During the 2016-17 season, I began to sharpen my focus on the HCA question. I examined the results of the 126 regular season SEC games. As Pomeroy indicated in his blog explanation, it seemed reasonable to focus only on conference games because of the consistency of opponents, and the even mix of home and away games played by each team. This analysis concluded that the average HCA in the SEC in 2016-17 was about 3.68 points.
Then Pomeroy reports in his aforementioned blog this summer that he has determined the average HCA for all conferences is about 3.75. Based on this analysis, I had already decided to modify my methodology for 2017-18 by shifting the HCA value from the 6 points I had been using to 3.68 points prior to Pomeroy's blog post and subsequent data. Furthermore, these findings give credence to prior fan observations that my historic use of a 6 point HCA value when the Cats are playing on the road may have been too large.
With Pomeroy's information, I have done additional research. First, I combined the data about home court advantage and home winning percentage from 1999 through 2017 and determined that there appears to be a relationship between home court winning percentage and the HCA value.
The logarythmic best fit to the Pomeroy data is shown on this graph of the data. I located data concerning the home winning percentage for all D1 teams for their current home venue. This data base included the results of over 140,000 college basketball games, and the average home court winning percentage from this vast database was 69%. When I applied the Pomeroy best fit relationship to those historic home winning percentage to calculate a theoretical HCA for each team, the result range of HCA was a high of 11.61 to -16.50 with a median value of 5.62 points and an average HCA of 5.82 points.
I then pulled out Kentucky and Kentucky's 2017-18 opponents from this database as shown below:
First, the logarythmic best fit relationship derived from Pomeroy's data has a very narrow range of values for both the home court winning percentage (59% to 62%) and HCA (2.5 to 4.5 points). The range of home court winning percentages for the full range of D1 teams is much wider, ranging from a high of 93.0% to a low of 21.7%. Therefore, I present these theoretical HCA values with a realization that the logarythmic best fit may not apply to such a broad range of home winning percent values. Nevertheless, I forge onward.
This season, I will track the effectiveness of predictions for UK games using the SEC HCA value of 3.68 points that I measured during the 2016-17 season. I will also track the effectiveness of predictions using these theoretical values for UK and its opponents, and at the end of the season, I will compare the two methods.
Submitted by Richard Cheeks
Submitted by Richard Cheeks