I received an email from Dennis of Mt. Shasta, CA, which alerted me that some of the split statistics were not being calculated correctly. Specifically, the vs. Winning/vs. Non-Winning splits were incorrect. There were a couple of issues that caused this.
First, the vs. Winning/vs. Non-Winning splits were only being updated for teams that played in a bowl game. I had a bug which caused those splits to not be updated for the other teams that did not play in bowl games. This has been corrected.
Second, there was a fundamental problem with how the vs. Winning/vs. Non-Winning splits were being calculated. When calculating the split for games between an FBS team and an FCS team, I was only considering the FCS team’s record against FBS teams rather than the FCS team’s overall record. For example, when calculating California’s split, I considered the Bears’ win against Eastern Washington (FCS) as a win over a non-winning team because Eastern Washington was 0-1 against FBS teams. In fact, Eastern Washington has an 8-4 overall record, so Cal’s win should have been considered a win over a winning team instead.
Unfortunately, I don’t track the statistics (including win-loss record) for FCS teams. Keeping up with just the FBS teams is enough to keep me busy. As a result, I don’t have enough information to properly calculate the vs. Winning/vs. Non-Winning splits when FBS teams play against FCS teams. My solution for now was to modify the definition of the vs. Winning/vs. Non-Winning split to only include games against FBS teams. On each page that includes that split, there is a notation indicating that the split only includes games against FBS teams.
I’m not really happy this solution. I don’t like how this split leaves out some games while the other splits include all games. I think the inconsistency might be confusing. However, I think it’s the best solution for now given the data that I have. Using the complete win-loss records for FCS teams is obviously the better solution, but the amount of work needed to do that may be significant.
Thanks to Dennis for letting me know about the problem.