BC Football: Win Probability Statistics For Remainder of Season
One of our loyal readers, TxAggie_2011 loves his statistics and put together some graphs to show what the rest of the season could look like for the Eagles. He scoured the internet, and put together some of the most reliable sites to create these graphs.
The first graph looks at BC's win
From the TXaggie_2011:
Obviously the underlying data is the work (and property) of others who are far smarter than I. I’m just a statistics enthusiast taking published game win probabilities, or interpreting them from ratings data, and looking at how those probabilities impact the 4,096 different possible ways a 12 game season could play out.
- There are 3 graphs and 1 table summarizing the results. The first graph shows the probability of winning x total games in 2019. The second graph shows the probability of winning x ACC conference games in 2019. The third graph shows the weekly evolution of the total number of expected wins based on an even-weight average of the 4 systems. For example, before the week 1 games the data indicated we should expect BC to win 6.05 games this season and after the week 1 games were all played the data indicated we should expect BC to win 6.59 games (+.54). This factors in not only the game BC played and won but changes in the win probabilities of the rest of the games on the schedule. Lastly, the table summarizes the current data used and shows (1) the individual game win probabilities, (2) total expected wins for the season, (3) the likelihood of having 6 or more wins at the end of the season to be bowl eligible, and (4) the “interest factor” for each game (using the average win probability). The “interest factor” is based on a couple of minor changes to Paul Kislanko’s Ranked Based Interest Factor formula. The result of the formula is a number between 0 and 4.5. Paul uses the Massey Consensus Rankings, only considers FBS teams, and scales the formula result to 0-5 stars. I’m using the Sagarin Rating ranks, considering FBS and FCS teams, and scaling the result to 1-5 stars.
- : Jeff Sagarin doesn’t publish win probabilities for all games remaining in the season. However, the difference two teams’ Sagarin Ratings (plus his published home field advantage, if applicable) can be used to calculate an expected point spread for that match up. I have an archive of the weekly Sagarin Ratings back to the 2015 season and used that data to back-test the actual game results vs the predicted Sagarin point spread and created an function to convert that to a probability. That is the source of the game win probabilities sourced from the Sagarin Ratings.
- : ESPN publishes the individual game win probabilities for all scheduled games, so those are sourced directly from BC’s FPI page.
- : Ken Massey publishes the individual game win probabilities for all scheduled games, so those are sourced directly from BC’s Massey Ratings page.
- : Similarly to what I’ve done with the Sagarin data, I have interpreted a win probability from Bill Connelly’s SP+ Ratings (formerly SP+). This is his first year at ESPN and he’s made some tweaks to his system and how he publishes the data. He used to publish a Google Drive spreadsheet that had data for every team on separate tabs, including the individual win probabilities of all remaining games. Unfortunately, he’s no longer publishing the spreadsheet now that he is at ESPN. Similar to the Sagarin Ratings, the difference between two teams’ ratings (plus a home field advantage of 2.5 points, if applicable) can estimate the expected point spread. I created a function to convert that point spread into a win probability and was able to match (or get within 1%) of all the games he published in his week 1 projections. I plan on using the same function going forward and hopefully it will match (or almost match) the weekly projections he publishes. If it looks like it’s way off or he makes additional changes to his system, I’ll reevaluate.