Apologies all for my late start this year. I was super lucky to be on vacation for August and September, so now I'm getting back into things. Rest assumed I was following the college football season from afar, and am thrilled to be back behind the spreadsheets.
Let's refresh a little on how things work. I'm focused on college football now, which means we've got predictive rankings on one side, and games of the week on the other.
Predictive rankings are meant to look forward - the model would always favor one team over any lower rated team. This is a good way to rank teams, certainly a more transparent one than whatever the AP voters user, but it does lead to some interesting results.
Games of the week are the 10 most "watchable" games. Watchibility is a metric I created that rates games as watchable or not, based on how good the team playing are, and how close the game is likely to be. For the 10 most watchable games, I dive a little deeper into who's likely to win, what the implications are for that teams CFP, Bowl, or Conference title hopes.
Finally I have dashboards for every team, those dashboards tell you about a team's schedule, odds of winning each game, and gives a little snapshot of the season. This annotated dashboard can explain it all.
One of the core pieces of the model is a concept I've named Posterior Win Probability, a powerful way to adjust teams ratings as we get new information (e.g. watch their games) throughout the season.
OK let's look at the week ahead. I like weeks like this where the good games happen to not overlap. We get Colorado Oregon on Friday night, leading right into the red river showdown, then Sparty vs. Wisconsin and WSU @ ASU, finally into the peak night games.
Then we bring it home with UW @ Arizona at 8(?!???!!!!) PM Saturday night.