Well, I’ve finished all my grading and will start my drive north on Thursday. So with a few extra days on my hands, and now that we’re 7 games into the bowl season, I thought it might be a good time to check in to see how well our favorite advanced stats have fared with their bowl game projections.

Although I have fun looking at these data (and maybe other people here will enjoy it too), I must confess that my primary motivation is that I’m hoping for some insights (even if just an inkling) into how well the SP+ and FEI predictions will hold up for the game against Alabama. You see, the advanced stats have Michigan as 8.4 point (SP+) and 10.5 point (FEI) favorites against ‘Bama and I’m wishing like hell that they are correct.

Before looking at the results to date, it’s important to keep two major caveats in mind. First, there have only been 7 games so far. Second, these predictions are based on each team’s performance throughout the season. However, the models do not know, and cannot account for, injuries or players that have opted out, say, to prepare for the NFL or because they’ve entered the transfer portal. In other words, the teams that are the basis for these models are, in some cases, not the same teams that will be playing the games.

If anyone is interested in the data for the SP+ or FEI predictions, they can be found here and here.

THE RESULTS

The graph shows the predicted winner and predicted margin of victory (blue dot) for SP+ and FEI, along with the actual margin of victory (red dot) for the predicted winner. Negative values for the actual margin of victory indicate incorrect predictions, while positive values indicate correct predictions. The distance between the predicted and actual margin of victory represents the projection error. 

So for example SP+ predicted ULL to beat Jackson State by 1 point. The blue dot represents the predicted margin of victory for ULL (+1). The red dot indicates the actual margin of victory for ULL (-3). The distance between the two dots (4 points) represents the prediction error. 

Just glancing at the graph, it doesn’t appear that either SP+ or FEI are particularly close in terms of predicting the actual margins of victory. If we just look at correctly predicting winners, FEI has a clear advantage with a 6-1 record compared to a 4-3 record for SP+. 

Personally, I’m a fan of graphs but I know that some people prefer to look at tables. And since I aim to please…

The table gives pretty much the same information as the graphs. But I also provide the average prediction error for SP+ and FEI. In general, there’s a pretty modest difference in the average prediction errors for both sets of advanced stats.  

So what can we say so far? 

• Well, the vast majority, 12/14, of the predictions so far have been coin flips (less than 3 points). And while FEI has been more accurate in predicting the winner, their prediction error has only been marginally smaller than SP+. I mean, sure, they accurately predicted Fresno State to beat NMSU. But they predicted a 1-point victory in a game that turned out to be a 27-point victory. Can we really say that’s more accurate than SP+’s prediction that NMSU would win (or Fresno State would lose) by .6 points? 
• I’ve got a problem with micronumerosity, and my doctor told me the only solution is more football. Luckily we still have 34 bowl games left to play. 

I'll try to keep these updated with one more post on the regular board before January 1. If anyone is interested to follow along at home, here’s the schedule and predictions for the games ahead.

 

 

 

P.S. I know I said in my last diary that I would take a break from making data posts. I guess I just can't help myself. 😊

Go Blue!