College Football 2020 (or some semblance of it) is underway and already there have been fireworks! How about the Sun Belt going 3-0 against the Little (Big) 12? If I were a betting man (and long-term readers know I’m not), I would expect to see more upsets than usual this year: practice isn’t the same, the delays, team illnesses and quarantines, opting out. I guess we’ll find out as the season continues, but keep in mind there won’t be many FCS/Group of 5 money games with Power 5s this year.
…..
And that brings me to the most difficult question. I’ve been asking myself about this for a while and I’ve had 4 different people ask me completely out of the blue in the last week or so: is the GCR going to be relevant this year? In every other season, I don’t publish until the 4th week or so because the odd upsets haven’t normalized yet and Tulane or some other team that upset a big school early shows up in the top 5. It takes a certain critical mass for the ratings to make sense. BTW, that’s true for any of the computer rankings I’ve seen. Some wait to publish, some publish with obviously premature rankings, and some label a certain preseason weight until critical mass is reached. Usually mid-to-late September is that point for the GCR. This year, it won’t be, so when do I start publishing? What do I do with the conferences that aren’t playing? How do I incorporate them, in a meaningful way, if/when they start their season? But the biggest question is how, with precious few non-conference games (and none of them really of note), do I compare say the ACC or Big 12 against the SEC (no non-conference games at all) or Big 10 (if they play)? If an American or Sun Belt team goes undefeated how does that compare to a Power 5 8-win team? In a regular season SOS takes care of that and there is enough cross-pollination to normalize that SOS. It’s what allows me to say one conference is better than another with confidence. You, oh great reader, do not have to agree – in fact, I love the debate – but I have a statistically sound point of view as a starting point. I don’t see how I’m going to have that confidence level this year. At this point, the Big 12 looks weak, the ACC looks like we expect it to. The Sun Belt surprised us by overshadowing Army (2-0 with big victories against ho-hum schools) or even UTSA winning on the road. We see a few FCS schools trying to get some games in. Air Force has 2 games on their schedule right now (Army and Navy). We see teams scheduling opponents for home and home series just to get 10-11 games.
…..
It’s just…odd. So, to answer the question: I don’t think we’ve learned a whole lot this season so far (not too surprising) and we certainly haven’t learned enough to know whether the GCR will be meaningful.
….
Here is my commitment. I started over to create the “new” schedule which I’m sure will need updating regularly, adjusting already for postponed or cancelled games. I will input each game and keep an eye on progress. Should it start making sense and if/when we get critical mass, I will publish – at least as much of the GCR functionality that makes sense.
…..
With that, I’d really like to hear your feedback, thoughts, questions, input, and/or complaints. That’s it for today – thank you as always for tuning in, G
I’m with you, Robert. I’m really struggling to see how you’re gong to have enough out-of-conference game data to have meaning. In any random set of data you need about 20 or more data points to make a meaningful conclusion. In your analysis, I don’t know where that 20 number would be plucked from. Is it the number of games that are played between two specific conferences? Probably not, because in a typical year, there wouldn’t be that many. Is it the number of out-of-conference games that are played by any one specific conference? Maybe, because all the out-of-conference games form a neural network when you include every conference’s out-of-conference games. You’re better at statistics than I am, Robert. What is the metric and what is the threshold to determine if the data is meaningful?
My gut prediction would be that the GCR is probably going to end up being meaningless even at the end of the season, because there still won’t be enough relevant data. However, you should still capture all the data. Maybe it will show something surprising.
Unfortunately, we will have to rely on the pundits for rankings this year. They will use their football knowledge/vision and historical biases to fill in the blanks, but maybe that’s the best we can do.
It’s disappointing for sure. I’ll still run the numbers but don’t feel comfortable publishing the rankings if it doesn’t make sense. Case in point: when I put in the current scheduled games, the SEC took most of the top 20 preseason SOS spots which makes sense because they are only playing conference games (no “patsies”). It’s not a true comparison so I can’t show it. I could publish pure rankings for each conference without comparing to other conferences, but not sure what that would show. Let’s see how the season plays out.