Changes are coming...
More accurate and more intuitive Elo ratings are here
Good morning and happy Thursday. What a weekend. From 9 am until well past midnight Saturday, we had uninterrupted college football coverage. There was a closer-than-the-experts-think Alabama-Florida game, a less-points-than-we’re-used-to Clemson performance against Georgia Tech, a do-you-believe-us-now Fresno St. dub over UCLA, a whiteout W for Penn State, and we got Saquon Barkley playing beer pong. What more could you ask for?
Let’s start the day off right with some CFB Stats talk.
This is also probably the perfect time to bring up some exciting updates to our Elo rating system. Up until this point, Florida would lose Elo points after losing to Alabama, despite putting up a great fight and leaving most people with a better impression of their talent than they had before the game. This seems wrong. Well, luckily we have Drew Bennison on our team, who came up with an innovative idea.
Before, a team would improve their Elo ranking anytime they won, with the caveat that they would improve less for games they were highly likely to win, and more for large upsets according to Elo’s win probabilities. Now, a team can actually lose points in a win if they play poorly, based on their post-game win probability, which is essentially looking at each team’s advanced stats and predicting how often team’s that end up with those stats win or lose the game.
Easier to just show you, let’s look at how teams in the Top 25 were affected by this new change this weekend.
So, Alabama lost points and Florida gained points by playing worse and better, respectively, than Elo expected them to. Likewise, Clemson lost a few points for an unimpressive win against their much inferior Georgia Tech opponent.
It’ll take some getting used to, but this has been tested and is the better way to evaluate teams and predict game results.
Last weekend, I put out an article on quarterbacks and interceptions. The basis: a text-exchange from almost three years ago where Brian asked if the Penn State quarterback, still Sean Clifford at the time, played worse after throwing an interception. It sure felt like he did, but I didn’t know for sure.
Well, a lot of procrastinating and some more coding experience led me to finally answer the question last week. And the answer is a resounding yes!
The idea is straightforward: quarterbacks that throw an interception lose confidence, may be more turnover-prone to begin with, and put themselves and their teams in difficult positions where they may need to pass more and take more risks.
I broke the data up into pre- and post-interception plays, and looked at some key stats like predicted points added, touchdown rate, completion rate, and more, and then found which quarterbacks performed the best and worst after making a mistake.
On average, almost every stat is negatively impacted after an interception. However, some quarterbacks handled things better than others.
Some household names like Mac Jones, Joe Burrow, and none other than Trevor Lawrence were excellent at bouncing back from an interception.
For the full analysis, check out staturdays.com.
Week 4 Picks
With the changes in Elo methodology, here are the updated Elo discrepancies compared to the Vegas win probabilities. These are places where Vegas may be over or under-confident in the home team.
A few I will be eyeing are Ball State—who put up a decent effort against a tough PSU opponent the other week—over Toledo, Boston College—who are on a three-game winning streak—and Georgia Southern. Navy and Colorado really spook me, but knowing me, I’ll probably be proven wrong this weekend now that I mentioned it. Never doubt Elo in the long run.
Below are the full win probabilities for all P5 conference games this weekend. I’ll be out in the Happy Valley fall weather this weekend, so you won’t hear from me as much on Saturday, but I’ll be back early next week to update everyone on all the happenings. Enjoy your weekend!
A note here: Wisconsin doesn’t get any home-field advantage in this game because it’s being played at Soldier Field in Chicago, which is more of a Notre Dame home-game. It’s the first time in a long time I’ve seen a perfect 50/50 matchup!