Talent Score for 2023
A simple metric that can help us predict the CFP participants and champion
Well, we made it! College football starts Saturday.
There are very few things that I produce each year that I actually get really excited for. As fun as sports and betting are, sometimes the work still feels like work, especially if it involves data entry and tedious computational stuff. I love this field, but genuinely, really getting excited? Hard to get there sometimes.
But Talent Score is a thing I actually get excited to look at. When it’s done, I kinda just stare at it for a few minutes, and maybe utter some incoherent grunts. “Ohhhh.” “Ahhhh.” It’s fun.
What is Talent Score? Glad you asked! Even if you didn’t.
When I first got really into college football, and especially it’s betting markets, about 6-7 years ago, I wanted to create a simple metric to help quantify the level of talent at a specific school. My theory was that if you could just take the question “how good are the players at each school?” and put that into a number, it would answer a lot of questions, no matter who the coach was or the quarterback was, or what the history of the program was.
My goal was very straightforward: simply combine two very fundamental rankings/metrics: recruiting, which is a measurement of how good the players are that you bring into your program, and returning production, which is a measurement of how many of those players you maintain and develop to continue playing for you. Get good players, ideally develop them, and then keep enough that they can produce lots of rewards for your school before either graduating or turning pro. Turn that whole thing into a score.
So, I did that. I took lots of recruiting averages, tried to figure out what the best way to aggregate them was (is getting lots of good players better than getting 1 great player, for example), tried to distill that as much as possible, then figure out returning production, inspired by Bill Connelly of ESPN, whose SP+ is the best publicly-available rating in the sport. Get really good players, keep them for some amount of time, hopefully a lot, develop them, and win.
Recruiting + Returning Production, weighted and distilled, equals Talent Score.
Before I go into how this experiment turned out, and a look ahead, it’s important to note we may be nearing the end of how much Talent Score can teach us. The transfer portal has warped our ability to measure the players at a school using these two metrics. If you hypothetically lose every player, then get non-recruited players (transfers) to fill all those spots, it’s clearly a blind spot in this rating. It will definitely register for the team losing the player, but maybe not so much for the team gaining a player. But although the portal is quickly becoming a huge aspect of the sport in so many ways, we aren’t yet at a place where it’s the only thing, and I think we can still proceed here knowing transfers do matter, while also understanding there’s a lot of predictive quality in recruiting and returning production. That’s kind of a disclaimer, while I figure out how to perfectly integrate transfer stuff into Talent Score. But I wanted to say that.
How predictive has Talent Score been previously? Very. What does it say this year? A lot. Let’s dive in.