FanPost

Is there a statistical correlation between talent and number of wins? Data suggest not so much.

Talent, talent, talent. There's such a prevalent notion that talent is the key to winning, and that question is really powerful throughout the year's coaching carousel. To scratch that itch, I evaluated team talent scores from 2015 to 2019 and several other factors to see if there really was a relationship between talent and the number of wins. (Talent rankings came from 247sports.) Long story short: there is no meaningful relationship between team talent and number of wins over that time.

  • I tried a number of different evaluations, such as limiting by conference, by windows of talent rankings, evaluations of talent + strength of schedule, and multiple variations of those combinations. There is not a relationship between talent and wins. The model was so bad that the prediction interval for a team with a talent score of 800 was between 3 and 14 wins. Here's what the basic graph for the period looks like. (For you math/stat oriented people, the R^2 value is 14.6%.)

Team Talent vs Wins, 2015-2019

  • I did find that there is a HUGE gulf between talent and conference. Median talent for all power five (PF) conferences for the entire period was more than 500; not one non power five (NPF) conference had a median talent score over 500 for that same period.

FGnVVUCX0AActwJ.0.png

  • Interestingly, I did find a strong relationship between, of all things, talent and strength of schedule. If you really consider what this means, it's easy to consider that the two create a self-licking ice cream cone. IE, the SEC is really THAT GOOD because of all the talent laden in the conference; this drives up the strength of schedule; and these heighten the value of wins while lowering the impact of losses. But y'all already know this. And, strangely enough, all of these high-end values (SOS and talent) align in the PF conferences with the SEC leading the dispersion. Here's what the regression (R^2 = 63.7%) and talent-SOS plots look like.

FGnX5nXXIAMrJOO.0.jpeg
FGnVgwJWQAoPvaI.0.png
FGnVkFpX0AcPbfj.0.png

The bottom line: talent is not a predictor of wins. I'm just guessing, but I'd think that a coaching staff and the system surrounding that coaching staff are far better predictors of success than talent. It's hard not to think of such examples: a talented UW team in 2021, USC over the last decade, and so forth. Furthermore, talent has a relationship with strength of schedule, and it's arguable that those two factors align over conferences and support existing biases.