Does Success in the Transfer Portal Lead to Success on the Field?
In this new era of college football, how important is the portal to building a successful team?
Deion Sanders made headlines going into last season when he became the head coach of Colorado and essentially flipped their entire roster. According to On3, Coach Prime brought in 55 new bodies out of the portal ahead of his debut season with the Buffaloes. This gave him the confidence to claim that his team - which went 1-11 the year prior - was merely “about seven to eight dawgs away” when asked on ESPN’s First Take how close he thought his team was to National Championship contention. Colorado went on to secure a 4-8 record in 2023. This was surely an improvement from 1-11 - which, let’s be honest, was not a particularly daunting challenge - but a far cry from championship contention. Clemson’s Dabo Swinney has often been criticized for his inactivity in that transfer portal, which some have cited as a key reason for Clemson’s recent “struggles”.1 Swinney claimed “We'd rather take a good high school player as long as we can get them, which we do, and you (sic) develop them”. The transfer portal as we know it today is still in it’s infancy. While it was established in 2018, only since 2021 have players had the ability to transfer once without sitting out a year. As of this year, players can now transfer as may times as desired without sitting out. How much of an impact is the transfer portal really having on on-field success in college football? Is it worth it for coaches to try and navigate the new frontier of roster building, or is the good old fashioned “recruit and retain” method still the optimal path?
Portal vs. Recruiting
I gathered data from 247Sports on team recruiting and transfer portal ratings for the 2022 and 2023 season.2 To measure a team’s performance that year, I gathered their Elo ratings from CollegeFootballData.com using their R package CFBFastR.3 I then ran a linear regression model using team transfer portal ratings and recruiting ratings to predict a team’s end-of-season Elo rating. The results showed that portal rating was not a significant predictor of Elo when paired with recruiting ratings. Recruiting rating on the other hand, was in fact a significant predictor. The model resulted in an r-squared value of 0.410, meaning approximately 41% of the variance in a team’s Elo can be accounted for by recruiting and the transfer portal. The majority of this, however, comes from recruiting. Recruiting rating has an 0.64 correlation coefficient to Elo rating. Below you can see a scatter plot of recruiting rankings versus Elo rating for teams from 2022-23.
Transfer portal on the other hand, has a mere 0.39 correlation. A similar plot compares transfer portal to Elo below.
As you can see, Colorado clearly underperformed when it came to transfer portal success translating to on-field success. Based on the results from this linear regression model, when considering just recruiting and transfer portal, the relative importance - calculated as the percent contribution to the the r-squared value of the model - for recruiting on Elo is 81%, while just 19% for transfer portal rating.
What about returning players?
We can’t forget about the players that actually decide to remain at their chosen institution. The impact of returning players is not to be discounted. When incorporating the predicted points added (PPA) returning to a team from the prior season, the adjusted r-squared for the model rises to 0.499, with returning PPA showing up as a statistically significant predictor.4 In this model, transfer portal rating is not significant. The relative importance of the variables in this new model are as follows:
Recruiting rating - 53.5%
Returning PPA - 32.9%
Transfer portal rating - 13.6%
While still not as important as recruiting, returning talent beats out transfer portal activity by a wide margin. The table below shows last year’s rankings in each of the discussed metrics, along with how the team ended up performing.
Takeaways
While the portal may be the flashy new trend in college sports, it may not be the best way to build a winning college football team. For instance, only one of the top eight teams by Elo last year ranked in the top ten for transfer portal rating, and only five of the AP top-25 teams had top-ten portal classes. The “recruit and retain” method may still be the best way to construct a winning program in college football.
Looking Ahead to 2024
Where do this year’s team's stand in terms of recruiting, transfer portal, and returning talent? The table below breaks down each team, and you can sort the table by any of the metrics.
Colorado again sits near the top of the transfer portal rankings, but with the 81st recruiting class. Meanwhile, Clemson is tied for last in transfers but holds the 11th best incoming class of recruits. Does this foreshadow on-field results? I don’t know about you, but the numbers aren’t convincing me. Sorry, Deion, but I’m still buying the Tigers over the Buffaloes.
If you consider a 9-4 season and #20 AP ranking to be a “struggle”. Clemson has, however, finished the season ranked outside the AP top-10 the past three seasons following a string of six top-5 finishes and two national championships. They have, however, won at least ten games every year since 2011.
247Sports did not start tracking team transfer portal ratings until 2022.
“Elo” is a rating system used to measure relative skill. Developed by Arpad Elo, it was first created for rating chess players but has since been extended by many in sports analytics to rate players and teams in Football, Baseball, and Basketball, among others. For more on the Elo system, feel free to read more here.
Per CollegeFootballData.com, Predicted Points Added uses Expected Points (the number of points that would be expected to be scored based on down, distance, and field position) to measure the outcome of a play. It takes the EP value from the beginning of a play and subtracts it from the EP value resulting from the play.
Tremendously insightful on an important question. A few minor observations.
First, in my uneducated view relative data is not the same as absolute data. So, while the portal might be smaller in overall impact versus recruiting for talent acquisition, it doesn't mean that the portal has no value at all. Thus, Dabo's ignoring of it feels a bit specious and misguided - he's unnecessarily leaving value on the table for Clemson.
Having said that, the additional insight that the net outflow from the portal appears more important than the inflow is also very important. It makes intuitive sense that better programs will take care of their more prized recruits creating more retention and leaving somewhat less talented players for the portal.
Consequently, Swinney in theory might be able to show that his 100% focus on recruiting results in a more cohesive culture with less net outflow via the portal resulting in great talent. This is at least plausible and found in other settings too such as uniformed soldiers performing better than mercenaries, corporate recruits in the private sector out of college doing somewhat better long term than transfers from other companies etc.
Certainly, having a more organic and deep understanding the culture and practices of a particular organization are valuable and build loyalty. Still, if I were a Clemson fan I would be unhappy at the stridency of Dabo's position here. He might elect to use the portal less, but not using it at all strikes me as overly narrow and poor judgment.
Lastly, I would love to see this analysis expanded to other sports. While the NBA with its fewer players is clearly different, baseball and football with their larger rosters and robust free agent markets and annual Draft would seem to mimic well the collegiate portal/recruiting dichotomy. While different financial restrictions and limited rosters at the professional level are complicating factors, the essential question would seem to be very much the same. Are free agents a better or worse value than home grown Draft picks in building a winner? I'm sure front offices would love some insights.
Thanks as always.