The Year 1 zero theory is well known through the Campus2Canton crowd. The researched theory is that unless freshman Wide Receivers hit certain thresholds in their first season, then they are very unlikely to get drafted or have success in the NFL. Therefore it is recommended that Devy/C2C players both A) trade those players after their freshman years before their value completely craters and B) avoid drafting these players once they’re into their sophomore or later years.
I’ve tackled this same theory by using a modeling technique called Logistic Regression, which estimates the probability of an event occurring. In this case, the aim was to estimate the probability a Wide Receiver would be drafted on Day 1 or 2 of the NFL Draft (Rounds 1 to 3) however, only using information available at the conclusion of their Freshman Year.
The variables that went into the base of the model were a combination of the 247 Composite Rating as a recruit, and their Year 1 Receiving Yards Per Team Pass Attempt, and whether they attended a Power 5 school or not. For those Wide Receivers that had sufficient information to have a Recruiting Wide Receiver Model Score then, this was used in place of the Composite Rating (and it also rendered Power 5 school or not useless too). The group studied was limited to those who had a 247 Composite Rating greater than 0.9000. The idea was that it would limit the population to the Top 300 recruits each year, which was the original process with the Year 1 Zero theory.
One way to measure the accuracy of a Logistic Regression model is by calculating the Area Under the Curve (AUC). That evaluates how well the model correctly predicted whether or not the target variable was met. An AUC of 0.5 means that the model was only correct 50% of the time when predicting whether the target variable was correct, which is a coinflip, which means the model would not be helpful at all. So if two Wide Receivers were to not be drafted on Day 1 or 2 of the NFL Draft, but one was drafted on Day 1 or 2, then this would be an AUC of 0.5. If one Wide Receiver was predicted to obtain this draft capital and did, and another was predicted to not obtain this draft capital and didn’t, then the AUC would be 1.0.
For the base model, the AUC = 0.743. For the model which included the Freshman Wide Receiver Model Score, the AUC = 0.85. So the base model was correct 74.3% of the time when predicting whether or not a Wide Receiver was going to be drafted on Day 1 or 2 of the NFL Draft when only using information available at the end of their Freshman year, and the other model and group was correct 85% of the time. Keep in mind that the heavy majority of Wide Receivers each year were predicted to not receive Day 1 or 2 Draft Capital, which makes sense. The aim of this model is to successfully find that small group that day finds success.
The above table has the 20 Wide Receivers who had the highest predicted probability of obtaining Day 1/2 Draft Capital since the Recruiting Class of 2010. You can see it was fairly successful, and if you obtained these Wide Receivers after their freshman years, then a lot of them were very good investments, even if you may have been a bit worried about Diggs and Amon-Ra St Brown after the draft!
The above table is this year’s Draft Class. Despite Jaxon Smith-Njigba’s poor Freshman season, his strong Recruiting Wide Receiver Model Score still encouraged us to believe in him. Overall I think that if I invested in the top of this list after their Freshman seasons then, then I’d be fairly happy with how their draft capital is going to shake out.
Finally, onto something that is actionable. The top five on this list would be my most recommended Wide Receivers of the 2025 Draft Class to trade for or to draft in Devy or C2C leagues. Funnily enough, Mike Vallerie has also come to the same set of top five names (albeit in a different order). There is a large drop in predicted probability from Barion Brown to Kobe Prentice and then from Tetairoa Mcmillan to Isaiah Bond.
So what is my recommendation on how to best use this information? Take the predicted probabilities and then adjust using whatever context you feel is appropriate. Using what you’ve seen on film, what someone you trust has seen on film, the team or coaching staff they play for, any injuries that occurred to the player or the team around them, or were they forced to sit behind some absolute studs (probably not when I look at this group of names, but in other years, yes – think Ohio State/Alabama).
As a data-based fantasy football player, I pay homage to both JJ Zachariason and Ben Gretch when I say how important it is to recognize that our data projection in a small sample game should have large error bars, and that is ok!
Go watch some highlights, and come up with your own reasons to either support or go against any (ok, most) data-based analysis such as this, and have some fun in this game that we call…. a game ✌️