Now that the NFL season has officially ended, #DraftTwitter is in full effect. There is no shortage of opinions on which players should go where, as film-grinders and number-crunchers let you know their favorite prospects. And this is all happening weeks before the NFL combine and months before the NFL draft. Even if you take a numbers-based approach to draft prospect evaluation, you still have find the most relevant stats and how much to weigh them.
In an attempt to do just that, I created a logistic regression model to predict wide receiver success, before the NFL combine and draft. Therefore, we can use the model today to judge 2016 prospects, and then update the model as more information becomes available in the coming months.
There are many different ways to measure success. You could have a fantasy-point threshold, or total career stats like receiving yards. At this time of year, we’re ultimately concerned with success for our dynasty rookie picks. I measured success as having one top-24 scoring season (PPR) in the first three years in the NFL. If a wide receiver has one top-24 early in his career, his market value is probably high enough at that point to make a profit on your initial rookie draft investment.
Armed with 15 years of college and NFL wide receiver data, I trained the model on 254 wide receivers to enter in the NFL from 2000 to 2013 (I didn’t go through 2015 in order to have three years of NFL experience). Then, I tested the model on another randomly chosen 136 prospects to make sure that the model had true predictive power.
In the process, I narrowed down dozens of collegiate stats to four that maximized predictive power:
1. Draft age
2. Career market share of receiving yards (i.e. a prospect’s total receiving yards for all collegiate years as a fraction of total team receiving yards)
3. Receptions per game (final season)
4. Receiving touchdowns per game (final season)
All four variables were found to be significant in the model, with a younger draft age, higher career market share, and more receiving touchdowns all being indicators of future success. Interestingly, fewer receptions per game is a positive in the model, likely because the model prefers receivers who accumulated yards and touchdowns through efficiency, not just high volume.
These four variables do a remarkable job predicting whether a wide receiver prospect will have a top-24 season in his first three years. The model correctly predicted that a wide receiver prospect in the test set would be a success or failure 80.9 percent of the time.
Remember, this model doesn’t incorporate any data from the NFL combine, and more importantly, the NFL draft. Draft position is clear one of the most important variables for predicting prospect success. That said, our model was only slightly less successful at predicting success than draft position (38.5 percent versus 42.3 percent).
Here are the top-15 “success” scores from the 2000-2013 data set.
| Name | Year | Draft Pos | Age | Career MS | Rec/Gm | RecTD/Gm | Top 24 | Predict |
| Larry Fitzgerald | 2004 | 3 | 21 | 0.41 | 7.1 | 1.7 | 1 | 0.91 |
| Chris Henry | 2005 | 83 | 22 | 0.47 | 4.7 | 1.1 | 0 | 0.88 |
| Antonio Bryant | 2002 | 63 | 21 | 0.39 | 4.2 | 0.9 | 0 | 0.82 |
| Calvin Johnson | 2007 | 2 | 22 | 0.42 | 5.4 | 1.1 | 1 | 0.78 |
| Charles Rogers | 2003 | 2 | 22 | 0.43 | 5.7 | 1.1 | 0 | 0.78 |
| Koren Robinson | 2001 | 9 | 21 | 0.36 | 5.6 | 1.2 | 1 | 0.78 |
| Hakeem Nicks | 2009 | 29 | 21 | 0.38 | 5.2 | 0.9 | 1 | 0.75 |
| Quincy Morgan | 2001 | 33 | 24 | 0.48 | 4.9 | 1.1 | 1 | 0.74 |
| Dez Bryant | 2010 | 24 | 22 | 0.35 | 5.7 | 1.3 | 1 | 0.72 |
| Lee Evans | 2004 | 13 | 23 | 0.43 | 4.9 | 1.0 | 1 | 0.70 |
| Sidney Rice | 2007 | 44 | 21 | 0.39 | 5.5 | 0.8 | 1 | 0.69 |
| Nate Turner | 2001 | UDFA | 23 | 0.47 | 5.5 | 0.8 | 0 | 0.68 |
| Deandre Hopkins | 2013 | 27 | 21 | 0.29 | 6.3 | 1.4 | 1 | 0.67 |
| Demaryius Thomas | 2010 | 22 | 23 | 0.43 | 3.3 | 0.6 | 1 | 0.63 |
The “Predict” column gives the model score (between 1 and 0) for each prospect indicating the likelihood of a top-24 PPR season in a prospects first three years, and the “Top 24” column indicates whether or not the receiver actually had a top-24 PPR year in his first three seasons. Draft position is listed in the table only for reference; it was not part of the model.
It should come as no surprise that our model favors some top-5 NFL draft picks, like Larry Fitzgerald and Calvin Johnson, who combine a younger age, dominant career market share, and touchdown-scoring prowess.
Since our model only is concerned with production and age, second- and third-round picks like Chris Henry and Antonio Bryant score just as high as top-5 picks, despite the fact that both had their draft position negatively affected by off-the-field issues.
It might surprise some to see Hakeem Nicks so high on the list, but he had incredibly strong college production at a young age. Nicks initially carried that through to the NFL, but mounting injuries turned the once promising star into a replacement-level player.
Our model likes did a great job of finding successful players who were in the 20-50 draft position range, like Dez Bryant, DeAndre Hopkins and Demaryius Thomas. The success rate for picks in this range is much lower than for our model, and there were some very early NFL draft selections that our model didn't look kindly upon (Tavon Austin only has a predict score of 0.09).
Now that we have our pre-combine, pre-draft model for predicting success in place, in the next article I will apply the model to 2016’s wide receiver draft class.
In the meantime, below is a sortable database that you can explore and see how all the receivers who entered the NFL from 2000-2013 fared in the model.
Kevin Cole is a Lead Writer for PFF Fantasy. You can follow him on Twitter at @Cole_Kev
[table id=1164 /]
© 2025 PFF - all rights reserved.