revolutionary tools.  groundbreaking articles.  proven results.

The Ultimate WR Prospect Metrics Guide: Which College Production Measurements Are the Most Predictive of NFL Success? – The Wrong Read, No. 54

Welcome to the 54th installment of The Wrong Read. We know that college production is more important than athleticism at the WR position, but what are the best and most predictive ways to measure that production? This article compares 45 different ways to measure college receiving production.

My goal this offseason is to put together some new WR prospect models to help us predict early career NFL fantasy scoring. As this is an ambitious task, and taking longer than I want it to, I figured I’d start with something a little simpler — an exploration of what metrics are actually predictive of NFL fantasy scoring.

So, I gathered up a bunch of different college production metrics for every receiver from the 2005 to 2016 draft classes who played in an NFL game. The following table shows the results of running each of the metrics through a linear regression against average PPR fantasy points in a player’s first three years in the league. Below the table I’ll include some observations about how best to interpret these findings. Before displaying the table, however, I should explain some of the terms that might not be familiar:

YPTA: Receiving yards per team passing attempt

TDPTA: Receiving touchdowns per team passing attempt

Weighted Dominator: Combines MS of receiving yards and MS of receiving TDs in an 80/20 split, following Anthony Amico’s work.1

Raw: Metrics that are not adjusted for games missed (i.e., no team games are removed)

Adjusted: Metrics that are adjusted for games missed (i.e., games that a player missed are removed from team stats)

Peak: The highest mark a player ever achieved in a given statistic.2

The Results

MetricR-Squared
Raw Career YPTA0.1899
Adjusted Final YPTA0.1852
Adjusted Career YPTA0.1831
Raw Career Weighted Dominator0.1809
Raw Career MS Rec Yds0.1779
Raw Career Dominator0.1769
Adjusted Career Weighted Dominator0.1655
Adjusted Final Weighted Dominator0.1647
Adjusted Career MS Rec Yds0.1641
Adjusted Peak MS Rec Yds0.1614
Adjusted Final MS Rec Yds0.1612
Raw Final YPTA0.1601
Raw Career MS Rec TDs0.1587
Adjusted Career Dominator0.1575
Adjusted Final TDPTA0.1568
Adjusted Final Dominator0.1567
Adjusted Peak Weighted Dominator0.1556
Raw Career TDPTA0.1518
Raw Peak MS Rec Yds0.1492
Raw Peak Weighted Dominator0.1472
Raw Final Weighted Dominator0.1458
Career Rec Yds per Game0.142
Raw Final MS Rec Yds0.1416
Raw Final Dominator0.1415
Raw Final TDPTA0.141
Adjusted Career TDPTA0.1396
Peak Rec Yds per Game0.1386
Adjusted Peak YPTA0.1376
Raw Peak TDPTA0.1365
Adjusted Career MS Rec TDs0.134
Raw Peak Dominator0.1318
Adjusted Peak Dominator0.1296
Adjusted Final MS Rec TDs0.1292
Raw Peak YPTA0.1285
Final Rec Yds per Game0.1266
Raw Final MS Rec TDs0.121
Adjusted Peak TDPTA0.1201
Final Rec TDs per Game0.111
Career Rec TDs per Game0.1108
Peak Rec TDS per Game0.105
Raw Peak MS Rec TDs0.1003
Adjusted Peak MS Rec TDs0.08925
Career YPR0.03044
Final YPR0.02639
Peak YPR0.009585

So how do we interpret these results? One thing to note about these metrics is that a lot of them are colinear. For instance, although raw career market share of receiving yards and raw career weighted dominator rating have nearly identical R-Squared numbers, that’s mainly because the former makes up 80 percent of the latter. In a multiple linear regression model, for instance, many of these metrics would end up being insignificant. Nevertheless, we can identify a few trends.

Possible Trends

First, it appears that career metrics are typically more predictive than those that account for only a single season. Six of the seven most predictive metrics are career-level measurements. This makes some intuitive sense, for two reasons: Not only do good career stats indicate production from a young age, but they also give us a more complete picture of a prospect’s body of work than a single season. Single-season stats may leave out valuable information that career stats capture.

This last point relates to the second trend: it appears that, at least on the career-level, raw metrics are more predictive than their adjusted counterparts. That is to say, when measuring career market shares and dominator ratings, it’s better to leave in games that the player missed. Again, it would appear adjusting for partial seasons on the career level gives us less information — a less complete picture — than accounting for games missed when evaluating prospects.

Interestingly, this does not appear to be the case when looking only at single seasons. When comparing peak and final season stats, adjusting for games missed generally increases predictiveness. So although on a career level, factoring out games missed appears to remove valuable information, failing to do so on a season-level too harshly penalizes a prospect who missed games due to injury.

The only thing I would offer to possibly reconcile this seeming discrepancy is that it may be the case that good adjusted peak or final season stats alone do not necessarily signal success at the NFL level (see Parker, DeVante). However, a prospect who missed games in his final season should not be penalized if he also has good career numbers.3

Possible Confounders

Before drawing too many conclusions from these numbers, however, it’s important to realize that many of these metrics interact in different ways to numbers I have not included. For instance, a receiver’s weight appears to have an impact on which measurements are most predictive.4

Even though career market share of receiving yards and career yards per team attempt measure roughly the same thing — namely, a player’s receiving yardage while controlling for how often his team passed — the subtle distinctions in how that receiving yardage is measured appear to make a difference, especially when we incorporate weight as a variable.5

It’s also important to note that we haven’t controlled for draft position at all, which we know to be among the most important variables when predicting NFL success. It’s possible that some of the metrics are being priced into a player’s draft position. For example, we know that bigger wide receivers tend to outperform their smaller counterparts, all else being equal.

 

How Does Size Affect NFL Production_

But once we factor in draft position, that outperformance disappears.

bigvsmalldpos
NFL teams favor bigger WRs in the draft, meaning that although bigger WRs might have an advantage over smaller WRs, that advantage is already (at least mostly) accounted for once we pull in draft position.6 This could very well be the case for many of the production metrics above.

The key takeaway from all this is that in general we should be giving more weight to career metrics than single-season metrics, and in general we should not be adjusting those career metrics for games missed.7 However, the even-more-key takeaway is that further investigation is needed. Stay tuned.

Want to become an elite rookie drafter just by knowing one simple trick? Check out the Wrong Read No. 53, where Blair Andrews points out a quick, evidence-based method for fleecing your draftmates.

  1. Anthony called it Adjusted Dominator — I’ve changed the name here for reasons that will become clear.  (back)
  2. Note that a player’s peak statistics might not all be from the same season. Peak adjusted market shares and peak raw market shares, for instance, often occur in different seasons if a player missed games in one season. Note further that for many players, peak and final-season stats are the same.  (back)
  3. And note that for many players, adjusted and raw stats will be identical anyway.  (back)
  4. For the purposes of the chart below, a “big” WR counts as a prospect weighing 215 pounds or more. Although we’d hardly call a 214-pound WR “small,” the labels are just for the sake of convenience.  (back)
  5. We see the same sort of differences when we incorporate multiple metrics into a single model. In a multiple regression including only big WRs, career numbers appear insignificant, while final season numbers take precedence. On the other hand, a model including only small WRs shows that career numbers are most important, while final season numbers drop out. But this is a discussion for another installment.  (back)
  6. This does not necessarily answer the question of whether the NFL draft accurately values receivers, since TDs are more likely to be scored by larger receivers and TDs are much more difficult to replace than yards. It also doesn’t necessarily answer the question of whether fantasy drafters are appropriately valuing receivers.  (back)
  7. Though perhaps we should be adjusting single-season metrics.  (back)
Find An article
Generic selectors
Exact matches only
Search in title
Search in content
Search in posts
Search in pages

recent and related...

in case you missed it...

Mecole Hardman’s ADP Is Weird for a Couple of Reasons

  When the Chiefs traded up in the 2019 second round, the RotoViz draft room was abuzz. Kansas City was obviously moving up to select the replacement for Tyreek Hill. This was the pick that was going to make a fantasy hero. The Chiefs love athleticism in their draft picks,

Read More

Which Rookies are High-Stakes Drafters Targeting?

  Having the top pick in a high-stakes dynasty league is typically a double-edged sword: According to recent history, the consensus pick at the 1.01 has at times in their careers been valued as among the most coveted assets in fantasy. But, unless you traded for that pick, having it

Read More
Connect
Support

rotovizmain@gmail.com

Sign-up today for our free Premium Email subscription!

© 2019 RotoViz. All rights Reserved.