# The Catch Radius Project: In Search of Better TD Production

For a while now I have been grappling with the idea of measuring a player’s catch radius. I was specifically looking to solve the  Red Zone TD Rate puzzle, since the metric can be flawed by usage and QB play. I wanted to see how it would correlate to WRs and touchdown production. So as I was contemplating the journey in my head, I decided to set up an experiment to see if there was any way to use a player’s Catch Radius to better predict the success of WRs. Using the Pythagorean Theorem, I devised a way to score a player’s catch radius. At RotoViz, we are always searching for better predictive metrics that will keep you a step ahead of the competition.  While the math may make your head explode, the results may have you chasing that little white rabbit down the catch radius hole.

## THE FORMULA

When it comes to measuring a player’s catch radius, there are two things that are important: A player’s velocity and a player’s ability to get vertical. The velocity measurement consist of the players 40 time, 20 yard short shuttle and 3-cone drill. These drills are used to measure a players agility and speed. We want to know how fast a player can get from point A to point B or how much ground a player can cover. The agility portion is especially important when it comes to route running and change of direction. The vertical score consists of a players height, arm length and vertical jump measurements. What we want to know is how high a player can get. What we have in the end (the simplified version) is (velocity score)^2 x (vertical score)^2 = Catch Radius Score^2.

## THE EXPERIMENT

I took 113 wide receivers on current NFL rosters and began gathering their data. The number 113 came about because that’s the number of receivers that I could find full data on and played significant snaps. Some players either didn’t complete certain drills or were injured during the combine and pro days. The 112 player sample was used to evaluate the correlation between a players catch radius and touchdowns per game.  I then plotted each player’s data on a scatter chart with TDp/G as the Y axis and Catch Radius Score as the X axis.

From this chart I was able to place the players into 5 categories based upon the mean, standard deviation and their individual scores. A Catch Radius Score of 200 is average. Players fell into the following categories:

>219 –  The players in this group averaged .215 TDp/G.
207-219 –  These players averaged .367 TDp/G
206-194 This group averaged .183 TDp/G
193-182–  This group averaged .205 TDp/G
<181–  This group averaged .221 TDp/G

### GROUP 1 (220+)

Only one player from this group has had significant production in his career,  Sidney Rice (222.35).. This group was littered with bust such as Jon Baldwin (229.87), Stephen Hill (222.07) and Greg Little (221.62). Of the group, David Nelson (222.37) and Justin Hunter (221.62) still have chances at a good career. While this group ranks the highest in scoring, they rank towards the bottom in touchdown production per game.

### GROUP 2 (207-219)

Vincent Jackson219.41
Julio Jones217.20
Marques Colston216.43
Brandon Marshall215.70
Torrey Smith214.99
Miles Austin214.98
Roddy White213.52
Dez Bryant213.49
Laurent Robinson211.34
Andre Johnson210.62
Victor Cruz210.61
A.J. Green209.88
Alshon Jeffery209.17
Mike Wallace208.88

As you can see from the table above, I have listed some of the stars of the group and their corresponding CR Scores. This group consistently out-produces the other groups in both touchdown rate and predictability. The coefficient of determination was consistently higher in this group when tested against the other groups. Past breakout players like Alshon Jeffery (209.17) and others score within this 207-219 range for catch radius.  Another player, Michael Floyd (209.15) going into his third season, is definitely on the RotoViz watch list for 2014. Players in this group also busted less than 9% of the time.  The two busts from this group are Darius Heyward-Bey (213.83) and Brian Quick (208.43).

Some players in this group that may be had for cheap in dynasty leagues include: Rod Streater (212.06), Andre Holmes (210.6) and Da’Rick Rogers (214.26), a player I warned everyone about here.

### GROUP 3 (194-206)

Group 3 is the largest group in the study consisting of 46 of the 112 players.  Dwayne Bowe (198.03) accounted for 4 of them. While this group has a few stars here and there, the overall production of the group is concerning. Two promising players by other metrics that fall into this group are Justin Blackmon (197.75) and Deandre Hopkins (202.32). Luckily for owners invested in those two, players like Hakeem Nicks (203.41) , Mike Williams (196.38) , and Pierre Garcon (195.31) give some hope of production in the future. Multiple season producers are few and far between when it comes to this group, so buyer beware.

### GROUP 4 (182-193)

Steve Smith (193.58) has been the most productive member of this group over his career. Other producers from this group, Greg Jennings (189.43) and Jordy Nelson ( 192.17) have had the luxury of catching passes from Aaron Rodgers. The only other producers are DeSean Jackson (181.27) ,who is an enigma at his size, and Stevie Johnson (193.55). The majority of players in this group are just guys. Productive members of this group have two things in common: they are their team’s number 1 option and see a high volume of targets . Moving forward, Nelson and Jackson are the only players in this group I would recommend investing anything significant in.

### GROUP 5 (<182)

All you need to know about this group is one name, Wes Welker (167.40). If you score this low, you’re probably not going to be winning any flag football games, much less producing on an NFL level. This group is made up of players that belong in a Willy Wonka sequel.  No one else in this group is even worth mentioning, so I won’t waste my time.

### CONCLUSION

The pool for this study consisted of roughly 70% of the NFL’s receivers. The players in Group 2, scoring in the range between 207-219, were not only the highest producing group, but also had the lowest bust rate among players. Players in this threshold have produced at a higher level in multiple seasons, as you would expect from a number 1 WR.

Two other observations were made when putting together the data and running the experiment for this article. When looking at total season points 86%, of players in group 2 had at least one season scoring 136pts or more. Group 2, while accounting for only 22% of the study group, accounted for more seasons hitting this mark than all other groups combined. They were also more likely to produce multiple seasons of usable fantasy production.

While its impossible to guarantee results with any one model, I think the Catch Radius Project has demonstrated it’s usefulness. Moving forward I will be targeting players within the 207-219 thresholds and cross referencing them with the Holy Grail to pinpoint my targets.

### THE ROOKIES

The interesting thing about this experiment is that the numbers are similar with college production as well when looking at the Combine invites.

>219 – Mike Evans was the only combine invite who scored in this range. He scored .653 TDp/G

207-219 – 11 players averaged .598 TDp/G

194-206 – 8 players averaged .345 TDp/G

182-193 – 16 players averaged .496 TDp/G

<182 – 2 Players averaged  .288 TDp/G

So which rookies rate between 207-219 in Catch Radius Score?  Well here’s the list:

1. Mike Evans – 222.73

2. Martavis Bryant – 218.21

3. Donte Moncrief – 213.22

4. Marcus Lucas – 212.33

5. Devin Street – 211.77

7. Jeff Janis – 210.63

8. Allen Robinson – 209.13

9. Brandon Coleman – 209.13

10. Kelvin Benjamin – 208.55

11. Jordan Matthews – 207.00

Notice Sammy Watkins is not on the list and Mike Evans may have run too fast a 40 time, but I’m still willing to invest. If you narrow down the rest of the list by using other RotoViz metrics like DR, then you have yourself a pretty good target list for dynasty leagues in 2014.

### Subscribe for a constant stream of league-beating articles available only with a Premium Pass.

1. It seems questionable to me that this metric has a successful bandwidth rather than just a cut-off. I’m usually skeptical of the hardline cut-off between success and failure based on a single metric since it seems like there’s a lot of slop in the combine results that go into determining these thresholds. In this case, there’s both a high-end cutoff and a low-end cutoff so it seems doubly troublesome. It’s not that this isn’t a worthwhile area of investigation but I would just caution against getting too reliant on a metric that has a “sweet spot” rather than just a continuous metric where higher/lower is better.

2. zjdlmt  i think i largely agree with this, but i think the value of this article is that 1) the headline is clear that it’s a project, which implies work in progress and 2) Scott is upfront with all of the results.

3. rotoviz zjdlmt  No problem and both of your points are valid. Since it’s a work in progress then that’s when feedback is most valuable and my feedback would be that it’s too limiting if you’re looking at a range of scores unless you have some concrete reason for doing so. For example, speed scores. It seems like huge speed scores are correlated with a higher rate of injury. I would understand if you defined a range of successful RB speed scores based on being good enough to be relevant, but not so good that they push you into the range of likely to get hurt a lot. In this case, it seems like the upper bound is just to eliminate some cases that failed and that’s not a sign of a good metric.

Again, I wouldn’t suggest abandoning things, but this seems like cherry picking to match existing data rather than being predictive which is what we’re all gunning for here.

4. zjdlmt rotoviz  i agree with that. i think it’s also the case that at a certain point we are all just finding a number of different ways to describe players that are athletic, so in some cases it actually makes sense to dial back the amount of credence that we give to a player’s specific type of athleticism.

5. I think all prediction models are questionable to an extent…I would never use a singular model for evaluation or prediction…there will always be players that both fit and break the mold in any model…I simply thought that the results of experiment were interesting enough to be aware of and to monitor in the future especially since the results were similar with college production…if the same group of players show up in multiple metrics I think it can be a great help in decision making and maybe breaking ties between similarly valued players…if production from players within the 207-219 threshold continues to be productive moving forward then the study may grow and hold more weight

6. Incredible work once again.  I’ve now added yet another function to my WR inspection algorithm.  Thank you for this.
The guy who stands out the most to me in every single analysis metric is Donte Moncrief.  Once again he falls in this list, and right at the sweet spot… not too high, not too low.  The comical part about it is, he will likely be drafted as a WR2 or WR3 for the team that takes him.

7. rotoviz zjdlmt  For me, the key point in the piece is that this could be a starting point for analysis or a portion of the analytical process. For most of the guys above 209, there were other red flags that could’ve suggested they were prospects to avoid. I think Evans differs from most of the 209 group, just as Sidney Rice does.

Anyway, I think you can see the continuum across the WRs, and I like seeing a lot of my favorites in the group above 207. Also, seeing Matthews right at 207 seems so perfect to me — he’s such a “boarderline” WR to me. Great production, good speed for his size . . . but I wish he scored more TDs in college, and my opinion on him really might come down to whether he’s drafted in the first round or the second.

8. Thanks for the feedback…the metric bounds were formed partly from the standard deviation and the mean…Mike Evans is a player that rates higher than the 219 and I expect him to succeed…the upper bonds unfortunately has a small pool of players partly because 6’5″ receivers with athletic ability don’t grow on trees…the picture will become more clear as time goes on and more data is available

9. ScottSmith610  this is absolutely a great start. nobody using numbers to make predictions is still using the first iteration of their model. so great start.

10. I love this outside the box thinking you writers do here. Who would of ever thought that they would actually use the Pythagorean Theorem in real life. Job well done.