The NFL offseason is dead. It's far too easy when football isn't on to get fascinated with workouts and potential instead of on-the-field performance, and there's no greater proof of that than the NFL Combine, held annually in Indianapolis as the last chance for incoming rookies to impress any potential employers, with millions of dollars on the line.
A few days ago, we took a look at what the offensive line workouts tell us about on-field performance, and today we'll look at wide receivers. Unlike the offensive line, a number of people have taken a look at how well receivers do after successful (or unsuccessful) days at the Combine.
By now, Combine skeptics well know the wide receiver duo of Larry Fitzgerald and Anquan Boldin, who both put up 1000 yards in 2008 on the way to a Super Bowl-and they both managed times in the 40-yard dash in the lowest ten percent of all current NFL wide receivers. They ran a 4.63 and 4.72, respectively.
Moreover, a quick look at the quickest electronically-timed 40-yard dash times reveals a whose who of nobodies (and former Oakland Raiders) along with luminaries like Chris Johnson and Champ Bailey. For every star who ran a sub-4.3 40-yard dash, there are nearly three who don't make it. Once you'll find Dominique Rodgers-Cromartie, another time, you could end up with Jerome Mathis (looking to find a job after the Hartford Colonials cut him, having failed to make a team).
In fact, over the past five years, only 14 1000 yard seasons came from receivers who placed in the top ten percent of 40-yard times, while 11 1000 yard seasons came from those in the bottom ten percent. Over the past five years, qualifying receivers in the bottom ten percent of 40-yard dash speed averaged 100 yards more per season than receivers in the top ten percent.
It seems that the 40-yard dash is useless, right?
Well, it's not entirely useless, but neither is it extraordinarily useful. When taken in context, however, it can be revealing. More than that, some workouts are much more powerful than you might believe, given the media attention that they (and other, less important workouts) receive.
The first step is to take a look at the work others have done and see how it performs. As always, DraftMetrics has been on the ball in figuring out whether or not Combine statistics provide any real value. Their measure of choice is the number of starts or seasons as a starter. Metrics are compared to their ability to predict one-year and three-year starters.
They initially found that the marquee event, the 40-yard dash, had very little impact on predicting future NFL success. Of the top thirteen players to run the fastest 40-yard dashes (top-ten times, including ties), only seven became starters for one season.
But, the event itself wasn't useless. The 20-yard split of that race did a pretty good job figuring out who was going to make it into the NFL. 74% of one-year starters ran the 20-yard split in 2.62 seconds or less, but that is a broad baseline. The average 20-yard spit of drafted players between 2006 and 2010 was 2.59 seconds, and 107 of 144 measured drafted rookies (74% exactly) in that time period posted a time equal to or less than 2.62 seconds.
The 20-yard split isn't useless-knowing that 74% of those scoring above a certain baseline become starters in the NFL is valuable information, but neither is it the most useful split. More on that later.
In their study, they saw that the vertical jump and the broad jump were moderately predictive of NFL success (as defined by the number of starts), but it didn't provide a lot of information. For them, the broad jump has a specific flag at 118, where 75% of all those who met that mark ended up starting in the NFL (defined as having eight or more starts in a season). In my sample, 72% of tested draftees bested a broad jump score of 118 and 33 of those (33%) started in their third season. Their sample is much larger and more holistic, so take the smaller sample done by me with a grain of salt.
DraftMetrics didn't combine any of the combine measures, which is a bit of a shame, as there's good evidence that it can be useful. Bill Barnwell of Grantland, and formerly of Football Outsiders, developed a statistic from the Combine that he calls a "speed score," which by itself has a very strong correlation for predicting running back success in the NFL.
When applied to receivers, the relationship isn't nearly as strong, but it is useful. Barnwell multipled weight by 200 before dividing that number by the 40 time taken to the power of the fourth, or simply this:
With running backs, the correlation (a number that describes the strength of the relationship, which runs from -1 to 1) was .45, fairly astounding for a simplistic number (0 means the relationship is meaningless, while scores of -1 or 1 describe a perfect relationship). For context, an obviously related statistic, the relationship between a receiver's reception yards and their team's total passing yards, was .59 in 2012.
What's really nice about the statistic, is that it averages at about 100, so it's easy to provide context. The best performances usually score a little above 120, while the worst performances end up with a total a little better than 80.
For receivers, Bill Barnwell's speed score was weakly predictive, but definitely significant. When looking at seasons for receivers three years after they were drafted, qualifying receivers in the top ten percent averaged 802 yards, while those in the bottom ten percent averaged only 662. In this metric, neither Boldin nor Fitzgerald are in the bottom ten percent.
That's not to say the statistic is perfect-productive receivers in the bottom ten percent include Danny Amendola,Brandon Lloyd and Wes Welker.
Eight 1,000 yard seasons were to be had by receivers in the bottom ten percent between 2008-2012, but those at the top comprised 19 1,000 yard seasons.
Aside from Barnwell, intelligent thinkers like those at Buffalo Bills Draft took a look at the combine numbers and found a distinct relationship for wide receivers. Instead of studying correlations, Tony Wiltshire applied a number of filters to see if there were indications of future success.
He took all receivers with a speed score greater than 100, a 3-cone time less than 7.15 and a college production ratio of 80, then saw if they could meet a certain threshold on a metric he calls "Explosive Power," which combines weight, height, vertical jump and broad jump numbers into one score. Of those above that threshold, 11 of 13 turned out to be good investments, with two more still up for judgment (Stephen Hill and Rishard Matthews). There's something to be said about this success rate, although only finding 15 players who met all the filters in 13 years of combine numbers gives it limited usefulness.
Still, it's a good read and I'll post an update on who fits in what category of prospect for his filters.
For receivers I again found a difference between predictive and descriptive statistics. Finding out who will grab significant snaps or put up yards in the NFL is once again different than figuring out which combine statistics do a good job predicting solid NFL players among starters.
Getting on to the field requires basic separation against top-tier competition, drive and impossible-to-ignore physical capability. In order to generate that separation, one would normally think of the agility drills as the best predictor. After all, movement in small spaces define how one can create passing windows out of nothing.
That's not the case, as it's explosion gets a receiver time on the field. The vertical leap and 10-yard split are among the two most important combine measures that receivers can do well in in order to start getting snaps. After the initial burst, it doesn't seem as if additional quickness down the stretch correlates well to grabbing more snaps.
The final equation for predicting NFL performance and participation from Combine numbers is easy enough to understand, if a little long:
All of this indicates that height, weight and quickness are important. The most important of any measure, however, is the vertical leap. Leaps below 28.5 and above 41.5 don't provide any additional predictive value, so there is a ceiling and a floor to how much they help.
Of the indicators used in the formula, height was least important, which implies one of two things. First, either NFL teams don't see a huge advantage in how high you can go-a receiver who is 5'10" with a vertical of 39" would typically and consistently see more snaps than a 6'4" receiver with a vertical of 37", despite the fact that the second receiver would have an advantage fighting for the ball.
The second implication, which seems more likely, is that the vertical leap measures much more than simply how high a receiver can jump. That measure of leg explosion from a stand-still manifests itself all over the field in the form of exploding out of cuts, getting off the line, beating blocks and accelerating through the ball.
This second implication is further confirmed by the fact that treating height and vertical leap as one variable (by adding them together to create a "jump ball score") is less predictive of how many snaps a player will take than vertical leap alone by a significant margin.
Shuttle speed and three-cone speed are nice, and it seems that short-area quickness impresses coaches enough to put players on the field, but it's the 10-yard split time that matters the most. Like vertical leap, it will measure acceleration and smoothness out of a stance. Strong 10-yard split times come from receivers who don't take a false step at the combine when running the 40, and that ability to avoid inefficient movement translates well to NFL play in a variety of ways.
In determining who takes snaps and who doesn't, the 40-yard dash is very weakly predictive, but the 20-yard split is very good, and the 10-yard split even more so-the relationship was twice as strong in the 10-yard split than the 40-yard dash.
This runs counter to some other research DraftMetrics has done, which indicates that 40 times are the strongest indicator of future starts. Given that this research doesn't align with later work that they've done on the importance of Combine numbers and that they come to their conclusion with the number of "one-year starters" while my work looks at the number of snaps a player takes, I'm not too concerned. They merely looked at the top percentile of each combine measure and saw how many one-year starters were in each batch. That is not an indication, holistically, that the 40-yard dash predicts performance, just that it allows players to get a shot on the field (think Jacoby Ford).
What's more interesting, however, is that among receivers who play a significant amount of snaps, completely different combine results matter, including one that might be the biggest surprise I've encountered so far. Out of every combine result, the single best predictor of success on the field (measured by total yards, yards weighted by receptions and touchdowns, yards per route run or any other reasonable measure) is the number of bench press reps a player took.
It's not very close either-bench press reps nearly rival draft picks in their predictive capability. Unfortunately, receivers have only recently started doing the bench press at the Combine, so it's difficult to see how it holds up historically (although 30 of the 38 receivers at the Combine participated). Nevertheless, I have enough samples (160 individual player seasons) to be weakly confident of this fact, at least in the short run.
When split into random samples, the correlation of bench press to performance remained equally strong, so it's not just the result of noise.
This might be a result of a number of things. For players recruited to play receiver in college, speed comes fairly naturally, but strength doesn't always appear for players who show up to college weighing between 175-200 pounds. That means that they need to invest a significant amount of time in the weight room in order to maximize their upper body strength. That sort of work ethic translates well to performance in the NFL.
Moreover, it takes a lot of strength to beat defensive backs off of press coverage and maintain timing, which is more critical in the NFL than in college-timing-based offenses are popular in the professional circuit, but can be easy to disrupt if receivers don't get off the line.
Beating defensive backs to the ball, both in the air and when pushing off, requires strength that allows a receiver to get to the ball unimpeded.
Still, it seems to be an outsized affect, but it is perhaps one of the few things that can't be covered with good instincts and technique. Receivers with poor speed can make up for their inability to fly across the field by using techniques to trick defensive backs or intelligence to sit in the spaces between zones. There are techniques to beat press coverage or get arms free for the catch (known as "creating real estate") but those can only get you so far without strength.
Other than that, the results of other workouts are fairly fascinating. For example, 40-yard dash times have positive descriptive power for yards per reception-the faster you are, the higher your yards per reception. That's no surprise, but the interesting bit comes from the fact that it has nearly identical negative descriptive power when it comes to receptions or receptions per snap. It almost implies that the faster a player is, the less likely he is to catch the ball.
The net correlation between 40-yard times and total yards is nearly 0, with a slight negative descriptive tendency.
My guess is not that the 40-yard dash is a bad thing to do well in, just that more players receive opportunities to perform on the field as a result of straight-line speed than other players with different, but equal, characteristics. That results in an oversampling of receivers with fast 40-times.
Another way to think of it is like this: when confronted with a choice to draft sixth-round receivers with generally equal grades (and therefore a theoretically equal chance to make a big impact in the NFL), NFL teams will generally choose the player with the better 40-time, perhaps in an attempt to grab "upside". This doesn't mean the 40 is bad, but that you might see a larger proportion of mediocre players play in the league with better times.
The relationship is relatively weak, though, which suggests that NFL teams don't have much patience for players who don't perform, regardless of straight-line speed.
The overall formula that projects success for receivers with significant snaps already in the NFL slightly exceeds the relationship between draft pick slot and performance (unlike the formula that attempts to project from the combine on). When combined with draft slot, the combine scores have a correlation that exceeds .5, which is very powerful indeed.
The formula above works just as well predicting yards as it does a weighted score that gives points for receptions, yards and touchdowns—you may even use it as a tiebreaker for your fantasy league. Like all the formulas used so far, it ranks prospects generally from 0-10, except in extreme cases (T.Y. Hilton scored negative 0.4, for example).
This unfortunately only works for 168 of the 550 people surveyed, given that few receivers participated in the bench press (which, like I said, correlated extremely well with success). Taking the bench press out weakens the formula, although it is still somewhat predictive. It just so happens that the bench press result by itself is more predictive than the rest of the formula. At any rate, the other important measures one can use to predict success are as follows (if bench press information isn't available).
Weight has been an enormously powerful predictor along with the bench press, and features prominently as a predictor not just of yards, but yards after the catch and touchdowns. The vertical leap is likely important here because, combined with weight, it is a measure of explosiveness, not just the ability to get into the air. With leg strength and a big body, it's hard to move receivers to places they don't want to go.
Height has been a surprisingly poor predictor, but it can make other measures more useful. By itself, it has a negative correlation with yards after the catch, and bears nearly no relationship to total yards. The fact that it has a negative correlation with receptions further speaks to how it doesn't provide as much of a benefit as one might think. Given that the separation between most receivers is about four inches, that's not a huge surprise.
But when added to weight (which is likely here important because it is an indication of muscle strength) it becomes powerful by itself, particularly when predicting touchdowns.
Not too different from Barnwell's speed score, it's more a measure of the ability to move a compactly formed body and less overall force. This predicts efficiency measures as well as total measures, and does a decent job figuring out whether or not a receiver's presence on the field is generally positive, even if he doesn't grab big yardage. It also generally correlates well with forcing missed tackles.
The Combine might be overrated, but neither is it useless. Have fun seeing where your favorite receiving prospect stands.