Earlier this week, I wrote about how our brains naturally form networks of connections. I also talked about how some of those networks are connected to nothing but themselves, not anchored to everything else we know.
That’s a neat idea in theory, but what does it look like in practice?
The story of Phlogiston
There is much knowledge we take for granted today, but there is always a time that predates that knowledge. In the late 1600s and early 1700s, for instance, we still had no idea what caused fire.
One popular theory at the time was known as the phlogiston theory. The basic idea behind the principle was that all matter contained something called “phlogiston”, and burning was merely the phlogiston being released into the air.
When something burned up and all that was left was ash, that ash was the substance’s true form, bereft of all of the phlogiston that was initially holding it together. When something burned, it was said to “dephlogisticate”. When a flame in an enclosed space burned out, it’s because the air had absorbed all the phlogiston it could hold and was therefore saturated- much like you can dissolve sugar into water until the water becomes saturated, at which point sugar stops dissolving.
Phlogiston theory explained every single behavior of fire. The problem, though, is that it didn’t predict any of these behaviors- it kept being adapted and updated after the fact with new details to explain any newly observed behaviors. And if something explains everything and predicts nothing, then it doesn’t really explain anything at all.
Football Phlogiston
Alright, so the 1600s were a weird time. Mankind was really just starting to experiment with science in some sort of rigorous and structured manner. We were still more than two centuries away from Karl Popper and the widespread acceptance of the importance of falsifiability.
Surely we now live in a more evolved and enlightened time. Surely we no longer carry with us beliefs that explain everything and predict nothing. Except, well… you know how we are with our faulty mental software and all.
If a team with a star receiver suddenly gains a quality second receiver, the star's numbers will surely go down because he loses targets. Unless his targets stay the same because the team passes more to offset. And even if the team doesn't pass more, maybe the star's numbers go up because defenses can no longer double-team him, resulting in easier coverages. You see, we have created a series of explanations for every conceivable outcome, but none of these explanations lend themselves to falsifiable prediction.
In essence, we have recreated phlogiston and applied it to football.
Now, in fairness, all of these component parts make perfect sense. Tougher coverage *should* lead to lower efficiency. More competition *should* lead to fewer targets. This is how football works. These facts are so self-evident as to practically be tautology.
The problem is in the interaction of these undoubtedly real effects. Again, the outcome could go in either direction. We have a ready-made story to explain everything, and therefore that story explains nothing. Not only is our ready-made story not right, it’s not even wrong!
Our belief about the interaction between targets and coverage, in other words, is an extensive and well-developed network… that we never bothered to connect to real-world results.
Dropping an Anchor
The entire network is so vast and sprawling that it’s really outside the scope of a weekly column to address it all in one fell swoop. I can, however, tackle one key component: does tougher coverage lead to worse efficiency?
Footballguys has target data going back to 2002. I went through that target database to find every instance where a receiver (a) finished at least 25 targets behind the team leader in year N, and (b) led his team by at least 25 targets in year N+1. My goal was to find guys who went from clear second options in the passing game to clear first options. The assumption is that defensive coverages would shift to reflect this new importance.
Please note that this search excludes players who were near the team lead in targets in year N. This might exclude some good instances of a player rising in prominence, but I wanted to avoid situations where players were 1A and 1B, (such as Ochocinco vs. Houshmandzadeh or Harrison vs. Wayne). Instead, I wanted to limit myself strictly to players who were true #2 options and then became true #1s.
In addition, I excluded every season where a player only finished 25+ targets off the lead in year N because he was injured. As an example, Jordy Nelson finished well off the target lead in 2012 and then dominated targets in 2013… but Nelson was one of Green Bay’s top options when healthy in 2012, so it’s hard to say that he was facing weaker coverages that year.
I also eliminated all players who switched teams, which tosses out perhaps the most famous example in the sample, (Peerless Price of the Bills and Falcons), but I wanted to control as much as possible for supporting cast. Did Price’s play fall off because of tougher coverages, or because Atlanta’s passing offense simply wasn’t as good as Buffalo’s?
Finally, I tossed out every player who did not have at least 20 targets in year N. I wanted to compare year N performance to year N+1 performance, so I needed to make sure they had enough performance in year N for the comparison. This also cleared out any players who were clear top options as a rookie, such as Anquan Boldin.
The Results
After all of the compiling and pruning, I was left with 29 names. Those 29 players averaged 65.7 targets for 500.3 yards as the secondary options in year N and 128.8 targets for 1026.1 yards as the primary options in year N+1. As a whole, their yards-per-target average actually rose from 7.61 to 7.97. In terms of individual players, seven players saw their yards per target decline by more than 0.5, thirteen players saw their yards per target increase by more than 0.5, and nine players saw their yards per target remain within +/- 0.5 of their year N average.
What does this mean? Well, there are a lot of lurking variables at play here. For starters, there’s a good chance that some of those players only became the #1 options because their efficiency improved. In addition, many of the names on the list were young 2nd or 3rd year players who were still improving.
On the other hand, there are plenty of established veterans who were thrust into the #1 role by injuries elsewhere on the roster. For instance, Harry Douglas had 59 targets for 396 yards in 2012, but 132 targets for 1067 yards in 2013 after injuries to White and Jones thrust him into the spotlight. His yards per target figure increased from 6.71 to 8.08 year-over-year.
Likewise, Julian Edelman was thrust into a larger role in 2013 with the departure of Wes Welker, jumping from 32 targets to 151 targets. His yard per target total declined, but only from 7.34 to 6.99, a value too small to be distinguished from random fluctuation given the sample size in year N.
So what can we take from this? For starters, it would be silly to say that players perform better against tough coverage than they do against easy coverage. There are other factors at work, instead, that are leading to this result.
One such factor is that the offense decides who gets the target only after the play has started. While receivers like Larry Fitzgerald and Calvin Johnson might get passes forced in their direction, quarterbacks likely are not going to force the ball to Julian Edelman or Harry Douglas unless they’re actually open. Tougher coverage might mean those players get open less often, but get the ball on a higher percentage of the the plays where they do get open, resulting in more targets without any damage done to per-target efficiency.
Either way, it’s pretty interesting evidence suggesting that the coverage a receiver faces has little impact on his per-target efficiency. Which hopefully helps build a link to our ideas of how coverage and targets interact to create fantasy value.
It’s not enough to completely anchor our old ideas, but it’s a start.
Second Thoughts
So… that Ben Roethlisberger guy is pretty good, it turns out. I’ve always been a fan of his talent, (in fact, he’s the only player I’ve owned since the startup of my very first dynasty league). With that said, he’s still an aging quarterback who absorbs a lot of hits and has never been a fantasy superstar. I always say that players are overvalued coming off of hot streaks, and Roethlisberger certainly qualifies. If someone is willing to value him as a top-12 dynasty quarterback, I would be happy to sell Roethlisberger off.
Reports have now come out that Houston will be moving to Ryan Mallett at quarterback, bringing the Ryan Fitzpatrick era to an ignominious end. I often value top backup quarterbacks ahead of the bottom 10 starters in the NFL for just that reason- just because you have a starting job today doesn’t mean you’ll produce any value before you get replaced. Better to gamble on guys who I think might suck than guys who I know for sure do suck.
On the other hand, Carson Palmer is another guy I’ve long held up as an example of a low-upside quarterback who, at this point, is who he is. Palmer’s had a long and interesting career, but he’s been on fire for the past month, ranking as a top-10 fantasy quarterback over that span. So far this season he has posted the second-highest passer rating of his career, and Arizona has rewarded him with a 3-year contract extension that will likely see him retiring with the Cardinals. He’s a tremendous feel-good story, but I stand behind my belief that he’s a low-upside fantasy quarterback who doesn’t produce anything you couldn’t get far more cheaply from a Kyle Orton type player. I’m rooting for him, I just won’t be owning him anywhere.
From the beginning of 2013 through the first six weeks of 2014 (22 games), Jacksonville had had a back get 10 carries and average 5+ yards per carry just twice. Denard Robinson has now done it in each of the last three games. Robinson is a great example of why you want to gamble on unknowns. Typically, they’re available quite cheaply, (Robinson was on waivers in most leagues), and while most will be complete busts, the ones that hit will provide a very nice, cheap return.
Golden Tate is doing a fantastic job of illustrating this week’s Big Takeaway. Over the first three weeks of the season, Calvin Johnson topped 10 targets each time, and Golden Tate averaged 5 receptions for 67 yards on 7 targets per game (9.57 yards per target). Since then, Calvin Johnson has three total targets and Golden Tate averages 8 receptions for 120 yards on 12 targets per game (10.15 yards per target). If he’s facing tougher coverages, it sure hasn’t impacted his per-target production.
On the other hand, if Calvin’s return corresponds to a commensurate decline in per-game targets again, Tate likely won’t see a bump in efficiency to match.
Don’t look now, but Rob Gronkowski is fully healthy and leads all tight ends in fantasy points and fantasy points per game. Meanwhile, some of the “safer” top tight ends such as Jimmy Graham and Jordan Cameron have been dealing with injuries of their own. I’m not going to say that I predicted that, because I didn’t- injuries are not predictable. That’s the point, though- this particular outcome was not really any less likely than an outcome where Graham and Cameron stayed healthy and Gronkowski got injured. Injuries are unpredictable. Unless a player has a specific chronic ailment, it’s best not to try.
Gronkowski is as young or younger than every other top-12 fantasy tight end except for Dwayne Allen. Even Travis Kelce is the same age as Gronkowski, (technically, Kelce is slightly younger, but both players have celebrated their 25th birthday within the last 6 months). Coby Fleener and Larry Donnell are both older than Rob Gronkowski. So not only is Gronkowski historically productive, he’s also shockingly young. Add in the fact that he appears to have some sort of mutant healing power and he’s my #1 dynasty player in all formats.