≡ Menu

The off-season is here.

But Football Perspective isn’t going anywhere. I prefer off-season writing to in-season writing, as football theory and history is more compelling to me than figuring out whether to rank the Lions ahead of the Bills. At the old Pro-Football-Reference Blog, we did some of our best work in the off-season, and I hope for the same results here.

Evan Silva just published a great piece detailing what each team needs in the off-season, but you’re not going to find that type of article here in the off-season. I will have some draft articles, but I don’t intend on staying topical all that often. My first big off-season project is to come up with a wide receiver ranking system.

I won’t bore you with all the details yet, but I think grading wide receivers (or for that matter, receiving tight ends) is much, much more complicated than people realize. I hope you guys are excited to participate the discussion, as I am in the early stages of this project and will go where the research takes me. One possible result I envision: the ultimate wide receiver ranking system does not exist, but a series of four or five ranking systems might give us the complete picture of a wide receiver.

Let me start with a question: which team had a better passing offense last year, Houston or Detroit? For now, try to ignore what we saw out of Matt Schaub in the post-season and just focus on the regular season results.

Avg Team338.5555.960.93700.623.74.314.

The Lions passed for more yards and had a better sack rate, but otherwise it’s a pretty clean sweep for the Texans. Whether you go by Y/A, AY/A, NY/A, or ANY/A, the Texans have the edge, and they also lead in things like passer rating and completion percentage. For the purposes of this post, I am going to ignore strength of schedule because I want to talk about the actual stats. SOS adjustments are easy to incorporate at the end.

The Lions passed for more yards, but that’s only because they threw more passes, so I’m not particularly inclined to give them any credit for that. It’s true that being very good on a large number of attempts can be more valuable than being great on a smaller number of attempts. Bayes’ theorem also informs us that by being very good over a large number of attempts could be an indication that such team’s true ability is more likely to be above average than by being great over a smaller number of attempts. But since the Lions were right around league average in Y/A, AY/A, NY/A, and ANY/A, I don’t know how much weight those points hold in this particular case. Another argument could be made that by shouldering a larger burden of the total offense, the Lions’ pass efficiency metrics would be artificially suppressed. I don’t really agree with that statement but do not feel like supporting my position with evidence at this time.

But I’m not willing to give up the fight on behalf of the Lions just yet. Football Outsiders, in their non-SOS adjusted ratings, ranked Detroit’s passing offense as slightly better than Houston’s. FO also ranked Detroit’s rushing offense as slightly better than that of the Texans.

Here’s another way to think of it. These two offenses were pretty equal. Houston ranked higher in points scored, but if you take away their five non-offensive touchdowns, the Texans would have outscored the Lions 381-372. Considering both teams had slightly above-average offenses in things like drive success rate, yards per drive, and points per drive, and the fact that Detroit’s passing game accounted for 75% of its offense compared to just 64% for Houston, perhaps that means Detroit’s passing game was better.


Calvin Johnson is still the man.

I don’t know the answer. My natural inclination is to pick ‘Houston’ but I think it’s a difficult question. Let’s call it a draw.

Andre Johnson was responsible for 39.5% of the Lions receiving yards, compared to 38.3% for Calvin Johnson. If we look at receptions, Johnson had 31.6% of Houston’s receptions compared to 27.4% for Megatron, while Calvin Johnson caught 5 of his team’s 22 touchdown passes compared to 4 of 22 for Andre Johnson.

Unless that one extra touchdown makes a big difference (it doesn’t), the natural conclusion is that Andre Johnson was a (slightly) bigger part of the Houston passing game than Calvin Johnson was of the Detroit passing game. If we agree that Houston’s passing game was better or equal to Detroit’s passing game, then we would have to say that Andre Johnson was the better (or more valuable) receiver. (I suppose another conclusion would be that Schaub is much better than Matt Stafford, and that if you switched the two quarterbacks, Detroit’s passing offense would have been much better.)

I think we can safely ignore the touchdown and reception data (at least in this case) and summarize the argument like this: Andre Johnson was responsible for 39.5% of a passing offense that averaged 6.3 ANY/A (or produced 231 adjusted net yards over average) while Calvin Johnson was responsible for 38.3% of a passing offense that averaged 6.0 ANY/A (or produced 38 ANY over average). Therefore, Andre Johnson was better (or had the better year, or had the more valuable year).


I noted earlier in the year that Houston’s Johnson averaged more receiving yards per team attempt than Detroit’s Johnson, although I also recognized there were some limitations to that study. But considering their respective contributions to their teams, you would need to rank Detroit’s passing offense as considerably better than Houston’s to say that Megatron had the better year (all else being equal). If the argument that Calvin Johnson had the better year boils down to nothing more than “Detroit threw 186 more passes” then there is no argument at all.

Finally, note the tag to this post: I hope to place a lot of posts at http://www.footballperspective.com/tag/wr-project/ over the next few months.

  • Good work and informative.

    • Chase Stuart

      Thanks Pat. Hope you enjoy the whole series.

  • Boston Chris

    Hi Chase, I’ve always loved your stuff back to the pro-reference days, and am thrilled to have stumbled across this blog recently. This project reminds of the time and effort put into coming up w/ AV. A tremendous, albeit necessarily flawed, achievement.

    I think at the outset it is important to keep in mind some of the issues any analysis is going to have. One that comes to mind is coverage. The second best WR on some (most?) teams are going to be overrated by any stats system I can think of because of how the defense is covering WRs. How do you account for the double teams Calvin Johnson sees? Does Andre Johnson see the same amount? Also, since we are focusing on WR play specifically, how do we account for usage? You are talking about on a basic level w/ counting stats as opposed to efficiency stats which is a start, but does 1 WR have an advantage over the other because of the type of pass plays that they are being asked to run?

    Anyway, I love your work, and am excited to watch this develop over the offseason. And just want to clarify that despite some of my misgivings about the things above, I do recognize that coming up w/ the best possible system still has tremendous value, even if we can’t get a “perfect” one.

    • Chase Stuart

      Thanks Boston Chris — glad you found the site! Yes, the origin of this particular post has the same roots as AV.

      I agree that we won’t be able to get a perfect one. I’m even skeptical that there is a best possible system, which is why I hinted that we might need multiple systems to really give the complete picture. I do know that just looking at receiving yards/receptions/touchdowns doesn’t get us very far.

  • First two things that pop into my head:

    1) Outside of the pre-points, isn’t “Andre Johnson was responsible for 39.5% of the [Texans] receiving yards, compared to 38.3% for Calvin Johnson” essentially the core of AV for WRs? In other words, using something like this in a WR rating seems to me to just be replicating AV.

    2) I don’t know how doable it is with your database pre-2000, but FO’s of the mindset that WRs should be rated based on all targets, not just their receptions. We have the target totals going back to 1991 on our site, but I’m guessing you want to be able to rate WRs before then too.

    • Chase Stuart

      Thanks for checking in Danny.

      1) Yes, this idea has its roots in the method behind AV. That’s not a bad thing, in my opinion, and the different method of pre-points is a substantive difference here.

      That said, I think you can make several arguments for what the right pre-points should be: passing efficiency or total offensive efficiency, for starters.

      2) If there are X number of sites/writers in our industry, I would rank X in placing value on targets. This will be another post (or two). I’m not sure if I will convince everyone or anyone, but I do find the emphasis on targets by well, just about everyone, to be misguided. I understand that I may be on a two-man raft here (with Drinen, naturally. Come to think of it, I can’t recall where JKL fell out on this one.)

      You are correct that the goal is to go back to 1932.

      That said, perhaps we can work something out with your target data. I think target data is generally overused, but it still has some value. I think I have target data through 2002 via FBG.

    • Neil

      Unsurprisingly, I’m in the same raft as Chase and Doug here. In order to get a target in the 1st place, a WR has to do *something* right. There’s a huge amount of selection bias if you ignore the volume of targets a guy gets and only focus on per-target production, because guys who aren’t as good at getting open by and large don’t get targeted as much.

      • Chase Stuart

        Continue to fight the good fight, my friend.

  • Neil

    Just noticed that Johnson & Johnson in 2012 had the 2 lowest TD catch totals of any WRs who ever had seasons with catches or yards of that magnitude (100+ rec, 1500+ yds):


    • Richie

      Andre Johnson shows up 3 times in the top 8 of that list. That almost seems like a trend. Though, 2008-09 Houston was a very different team than 2012 Houston.

  • Calvin averaged 1.8 yards more per reception than Andre did on almost the same number of receptions. I don’t think you can ignore that when comparing player vs. player. Andre averaged 4.2 yards more than generic other Houston receiver, Calvin averaged 6.3 yards more than the average other Detroit receiver.

    Andre’s catch % was considerably better, but even doing yards / target Calvin did better than Andre did vs what their other teammates did (3.59 yards / target better than teammate for Calvin, 3.54 yards / target better than teammate for Andre). Calvin’s catch % was worse than his teammates by 1.6%, Andre’s catch percentage was 6.5% better than his teammates catch %. Even with those advantages for Andre, Calvin was still gained more relative yards per team target than Andre did as Andre’s advantage in yards / target over Calvin is less than the difference between Houston teammate yards / target over Detroit teammate yards / target.

    (Detroit team targets: 730, Houston: 549. QB spikes, passes with no intended receiver make up the other 10 Detroit pass attempts. Houston had 5 of those. Other math regular stats.)

    If the hypothesis is that Detroit’s and Houston’s passing games were roughly equivalent, I think the edge has to go to Calvin.

    • Chase Stuart

      The natural response is:

      If Calvin Johnson is the better receiver, how come he was only targeted on 27.9% of Lions’ passes while Andre Johnson was targeted on 32.6% of Texans’ passes? (According to Footballguys, Calvin Johnson had 204 targets while Andre Johnson had 179.)

    • BigTex’sAshes

      While Calvin averaged more yards per reception Dre averaged more yards per route run (according to PFF). Not sure how much that would factor in to this.

      • Chase Stuart

        I would also add that while Megatron may have averaged more yards per reception, isn’t that offset by Andre having 31.6% of Houston’s receptions compared to 27.4% for Calvin Johnson in Detroit?

  • Boston Chris

    I think that I have problem w/ accepting as true the assumption that # of targets means a player is getting open more often. I think pfr had studies about how 1st and 2nd round QBs got more opportunities despite not showing any better football skills than their lower drafted brethren. How do we know there are not similar dynamics at play w/in the WR position. Higher drafted WRs get thrown to more often, b/c they are the first look on a particular play, b/c there’s an inherent prejudice towards them in the system for example. Or established WRs get more looks than they deserve, even if the up and comer is actually “performing” better on the other side of the field. I’m not saying there’s not something to the idea of more targets means better WR, I just think that there is a lot more to it than that as well. You often hear about various QBs who force WRs the balls. Indicating that they may not have been open.

    Just a lot to consider on that one, and wanted to point out that the assumption that more targets (or higher target %) should not be seen as prima facie meaning better WR.

    • Chase Stuart

      I would agree with this. There will certainly be exceptions. I think the key is that, on average, the assumption means (within a team) that more targets = better receiver.

      • Boston Chris

        I would agree that in most instances the most targets, within a team, gives us the rank order of best receivers on that team. However, once you bring different teams into the mix, then we’re talking about different systems of football, so now I’m not as comfortable saying that just how many more a WR gets than others on his team can be compared across teams (and different QBs/coaches/offensive systems).

        Definitely some value in the number, just hard to tease out across different systems….perhaps a preliminary look at this can be done by looking at WRs changing teams, although even that is difficult as their teaammates at WR don’t go w/ them. Hmmmm….

  • Richie

    The off-season is here.

    Booo. This is your worst post yet.

    • Richie

      football theory and history is more compelling to me than figuring out whether to rank the Lions ahead of the Bills.

      100% agree.

  • Richie

    It is interesting to see the similarities in passing stats between Houston and Detroit (and between Johnson & Johnson). But, didn’t those teams accumulate those stats in very different environments? I think Houston spent most of the 2012 season leading games, while Detroit was in a lot of close games and/or trailing. That would imply that much of Detroit’s passing game was either in desperate situations or garbage time.

    Of course, it could be argued that it would be more difficult for Calvin Johnson to perform well when the opposition knows the team is passing on every down, and it could be argued that much of Johnson’s gaudy numbers are also due to trailing in so many games.

    I used the play finder to look at how their passing offenses compared, looking only at first half stats in games with a scoring margin of +/- 14 points.

    Schaub: 174-266 65% 2,036 16-6 15sk 7.7y/a 99.1 rate
    Stafford: 186-314 59.2% 2,037 6-12 14sk 6.5y/a 68.9 rate

    So with the game still in question early, Schaub was much better than Stafford.

    Andre: 58-843-3 14.5 ypc (caught 73% of targets)
    Calvin: 52-785-1 15.1 ypc (caught 58% of targets)

    According to the play finder, 2 of Staffords pass attempts had no targets. He completed 58% of the passes he threw to Calvin and 60% he threw to all other receivers. So it would seem that Calvin’s lower catch percentage has more to do with the QB than the WR.

    • Chase Stuart

      Good stuff. I agree with the thinking, and I’d like to work Game Scripts into this at some point.

  • George

    Really cool idea to try and rank WR’s but I think you may have hit the nail on the head in that the ultimate WR ranking system does not exist because of the number of variables that you just cannot account for (e.g. game states in terms of score, score in relation to down, score in relation to down in relation to position on the field – any of the first three in relation to the coverage and then I suppose you have the quality of the coverage, weather – e.g. would Falcons, Texans or Lions receivers stats be a better than San Francisco, Seattle and Washington receivers stats when graded, type of offence – e.g. spread or pro style etc.). I think as you say it will come down to 4-5 rankings to provide an overall figure (think of it like the BCS – some will be bias free, some won’t).

    As a starting point perhaps there is an adjusted variable that you could maybe isolate out and define weights for (as there will be enough data), and then you could take player data in comparison to the league average (and then when you get unexpected weights dig further to find out why – e.g. playing on a bad team with no other talent so in double coverage all day, outdoors etc.). Develop a base line perhaps over a large enough data set over a number of years and then you could apply the base line to the current year. Maybe doing it by the team data may yield different weights than individual data (and may account for some of the variables e.g. such as the quality of the receiving talent). Good luck.

    • Chase Stuart

      Yes, we’re on the same page here. The first system I’ll probably use a revival of my old one. If nothing else, it’s a starting point, but there were some flaws to it.

  • Thad

    Have you considered looking at rec/ total incomplete passes or yards/total incomplete passes. A quick check shows Rice in 94, Hennigan in 61, and Chandler in 82 do quite well.

    • Chase Stuart

      What are you thinking, Thad?

  • Chase Stuart

    Chris from above was having problems with the comment feature, so he asked me to post this.

    Calvin saw a lot more long pass attempts than Andre did. They are a little more difficult to catch than shorter passes, which will affect a receiver’s catch %.

    Calvin, passes caught 15-24 yards from LoS: 33 (55 targets, 27% of all targets), 608 yards

    Andre, passes caught 15-24 yards from LoS: 11 (22 targets, 13% of all targets), 189 yards

    That’s a 419 yard difference, very close to the 366 receiving yard difference between the two players. They saw about the same % of pass attempts of length 25+ yards.

    Detroit’s #2 receiver, Pettigrew’s average pass length was 5.6 yards, Houston’s (Daniels) was 6.5 yards.

    Whether you measure yards per target, yards per reception, or yards per target by avg. teammate Calvin comes out on top over Andre. Not sure what else there is to measure.

    Quick chart I made graphing catch % by pass length distance, 2012 regular season:

    The peak at the right is a small sample size artifact. The sharp drop at the left at 0 yards represents batted passes. A smoothed estimate of catch % likelihood:

    <1	1	4	7	13	19	28	37+
    80%	0.75	0.7	0.65	0.6	0.5	0.4	0.3
    • Chase Stuart

      Thanks for running those numbers, Chris.

      I have a few comments to this, but let me start with this one. My goal is to understand where you’re coming from, so:

      Assuming Calvin Johnson was better than Andre Johnson, why do you think wasn’t Detroit’s passing game better than Houston’s?

      • Calvin’s supporting receiving cast was much worse then Andre’s. That’s the majority of the pass attempts for both teams.

        Since I wasn’t expecting that chart to get published, I should mention that the sharp drop to the left (roughly 10 on the x axis) is 0 yard pass length attempts, stuff to the left are screen passes behind the LoS, stuff to the right are beyond the LoS. The X axis labels aren’t yards past line of scrimmage. They are discrete pass length intervals.

        • Quarterback matters too. Demaryius Thomas and Eric Decker weren’t exactly fantasy stars before this year.

        • Chase Stuart

          I’m not sure I buy that about the supporting cast. I’d say Schaub and Stafford are pretty similar, and roughly equivalent to each other.

          LT Duane Brown and C Chris Myers are great players, but the Lions have a pretty good pass-blocking offensive line. According to Pro Football Focus, Gosder Cherilus graded out as very good, and Rob Sims, Dominic Raiola, Jeff Backus and Riley Reiff were all above average.

          I’d like to leave the running game aside for two reasons. One, I don’t think there’s much correlation between the success of a team’s running and passing games, but since that is often a contentious point, I don’t want to get too sidetracked. And two, by at least some measures, there wasn’t much of a difference between the two team’s running games despite the common perception.

          So then you get to supporting cast. Megatron had Pettigrew and Scheffler and Joique Bell and Titus Young and Ryan Broyles. Houston had Owen Daniels, Kevin Walter, James Casey and Arian Foster.

          I suppose I would give Houston the edge in supporting cast, but it’s close. I would not agree that it was much worse. Would you say Houston has the advantage at QB, OL, and receiving options outside of the #1 guy?

          Here’s the other point: If you think that Andre Johnson had a better supporting cast of receivers, why did he catch a higher percentage of his team’s yards?

          Houston did have a decent-sized edge in terms of ANY/A, so even if you argue that Houston had a better supporting cast, one could still argue that Andre Johnson’s year was better.

          • sn0mm1s

            I don’t think that success of a team’s running game has much correlation with the passing game either – however the respect given to the RB might. I didn’t get to see many Houston games (and really wasn’t paying attention to the defensive personnel in any case) but if Houston’s opponent considered Foster a threat they might be less likely to run a nickel or dime defense.

            Also, the coverage teams employed on CJ was quite a bit different than what AJ saw. I know I saw that punt gunner style coverage on CJ at least a dozen times this year. I remember seeing the Cards use it around midfield – not just the goal line. I don’t recall seeing any of that style coverage being used on AJ. I am also pretty sure Stafford said that in at least one of those cases he doesn’t even look CJ’s direction when that coverage is used.

            • Chase Stuart

              I think your points make some sense; unfortunately, I don’t think there’s any way to incorporate them into a study of receivers over several decades.

  • Thad

    I am thinking teams with 295 incomplete passes generally do not have good passing games

    • Chase Stuart

      That’s probably true.