SITE NEWS: We are moving all of our site and company news into a single blog for Sports-Reference.com. We'll tag all PFR content, so you can quickly and easily find the content you want.
Also, our existing PFR blog rss feed will be redirected to the new site's feed.
Pro-Football-Reference.com ยป Sports Reference
For more from Chase and Jason, check out their work at Football Perspective and The Big Lead.
ShareThisJosh Cribbs, Superstar, Part 2
On Monday and Tuesday, I looked at some of the best seasons by returners in league history. As noted yesterday, Josh Cribbs' 2007 season was truly remarkable. He added 486 more adjusted yards than the league average kickoff returner would have produced, and another 136 adjusted yards above what the league average punt returner would have compiled with the same number of returns. Those 622 adjusted yards over average was the single highest total by any one player since the merger.
year KR VAL PR VAL RET VAL Josh Cribbs 2007 486 136 622 Michael Lewis 2002 297 211 508 Dante Hall 2003 266 211 477 Mike Nelms 1981 347 123 470 Brian Mitchell 1994 242 181 423 Mel J. Gray 1991 247 175 422 Billy Johnson 1977 118 283 401 Mel J. Gray 1994 347 42 389 Billy Johnson 1975 76 312 388 MarTay Jenkins 2000 392 - 9 383 Terrence McGee 2005 373 0 373 Derrick Mason 2000 208 158 366 Tyrone Hughes 1993 178 182 360 Ron J. Brown 1985 359 0 359 Eddie Brown 1976 108 249 357 Brian Mitchell 2002 222 134 356 Eddie Drummond 2004 220 129 348 Raymond Clayborn 1977 340 0 340 Ron Smith 1973 155 173 328 Jerome Mathis 2005 356 - 30 325 Jerry Azumah 2003 325 0 325 Glyn Milburn 1995 260 61 321 Brian Mitchell 1995 227 89 316 Cecil Turner 1970 322 - 9 314 Tim Brown 1988 288 25 313
When comparing returners to returners, using adjusted yards (adjusted meaning simply giving an additional ten yards for every TD score) compared to league average works well as a measure of accomplishment and value. But what if we want to compare returners to QBs? Well, that's a little more complicated.
For example, here's a list of the top ten QBs in yards over average last year, using adjusted yards per attempt compared to the league average.
1560 Tom Brady 703 Peyton Manning 669 Tony Romo 650 David Garrard 626 Brett Favre 612 Ben Roethlisberger 474 Jeff Garcia 428 Donovan McNabb 413 Matt Hasselbeck 331 Jay Cutler
Guys like Derek Anderson, Drew Brees and Carson Palmer don't even make this list. And while I think this does a nice job of measuring the value added by each QB intraposition, it doesn't do a good job of showing the relative value of Brett Favre compared to Josh Cribbs. It doesn't do a good job for lots of reasons, but the simplest reason is because it basically says they were equally valuable last year.
The flaw in the math is that an average QB is actually pretty valuable. An average kickoff returner isn't really valuable at all. Why? Well, anyone can find a KR that can come close to the league average; there aren't league average QBs just hanging around.
So how do we compare the two? The league average adjusted yards per pass ratio last year for quarterbacks was 5.86. More importantly, 85% of the passes thrown last year were by QBs who averaged at least 4.75 AY/A. I feel pretty comfortable using that as a replacement level number. That's in between what Vince Young and Matt Moore averaged. I may be wrong, but I think almost every team can find a QB to average at least 4.75 AY/A, so if you're not averaging that, you really have almost no value. Only four teams had their QBs average under that -- Oakland, Carolina, St. Louis and San Francisco. In other words, I feel pretty comfortable with Miami's team QBs (4.89 AY/A) getting negligible value in this system.
Conversely, 85% of all kickoff returns last year were by kickoff returners who averaged at least 20.00 AY/KR, which is only 2.77 adjusted yards below league average. Additionally, about 85% of all punt returns were by players who averaged at least 6.15 adjusted yards per punt return, which was 3.14 adjusted yards below average.
What's that all mean? If you take 20.00 AY/KR and 6.15 AY/PR to be "replacement level", that means Josh Cribbs added 649 adjusted kickoff return yards above replacement and 231 adjusted punt return yards above replacement. So he brought in about 880 return yards above replacement level last year.
How does that compare to say, the adjusted yards over replacement level added by the top QBs?
2067 Tom Brady 1155 Peyton Manning 1126 Tony Romo 1095 Brett Favre 967 Ben Roethlisberger 906 Matt Hasselbeck 935 David Garrard 843 Donovan McNabb 741 Jay Cutler 761 Jeff Garcia 645 Drew Brees 675 Kurt Warner 627 Carson Palmer 597 Derek Anderson 553 Jon Kitna
Cribbs ranks ahead of all but seven QBs. Is it actually possible that Cribbs was more valuable to the Browns than Derek Anderson? My gut reaction, along with my second, third and fourth thoughts, tell me the answer is 'no'. But consider this -- our guts would also tell us that say, Brodie Croyle and Josh Cribbs would be way worse than Derek Anderson and a mediocre punt and kickoff returner, right?
Well, maybe not. Derek Anderson had 529 pass attempts last year, and totaled 3,787 yards, 29 TD and 19 INT. That's an average of 6.11 AY/A, or 3,222 adjusted yards. If we give our mediocre kickoff returner (let's call him Dominic Rhodes) 59 KO returns, we'd expect 1,180 adjusted yards. If we give our mediocre punt returner (let's call him Adam Jennings) 30 punt returns, we'd expect 186 adjusted yards. Therefore, our KR, PR and QB will give us 4,588 yards.
Now, let's look at Brodie Croyle and Josh Cribbs. We know Cribbs gave us 1829 adjusted yards on kickoff returns, and 415 adjusted yards on punt returns. If we give Brodie Croyle 529 pass attempts -- that's what Derek Anderson had last year -- then Croyle would end up with roughly 2,898 passing yards, 14.2 TDs and 14.2 INTs. Those numbers, of course, aren't very good. That's only 2,401 adjusted yards.
But Cribbs (the KR), Cribbs (the PR) and Croyle give us 4,759 adjusted yards on the season. And that's almost 200 more yards than Derek Anderson, Dom Rhodes and Adam Jennings gave us. Looking at it in that light, I can sort of begin to see how Cribbs was better, or more precisely, more valuable to the Browns than Derek Anderson. And that's not really a slight on Anderson -- Cribbs was more valuable than Donovan McNabb, Jay Cutler, Jeff Garcia and Drew Brees, too.
As mind boggling as that might seem, keep two things in mind. One, don't use Cribbs as a proxy for "the best kick returner in the league." Because this wouldn't be true, well, ever. Cribbs had by far, the single greatest season by a returner since the merger. So you might want to throw what you normally think of great returners out the window. Two, we should understand that Cribbs' season was an anomaly, and he had some luck, and he's obviously not "that" good. But it did happen and he was valuable. Tom Brady's not going to throw for 50 TD and 9 INT ever again, but when valuing his season, we need to ignore the anomalous nature of it.
This got pretty lengthy, so I'm going to table any discussions of kickoff returners and correlation coefficients for another day. There should be more than enough to chew on, here. If you're a Chiefs fan, it's easy to think that Kansas City would have been a whole lot better if they had Jay Cutler or Donovan McNabb instead of what they actually had at QB last year. But KC might have been better off if they simply had the Cleveland special teams unit instead.
This entry was posted on Wednesday, June 11th, 2008 at 6:20 am and is filed under General. You can follow any responses to this entry through the RSS 2.0 feed. Both comments and pings are currently closed.
2 things about returner yards : they're mostly on the own end of the field, & they don't affect the down count.
Figuring the relative values for types of yardage would take either a humongous data set or monte carlo run.
Another thing when comparing return yards to QB/RB/WR yds--total number of touches. A RB who averages 4 ypc times 300 carries has a 1,200 yard season. A backup RB who has 80 carries (5 per game), and averages 4 ypc while running in garbage time, 3rd & longs, and a few "regular" carries mixed in, is not nearly as valuable. When comparing a returner to a QB or anyone else, this MUST be taken into effect. While I think Chase's metric does a decent job of that, it is extremely difficult to measure.
I also agree with #1 in that the first about 15 KR yds. are free (although punt return yds. aren't). No other position gets those "free" yards. (All players get those "free" yards at the end of a long TD run/pass/return where speed wins out over technique.)
Not to take anything away from Chase or this blog, but football outsiders stat of VOA and DVOA (defense adjusted) is a very similar metric. For returns, it also takes into account the field position after a return (e.g., a 30 yd return that starts 5 yds deep in the end zone is OBVIOUSLY less valuable than a 30 yd return that starts at the 10 yd line).
Completely off-topic comment:
Doug, I tried listening to your interview on the Footballguys' podcast, but I had to give up after about 15 minutes because I just couldn't hear you. Too bad your call recorded at such a low volume.
Hey Downpuppy,
I don't think it really matters that the returns are mostly on their own end of the field, when you're comparing a returner to another returner. So if a replacement returner gets 20 AY/KR, and Cribbs got 31, that's still eleven more yards. It doesn't matter too much if it brings you from the 24 to the 35, or the 20 to the 31. I agree that yardage inside the ten yard line is the most important, but the majority of QB yards don't really come there, either.
It's definitely a herculean task, but I'll play the approximate card here.
Joseph,
I think my method does take it into affect. For example, a replacement RB might be three yards per carry. That's why I've long been a fan of the stat, rushing yards over 3.0 YPC. (There was a lengthy post about that here: http://www.pro-football-reference.com/blog/?p=361).
So a RB that goes 80/320 brings only 80 yards over replacement, assuming 3.0 equals replacement. Your 300 carry RB brings 300 yards over replacement, making him nearly four times more valuable than the other guy. Similarly, I don't think the "free yards" is much of an issue, since we're giving the replacement level returner the same amount of free yards.
I agree with your example about a kickoff return at the 10 or five yards deep in a player's own end zone. Unfortunately, for now, that's going to have to be treated like a QB's pass that bounces off his WRs hands and into the opponent's -- we're going to have to use some arbitrary numbers, shrug our shoulders, and hope things even out over time.
Re 2: It's a little confusing, but I think the initial free yards enjoyed by a returner are accounted for in the analysis. The yards-above-replacement takes care of that effect.
A couple other thoughts:
1. Kick and punt returns are very chaotic. I'd bet randomness is far more of a factor than with passing. Cribbs is obviously gifted, but a great deal of his accomplishment last year was likely due to luck.
Along those same lines, returners have far fewer reps, so a just a couple long returns can skew the data heavily.
2. Total yards above replacement might not make sense here. KR attempts are based on how many scores his defense gives up. QBs are far more complicated. A QB with a good running game and a good defense doesn't have to throw very often. On the other hand, a good QB creates his own attempts by getting first downs.
Also, Derek Anderson may not be all that he's made out to be. His NFL QB Rating (as flawed as it is) was only slightly above average. He had 2 real big games (vs. CIN and SEA) but was otherwise noting special. Definitely better than expected, however.
DPAR says that the Browns' return game was worth roughly the same as Jon Kitna last season, and about two thirds as much as Derek Anderson. I'm guessing that stat is somewhat depressed by the inclusion of returns by players other than Cribbs, but I'm guessing not that much (and the unit does get credit for teams squibbing away from him). That's something I'm more comfortable with - the idea that a full season of historically awesome return play might be as valuable as a league average starting quarterback (which Kitna pretty much was - the only such player to start a full season in 2007) than the notion that it might be as valuable as an actually good one.
On a side-note, I'm still not convinced that Anderson is really that good. As noted in the entry in question, he didn't come out well of this site's recent study into adjusted quarterback performance. Throw in stellar play from his supporting cast, and considerable luck in how few injuries Thomas, Edwards and the rest suffered, and I think quite a few people will be very disappointed with Anderson's 2008. The Browns should have sold high, if that was indeed an option, and retained Anderson at the top RFA tender if it wasn't.
Then again, I think it's safe to assume that Cribbs benefited from some excellent blocking on those return units.
Also I'm pretty sure that research has been done, by FO and others, on the relative value of yardage in different areas of the field. I can't find it, though, either because it was in a PFP or because the search function on their site sucks.
Return yards are very different in value from passing or rushing yards.
Adding two yards to a kick return is pretty pointless. Adding two yards to a dumpoff reception on 3rd and 5 is huge.
The down count is tremendously important, and no return man can ever affect it. Only offensive players can. Josh Cribbs isn't worth a third of what Jay Cutler is worth.
Yaguar,
Sure, adding two yards to a kick return is pretty pointless. But adding 11 yards to 59 kick returns becomes really important. The down count is important, but your team still fields a QB. It's not like we're playing Josh Cribbs at QB. So the question really isn't Josh Cribbs vs. Jay Cutler, its Josh Cribbs and whatever you can find to play QB, vs. Jay Cutler and whatever you can find to play QB. That might turn into something like Josh Cribbs and Brodie Croyle vs. Jay Cutler and Dominic Rhodes.
I agree that Josh Cribbs, 2008, won't be worth a third of what Jay Cutler will be worth. I don't agree that Josh Cribbs, 2007, was anywhere near only one-third as valuable as Jay Cutler, 2007 (ignoring, for simplicity's sake, that Cutler had the hardest schedule int he league last year.)
I would gladly take Cutler and Rhodes. Think about all the drives Brodie Croyle stalls. Josh Cribbs can't do anything about that. If Cutler is worth an extra first down every couple of drives, which he almost certainly is, that's already enough to be more important than Cribbs.
It's a good point, Yaguar. I've had some in person discussions with others, too, and I think there needs to be some tweaking to account for the value of keeping drives alive.
Surely DPAR already does that? If we buy the notion that QB DPAR does a pretty good job of reflecting the value of a QB's play to his team - and I think, unlike receiver DPAR, it pretty much does - then we're left with Kitna as a value comparison. That's still an astonishing amount of value for a returner/return unit.
I was thinking... Something I've noticed here is that y'all turn points into yards. A touchdown is +10 yards, an interception is -45 yards, and so on. Then you use yards to compare players. I understand it, but it seems a little strange now that I think about it. What we're concerned about is really points on the scoreboard, yes? So wouldn't it make more sense to think in terms of virtual points rather than virtual yards? Isn't that what we want to know? Does Cribbs + Schmo result in more points on the scoreboard than Schmo + Cutler?
So one can consider this in sort of an "average drive" frame of mind. If you think of the drive as starting from the kickoff return, then you've got X return yards and Y yards from your offense, and the total drive is X+Y. If the field was infinitely long, you could compare them reasoanbly directly -- Cribbs adds A yards to a drive compared to an average returner, Cutler adds B yards to a drive compared to an average QB. For a non-scoring drive, you can probably compare them like that too. Since no points were scored, the yards at the beginning and end of the drive are essentially the same.
But with drives into the red zone, it gets weird because red zone yards are different than the others. The post you linked to the other day suggests this too.
With a kick return, those yards are essentially free. The returner has already passed everybody on defense by that point. If our field were infinitely long, there's no reason he couldn't run 2 more miles.
One might make the assumption the "red zone" for the kick returner is probably the 20 to the 50 yard line (or some such -- I've never really paid attention that closely), and nearly any return past the 50 is a touchdown. He should get credit for the touchdown, but the extra 50 yards are negligible. So in a sense, a kick returner using your system is getting 60 bonus yards for a touchdown instead of 10.
It seems to me that if you really want to compare the value of kickoff returns, you'd want to have a function that describes how many points you score from a given position on 1st down. So if you want to compare three returns to midfield with a touchdown return and two touchbacks, you'd get something like
7+f(80)*2 vs f(50)*3.
Heck, I don't know. It's easy to dream up ideas, not so easy to research and implement them.