Innovative Statistics, Intelligent Analysis
Click here to buy PDF version.
Click here to buy PDF version
Like our page on Facebook and get Football Outsiders links directly in your Facebook feed.
Official Account: @fboutsiders
Scott Kacsmar: @FO_ScottKacsmar
Ben Muth: @FO_WordofMuth
Aaron Schatz: @FO_ASchatz
Vincent Verhei: @FO_VVerhei
-- plus --
Ian Boyd: @Ian_A_Boyd
Bill Connelly: @SBN_BillC
Cian Fahey: @Cianaf
Brian Fremeau: @bcfremeau
Tom Gower: @ThomasGower
Bryan Knowles: @BryKno
Rivers McCown: @RiversMcCown
Chad Peltier: @CGPeltier
Andrew Potter: @BigHairyAndy
Rob Weintraub: @robwein
Carl Yedor: @CarlYedor61
07 Dec 2010
I wrote this Monday afternoon knowing full well that it would cause Brady to have a four-touchdown game. So you can thank me, Patriots fans.
Posted by: Mike Tanier on 07 Dec 2010
53 comments, Last at
09 Dec 2010, 8:25am by
I'm as big a Tuck fan as any, and agree about how well he's played this year, but... fumbles RECOVERED?
what's wrong with fumbles recovered? recovering fumbles provides value to your team. the MVP award is descriptive, not predictive.
It's the same problem as using Runs as a measure for MVP in baseball. Players on defenses that cause a lot of fumbles are going to have more fumble recoveries than players on teams that don't.
Vick and Brady aren't benefiting from their teammates?
Actually runs are a better indicator than you would think in basbeall. At the very minimum, it means the guy is getting on base a lot, which is the most important thing for a hitter. Granted there are better advanced metrics that measure offensive performance better than runs, but as far as the base stats go, runs is pretty good one, certainly better than RBI or BA.
the MVP award is descriptive, not predictive.
This point nullifies most of the arguments below.
Thank you, Mike Tanier.
And a belated happy birthday, too.
"So the next time someone suggests Vick or Brady for MVP, tell them that you think Hillis, Tuck, or Mangold deserves serious consideration. You won’t win any arguments, but you will definitely stand out from the crowd."
I think that last line is known as the motto of the "columnists aren't paid to be correct, consistent, logical or reasonable, just provocative" faction. Its kind of funny that Football Outsiders was started because of this kind of nonsense and now its just another source of it. Pure filler that isn't even really sincere:
From "Football Outsiders Basics"
"fumble recovery is not a skill" and "Pouncing on the ball as it is bouncing all over the place is not a skill. ...The odds of recovery are based solely on the type of play involved, not the teams or any of their players."
Next week how teams need to run the ball to win.
There are lots of random events in football. They're still good if they happen for your team.
Let's show some love for team captains who win overtime coin tosses!
I'm pretty sure he wrote this for NBC Sports. Yep, it's got an NBC logo on it. It's on the msnbc web site. What's that at the bottom say?
"Mike Tanier writes for NBCSports.com and Rotoworld.com and is a senior writer for Football Outsiders."
Ah. So, perhaps he's allowed to have an opinion outside of Aaron's, informed by his own experiences. That's allowed here, right? Different opinions on the value of certain statistics? I'm pretty sure Football Outsiders was started to hear alternative opinions, too. So...Tanier's wrong for not towing the line and treating FO alone as his gospel? When has Tanier ever towed the damn line? He's one of the more interesting writers here because he's regularly contrarian to the popular wisdom (and often right). I don't think fumbles recovered is the most relevant statistic, but I'm not going to say someone who disagrees is flat out wrong. It does offer some value, and it shows that Tuck is right there in the backfield with Umenyiora to finish up on the play when the latter makes some of those forced fumbles.
Except... does he really disagree with the notion that fumble recoveries are largely luck? This is one of the most cut & dry FO fundamentals, and with a pretty good bit of research behind it. It's one of FO's defining conclusions, and cutting against it requires better explanation.
I know that it's considered dogma at FO that "fumble recoveries are mostly luck", but it's really not a theory that stands up very well. For starters, what does it mean to be "caused by luck"? Well, strictly speaking it means that you have some kind of (pseudo-) random number generating system that produces results purely by luck. Like a deck of cards, a slot machine, a pair of dice, or a roulette wheel.
But that's not what happens with a fumble recovery. A fumble recovery is caused by being alert as to where the ball is an being able to get control of it. See, for example, the fumble last night where Sanchez coughed the ball directly to a Pats' rookie D-lineman, who managed to not get control of the ball.
You can say that "we can model fumble recovery as a function of luck", but so what? We can model baseball hitting as a function of luck, but that doesn't mean that it is a function of luck. It is a function of the skill of being able to put the bat on the ball. The persistence of statistics is not the same thing as "luck".
But, but! says FO, we cannot dismiss luck as the causal factor statistically! But in this argument, the FO head honchos are misunderstanding the purpose of a null hypothesis. "Happening by chance" is a null hypothesis and its purpose is to be used in scientific studies where the data can show that the null hypothesis is rejected. You're not supposed to accept the null hypothesis!
Philosophically, you cannot say "well, we have modeled this physical system according to such and such a model, and found that the models we've tested have not outperformed randomness" and the conclude that randomness is itself the explanation!
But it's worse than that. Whether a fumble is recovered or not has been shown to be related to what kind of play was being run. This site
says that 55% of fumbles committed on rushing plays are recovered by the defense, while 47% of those caused on passing plays are recovered by the defense. Curiously, fumbles caused by strip-sacks are lost by the fumbling team 47% of the time and fumbles after completions are also lost 47% of the time.
So, whatever "pure luck" is involved here, it seems like it's situational luck. Now I'm all in favor of using conditional probabilities when necessary, but the more complicated you have to make your random model, the less reasonable it is to say "it's all luck". And the 8% disparity among fumble recovery rates suggests that area of the field matters, as does type of play. So how does one allow those variables into the model, but reject pre-emptively the possibility that player skill is also a factor?
Since the first time I've heard this theory, I've mocked it. Saying "it happens by chance" is not an explanation, it's the opposite of an explanation. It's a punt, an admission that you don't know exactly what is involved.
I believe the meaning of "fumble recoveries are luck" on this site is simply that past history for players or teams is not correlated with future fumble recovery performance. Nothing more, and nothing regarding cause and effect. So if player X has a lot of fumble recoveries this year, that's valuable and it COULD be that he is particularly skilled at recovering. But based on lots of history, that doesn't help you predict how many recoveries he'll have in coming years.
In a case like this, saying that something is not predictive (i.e. luck) doesn't mean there's no skill involved.
Soooo ... many paragraphs that seem to boil down to "fumble recoveries are NOT mostly luck." Which is interesting, given that a) FO never said they were in the first place and b) you don't show any more proof to support your side than what I think you believe FO shows to support theirs.
This is FO's posted position on fumbles: There is no correlation whatsoever between the percentage of fumbles recovered by a team in one year and the percentage they recover in the next year. The odds of recovery are based solely on the type of play involved, not the teams or any of their players.
Farther down: Fumble recovery is a major reason why the general public overestimates or underestimates certain teams. Fumbles are huge, turning-point plays that dramatically impact wins and losses in the past, while fumble recovery percentage says absolutely nothing about a team's chances of winning games in the future. With this in mind, Football Outsiders stats treat all fumbles as equal, penalizing them based on the likelihood of each type of fumble (run, pass, sack, etc.) being recovered by the defense.
Other plays that qualify as "non-predictive events" include blocked kicks and touchdowns during turnover returns. These plays are not "lucky," per se, [emphasis added] but they have no value whatsoever for predicting future performance.
Congratulations. You're mocking something that FO didn't say. Your "argument" is almost exactly what they did say. You win the internets.
You know, it's hilarious that you cite FO-basics without citing the heading for the section that you are quoting:
Recovery of a fumble, despite being the product of hard work, is almost entirely random
Talk about intellectual dishonesty!!! Why are you wasting my time with this crap?
From Methods To Our Madness:
A fumble is worth anywhere from -1.70 to -3.98 [success] points depending on how often a fumble in that situation is lost to the defense -- no matter who actually recovers the fumble.
FO agrees with Burke that fumbles in different situations yield different recovery rates (hence the different success values). The other point that FO makes is that a team's past fumble recovery rates don't predict future rates. To that extent, it's random or luck or whatever you want to call it.
Trying way to hard, and ignoring empirical evidence and rational analysis. Other than that, a perfectly reasonable opinion.
What he said was exactly correct.
Unfortunately, its exactly what FO feels.
And back to the point of the article: The fact that Tuck has recovered four fumbles so far this season is in fact valuable to the Giants. It's non-predictive, and it shouldn't be used to claim that Tuck has some special fumble-recovering skill - but it is nonetheless valuable.
I'm sorry that "math is hard", JonFrum. I provided a "rational analysis" which you dismiss as "trying way to(sic) hard". And the point of my argument is that there is no such thing as "empirical evidence" that a process is "random".
I would hazard a guess that most of the posters here don't know just how flexible the concept of "random" is in a mathematical setting. You cannot prove that something "is random" since there literally is no way of finding counter-evidence. Any kind of data set you look at could have been generated by a "random distribution", since "random" is a sufficiently pliable concept that it includes constant functions, exponential functions, etc., etc.
That's why the way actual scientists use the concept of randomness is to demonstrate that the data presented does not fit a given (usually simple) random model. Data that is not random is what is of interest.
I don't think the problem is writing a different kind of column for a different audience. People are asking whether he's following one principle here, and ignoring it elsewhere. No reason why he can't disagree with another writer here. The question is - is he disagreeing with himself (writing here)?
Which is probably why Tanier wrote it for another website, and not this one. It's worth an XP simply because it's Tanier, and we want to read what he has to write.
If it had been posted on FO, rather than linked too, your critique would be accurate.
I suppose an article like this is intended to put attention on positions other than QBs, but I can't help and wonder why Matt Ryan is on the list and Rodgers and Rivers aren't.
I suppose in Rivers' case you can look at poor team record and say that an MVP should be playing for a playoff contender, but Rodgers has moderate leads over Ryan in both DYAR and DVOA, even before considering that he's a very good running QB.
Still looking at DYAR, Vick shouldn't even be in the discussion, at least unless he can maintain his level of play and stay healthy for every remaining game. He just hasn't played enough to make up for the fact that his passing DVOA is basically identical to Rodgers and Rivers, as well as that Kolb has actually been pretty good (15.1% DVOA), so it's ridiculous to say that the Eagles have no passing game without Vick.
Truly, going striclty by most VALUABLE player, Vick would not be anywhere near the top. Take away the other candidates from their teams, and the teams become mediocre to terrible. Take away Vick and the Eagles are still very good.
Another dark horse, even if he is a QB: Sam Bradford.
The Rams without Bradford might be worse than the Browns sans Hillis. He has completely and to the extent possible in football, singlehandedly, turned around the franchise.
But he should get ROY, and Spagnulo should get COY, and you won't even hear so much as a medium-sized "eh" from the folks in St. Louis.
"he Browns sans Hillis. He has completely and to the extent possible in football, singlehandedly, turned around the franchise.
The Browns sans Hillis? Weren't we just saying last year that Jerome Harrison was the real thing? Didn't he get cut? Or traded for a bag of footballs or something? Isn't Cribbs basically giving you the same YPC?
Hillis isn't that valuable. I'm not even sure hes thats good.
Joe Thomas is good. Joe Thomas is valuable.
Harrison was traded to the Eagles, where he had 100 yards in the blowout against Washington, all of which came after the game was esentially decided.
I'm tired of seeing the coach-of-the-moment who took a franchise with years of bad records, was handed a bunch of high draft picks, and manages to turn them into a middling team being rewarded with "Coach of the Year". I'm not knocking Spagnuolo, but Belichick has done a fantastic job this year building a defense out of nothing while managing the departure of Randy Moss, among other teams, and has created a 10-2 team.
It's a lot harder to build a 10-2 team than a 6-6 team. Detroit Lions notwithstanding, any franchise given a few years of high draft picks and moderately competent management ought to improve to a middle-tier team.
Bradford can be offensive ROY, but he certainly hasn't "singlehandedly" turned around the Rams. He doesn't even play defense!
(BTW, you cannot simultaneously claim that Bradford has turned around the team "single-handedly" while claiming that Spagnuolo should be the CoY. Well, I mean, sure, if you were a lawyer there would be no proscription against proposing mutually contradictory theories, but in terms of consistency, if Bradford is solely responsible for the Rams' improvement, than Spagnulo doesn't deserve CoY. All he's done is play Bradford!)
Take a look at the Rams roster from this year and the Rams roster from last year.
A few role players, and Sam Bradford. That's it.
Even then, I still put the caveat "to the extent possible in football" in my statement. Of course, you'd rather ignore that so that you can have some good, cathartic internet outrage.
"you cannot simultaneously claim that Bradford has turned around the team "single-handedly" while claiming that Spagnuolo should be the CoY"
So you're suggesting that a coach and a QB can't both do a great job in the same year? Whatever.
They are mutually exclusive given the words you used. If Bradford has single-handedly turned the Rams around, then Spagnolou deserves no credit. If Spags is deserving of COY, then Bradford hasn't single-handedly turned the team around.
And yes, let's stop with the "coach with the biggest turnaround" as COY. There are more deserving coaches for the award.
"given the words you used. "
Except that you're not looking at the words I used.
So in 2007, did Tom Brady not deserve the MVP or did Belichick not deserve COY?
What about 1999? Who didn't deserve their award? Kurt Warner? Or Dick Vermeil?
The word you used which is sparking this debate is "singlehandedly", which implies that Bradford alone is responsible for all the Rams' success this year. If that were true, Spagnolo wouldn't deserve the COY.
Except that I didn't use a word.
I used a phrase. Go back and read the post. Not RickD's. Mine. You'll see a phrase, not a word. The phrase is "to the extent possible in football, singlehandedly."
I specifically put that qualifier in to give credit to people on this site for having some intelligence. Apparently I overestimated.
Well, to me, the qualifier "to the extent possible in football" attached to "singlehandedly" doesn't mean any more than the acknowledgement that there are actually 10 bodies on the field wearing St. Louis uniforms blocking, running, and catching passes. "Singlehandedly", even with your qualifier, certainly doesn't co-exist with the belief that he has the best coach in the NFL this year on his sideline.
Anyway, I was trying to harmonize an argument, I could care less. The Rams appear to have a great shot at the playoffs instead of drafting in the top 5. You are entitled to enjoy that fact with unbridled enthusiasm and passion.
Let me get this straight: your contention is that you used the qualifier "to the extent possible in football" so that means that your usage of the word "singlehandedly" is acceptable?
And you're calling my intelligence into question?
I'm curious, do you use phrases like "very unique"? "Mostly virginal"?
Don't blame me for your bad writing.
The Bradford hype is starting to feel almost Sanchezian. He's promising, and the fans haven't had a lot else to cheer in recent times — but let's not get carried away.
Could the Rams have gotten to 6-6 in the NFC West with Jon Kitna at quarterback? Shaun Hill? Maybe they could.
What has really turned around the Rams' franchise is not having to start Keith Null.
"Could the Rams have gotten to 6-6 in the NFC West with Jon Kitna at quarterback? Shaun Hill? Maybe they could."
No. They could not.
I understand being skeptical, but this is not Reggie Bush. This is the rare instance where the player is earning the hype.
In this years NFC West, the Rams absolutely could be 6-6, and even win the division, with Jon Kitna or Shaun Hill. They'd have an outside shot with Jon Heder or Faith Hill.
In fact, remove the Cardinals from the conversation, and the other three teams could play musical quarterbacks and likely be right where they are now.
Is Bradford a better QB than the Smith brothers in San Francisco? It sure looks like it. But I don't think the Rams are also-rans in the division without him. Chris Long and the rest of the defense is playing much better. Mark Clayton was and Danny Amendola is playing much better. The offensive line seems to be playing much better, so it's not as simple as finding the new guy on the roster and saying he's responsible for their improvement.
If Bradford hadn't been ready to play, or if one of the myriad of hits he took had re-injured his shoulder, this team is down there with Buffalo.
Mark Clayton is your rationale? He wasn't even on the roster at the start of the season, never started a game for them, and only lasted 5 games before he got injured. Danny Amendola is the exact same player he was last year, only this time he has a QB who can get him the ball. Amendolas numbers are better because the QB play is that much better.
"it's not as simple as finding the new guy on the roster and saying he's responsible for their improvement" - except that sometimes, like here, it really is that simple. It's easy to try to be too smart for the room and say that the obvious can't be right because it's too obvious. But sometimes, like here, the obvious really is true.
There's one little problem with all of this: The Rams have a terrible offense. St. Louis is a .500 team in the weakest division I've ever seen in my life, and they've gotten those six wins by beating other bad teams, usually by close margins. They've scored more than 20 points twice all season, against the Redskins (25th ranked defense in the league by DVOA, 23rd by points) and Broncos (32nd by DVOA, 29th by points). DVOA ranks them 28th overall, and going by points scored gives you a comparatively generous 25th. If you want a reason for their improved win-loss record, look at their 19.7 points allowed per game, +5 takeaway differential (partly due to the offense, yes - minimizing turnovers is Bradford's strong point thus far), and the 24-48 record of the teams they've beaten, including a grand total of zero who are better than even at this point.
The only thing separating this team from Buffalo is that Buffalo plays in a division that features actual football teams.
DVOA says the Rams' pass offense is slightly worse than the Bills', so they probably are "down there with Buffalo". However, on defense St Louis has a better ranking than the Bills' — and is much improved from 2009. I haven't seen enough of their games to pass judgement, but the stats suggest the reasons for the turnaround (to .500) are:
1) defense now playing at a respectable level (credits: Spagnuolo, Long, Laurinatis, run defense generally);
2) easy schedule;
3) pass offense now 'poor' rather than 'historically awful' (credit: Bradford).
I actually went to have a look at Bradford's numbers to see the difference. 300 fewer yards, 4 fewer touchdowns, 3 more picks, 3% lower completion percentage.
Sorry, compared to Ryan.
I just don't get this article. Matt Ryan's value exceeds his stats because he isn't used much... but Peyton Hillis' value is high because he is used a lot? Furthermore, if we're talking about players who are valuable because their backups stink and their teams would be screwed without them, is it even worth mentioning any player besides Peyton Manning?
I guess this is water-cooler-discussion material, but come on, Mike, you're better than this.
They're not mutually exclusive arguments. To argue for the less used player, you would argue that they were used mostly in high-leverage situations. Classic example was how Casey Stengel used to save Whitey Ford to pitch him against the best teams in the league, and let his middling pitchers beat the cellar dwellers. To argue for a more-used player, you do what Tanier did with Hillis or with what baseball writers do when they vote for a position player over a starting pitcher for MVP.
As for the article itself, I get tired of the intellectual exercise of "watch me make the MVP case for player X!" Please! Nobody really thinks Peyton Hillis is the MVP. The MVP of the Browns? Arguable - like many I would go with Joe Thomas. But of the entire league?
This from an FO writer?
Is Peyton Hillis "more valuable" than, say, Arian Foster? "How many wins would the Browns have without Hillis?" How many wins would the Texans have without Arian Foster?
Let's keep it simple: the MVP should be the player performing at the highest level and having the greatest impact over the course of the season. And right now, that's Tom Brady. His team isn't 5-7 because...he's never been on a 5-7 team!
Sorry if that's "boring".
So you are mortally offended by an article who's conclusion is that these ideas are meant to spark conversation? It doesn't even make the claim that these players are more valuable than Brady or Vick, simply that it's an interesting thought experiment. He labels them as "dark horse candidates", "part of the debate" and "not ideal". Seriously, dude, chill out. It's just an interesting idea.
"Mortally offended"? Huh?
No, I'm tired of people advancing arguments "to spark conversation". I expect that from the crew on NFL Live, or Sunday Countdown. That's not what I'm looking for at FO.
Technically this article isn't on FO.
I know it's hip to always point out someone other than the mass media but Mathews is more than a pass rusher. The guy blows up the run, covers when asked and has been the anchor of a good defense that has lost half its starters along with coping with a Charles Woodson who isn't CHARLES WOODSON this season.
The guy is a whirling dervish out there. CMIII deserves the attention even if it involves the likes of Joe Buck.
Does momentum exist in college football? It sure seems that way for the Louisville Cardinals.
See All XP | NFL XP | College XP
© Football Outsiders, Inc. // Site powered by Stein-Wein // Partner of USA TODAY Sports Digital Properties