Football Outsiders (& fumble recovery, Champ Bailey, etc.)

Cafe Society
1

This post is a reply to some comments made in this thread. They concern the website (and published works of) Football Outsiders (FO), which does advanced football statistics and analysis – I’ve been pimping them here and elsewhere for a few years now. I missed the linked thread the first time around, and found it by accident tonight when I ran a search for my screen name. Since the original thread is a few months old and this post is quite long, I decided to start a new thread rather than resurrect the old one.

Aside from the immediate topics of this thread, if anyone has any comments, questions, or criticisms of FO, I’d be interested in addressing them.
Ok, on with the show:

They really don’t. They make lots of predictions for each team, as well as hundreds or fantasy football projections every year, so of course you’ll be able to point to instances in which they were way off. OTOH, there also plenty of times that they predicted seemingly unlikely things to occur, and they did, for exactly the reasons they gave.

For instance, after 2003, when the San Diego Chargers were coming off a miserable four year stretch in which they had a record of 18-46 and a lame duck QB, FO’s projection system said that the Chargers would completely turn it around in 2004 and have one of the best offenses in the league (as well as an improved defense). I think it’s safe to say that they were the only ones making such a prediction, and they were absolutely right. Their system noticed that San Diego’s 3rd down performance was way out of line with their 1st and 2nd down performance, and that 3rd down performance tends to regress to a team’s mean, which has a disproportianately large impact on their overall success.

Or consider the 2005 Atlanta Falcons. Halfway through the season they were 6-2 and tied for the best record in the NFC, yet FO had them ranked as the 17th best team in the league. They caught a ton of flack for it, including a major flame war in the discussion threads brought on by some visiting Falcons fans. But they were absolutely correct – their stats (DVOA in particular) noticed that Atlanta had faced weak competition and had had great luck recovering fumbles (they had recovered 15 out of 19 fumbles in their games – more on fumble recovery later, btw). Over the second half of 2005, the Falcons went 2-6 – their level of play remained the same (at the end of the year they were still ranked 17th in DVOA), but they faced a tougher schedule and, more importantly, their fumble recovery luck was completely reversed: they recovered just 5 out of 18.

Other noticeable predictions, all from this year: Oakland would have one of the best defenses in the league (they improved from 23rd to 9th), Philly would bounce back, and Minnesota would get a lot worse. To be fair, they also predicted that Seattle would be the best team in the league – I don’t think anyone knows why they regressed so badly (even when healthy they played mediocre football).

First, the article you refer to isn’t like the official canon of Football Outsiders – it’s just a guest column by a reader who used their numbers to do a lot of research and draw his own concluions. Second, FO does now differentiate between run support and pass defense plays in their defensive player stats. Third, FO is always quick to point out that a lot of their numbers are in their infancy, with much room for improvement. Finally, even the writer of the guest column in question isn’t taking those numbers as gospel: “. . . The second possibility is that there is important information that is being missed by the PFP numbers. And until we can explain what is going on with Kelly “Lockdown” Herndon, we can’t accept Bailey’s PFP numbers as being a truly accurate reflection of his play.”

KC Joyner puts out his own book and writes for ESPN.com, so don’t hold his flaws against FO.

Come on, the article gives a perfectly good explanation of why you might discount Bailey’s contribution to Denver’s good pass defense: “Then again, if Bailey was so effective, why doesn’t it come through more in Denver’s DVOA numbers? Yes, Denver was fourth in the league in passing DVOA, but a closer inspection of the numbers provided in Pro Football Prospectus 2005 shows that Denver’s high ranking was almost entirely due to its tremendous ability to cover the tight end (-43.6%, 2nd) and running backs (-48.8%, 2nd). Their DVOA versus #1 receivers was only 7.4%, a substantial improvement over the year before Bailey’s arrival, but still below average. Denver’s DVOA against #2 receivers actually got significantly worse in 2004, dropping from –9.1% to 14.4%, which both highlights the strangeness of Herndon’s PFP numbers and throws into doubt the notion that Bailey’s presence allowed the team to roll coverage effectively to the other side of the field.”

Sadly, no. FO’s website needs a major upgrade. They are at least taking a step in the right direction: some sortable stats.

Their numbers from their latest book suggest that, last year, Shawn Springs was the best CB, while Adrian Wilson was the best Safety (by a lot). They also indicate that Champ Bailey was excellent.

Ellis Dee, on Fumble Recoveries

The thing that some people mis-understand about FO’s stance on fumble recoveries is that it’s not based on the overall 50% recovery rate or the fact that there’s a lot of variance in the number from one year to the next. It’s based on correlation coefficients, which consistently show that there is no correlation between a team’s rate of successful recovery one year and that same success the next. Likewise, there’s no correlation between a team’s recovery rate in the first and second halves of the same year.

Just about any other football stat, despite plenty of variance, will show a meaningful positive correlation from one year to the next – teams that intercept lots of passes (or to cause or avoid lots fumbles) in Year X will tend to do the same in Year X+1, in spite of lots of inevitable exceptions. OTOH, over the past few years there has actually been a small negative correlation between inter and intra-year fumble recovery rates. It seems silly that good fumble recovery teams are likely to turn into bad recovery teams in the future – more likely there’s just no connection at all.

Two things: First, you have to seperate fumbles from fumble recoveries. Causing and avoiding fumbles are repeatable skills that show postitive correlations from one year to the next. Once the ball is on the ground, however, the numbers indicate that no NFL team is any better at recovering it than any other.

Second, a fumble recovery is indeed the product of hard work – teams presumably run loose-ball drills and instruct their players to use this or that technique to recover the ball – but it is still a non-predictive event. A team’s success (or lack of success) at recovery over one period of time tells us nothing about the likelihood of that team’s success over a subsequent period of time.

FO accounts for a fumble’s type and location on the field when measuring the likelihood of recovery. Their position on fumble recovery has absolutely nothing to do with the fact that overall recovery rates are about 50/50 between offense and defense. If every one of an offense’s fumbles occured in the secondary, and that team lost 70% of its fumbles, FO would say that they’ve had normal recovery luck, since that’s about how often fumbles downfield are recovered by the defense.

Other “non-predictive events” include blocked kicks, as well as fumble and interception returns. In these cases, it’s not that no team is better than any other, it’s that the events are so rare and, in the case of turnover returns, so dependent on circumstances beyond the team’s control. Because of these factors, whatever differentiation in skill that exists is unable to break through all the random noise, and is thus useless for making predictions about the future.

2

First off, is there a positive correlation in FG kicking %? I read an involved Wall Street Journal article that argued that FG% fluctuates wildly from season to season, so you should basically only judge a kicker by his kickoff distance, which is fairly consistent.

I’m confused. Is it 70%, or random?

Fumble returns are non-predictive because they are too rare to quantify, but fumble recoveries aren’t? That’s a joke, right? Please tell me you see the disconnect there.

Aside from all that, I’m interested in your feedback to my last post in the [post=8131939]He Hate Me keeper league thread[/post].

3

Why doesn’t the “dependence on circumstances beyond the team’s control” kick in until the moment after a fumble is recovered? Seems to me those exact same circumstances impact the chance to recover in the first place.

4

How often does a player actually pick the ball back up and run with it? Usually, a giant pile forms on top of the ball.

5

I believe it was a New York Times article written by Aaron Schatz of FO (Google turns up no such Wall Street Journal piece). It would cost you $4.95 to read it now, but here’s the page on FO that originally linked to the article. It includes a few paragraphs summarizing it, as well as a very good discussion thread about the idea and various other Kicker stats. Here’s what Aaron Schatz wrote about it there and in the discussion thread:

6

Great OP VarlosZ.

While I really do enjoy Ellis Dee’s take on football and other entertainment, I’m sorry to say that I don’t think he really grasps statistics that well.

I know people have the opinion “you can make statistics say anything you want” and sometimes we do that around here. . .one guy cites a team’s “total offense”, another cites the team’s “scoring offense” and a flare-up ensues.

But those are “sports statistics”. FOs (and Bill James, and that guy who wrote the basketball book that “trashed” AI) are doing real statistics. People who don’t know statistics sometimes need to trust that people who DO know statistics know what they’re talking about and they’re not just making it up as they go along. The FO guys never set out to prove they’re right, or win arguments, or even predict football games.

They set out to look at the game of football statistically, and answer questions like “how much of a run can you attribute to the OL and how much to a RB?” and “Who really has a better offense between teams A & B when you adjust for the different defenses they’ve played?”

And, one interesting thing they’ve answered along the way is a question like : Team A is 3-6 and they’re -5 in fumbles. Team B is 6-3 and they’re +5 in fumbles. Is Team A really that much worse than Team B or are their records just a by-product of a random event, i.e. fumbles.

They have ways of testing and answering such questions, and if you don’t know statistics, well, I hate to use the old “appeal to authority” but you just have to trust that some people do know how to answer that question.

7

Both.

It’s 70% because that’s about how many fumbles in the defensive secondary are recovered by the defense (as they’ve just got more bodies around the ball than the offense at that point, in general). It’s random because there aren’t any defenses that actually recover 90% in the long run, nor any offenses that only lose 50%. NFL teams show no tendency whatsoever to recover loose balls at a greater (or lesser) rate than what the location of the fumbles would suggest. A team that does better than average one year is no more likely than any other team to do well the next.

In the Week 9 DVOA Ratings, FO addressed a poster with similar objections to yours. The relevant portions:

Incidentally, according to their 2005 book, the correlation between NFL teams’ recovery rates in 2003-04 was even more strongly negative (-.29 on offense, -.40 on defense). As I said, though, the inter-year correlation between both fumbles and forced fumbles was meaningful: +.39 on offense, +.26 on defense (again, that’s just for 2003-04).

Fumble returns are non-predictive in part because they are rare (most fumbles are not returned). They are also heavily dependent on circumstances which are totally out of the returner’s hands, i.e. the location of the and direction of the football and returner, as well as that of the other 21 other players on the field (as well at the tackling ability of the other team’s WRs and RBs).

The argument is not that no team has ever been good or bad at turnover returns – yeah, of course I’d rather have Deion Sanders returning an interception than Ted Bruschi. The argument is that the opportunity for a meaningful return is so rare, and so dependent on circumstances beyond the returner’s control, that a team’s past success returning fumbles and interceptions just doesn’t tell you anything meaningful about their chances of victory in the future.
You’re right that I could describe recovery the same way I describe returns – there probaby is some miniscule difference in recovery skill, hidden way, way below the threshhold of detection. However, there are two reasons that I feel more comfortable labeling fumble recovery as “random.” First, there are a lot more recoveries than there are opportunities for a meaningful return, so it should be easier for a difference in skill break through the random statistical noise. Second, the correlations for turnover return success at least hover around zero (i.e. completely independent from past results). If there was a fumble recovery skill that meant anything at all, wouldn’t it at least be able to get those correlations between past and future recovery rates up to “totally random”? Instead, the correlations are somewhat negative.
Whatever particular terms and phrases we decide best describe certain situations, my argument is fundamentally about what past performance does and does not tell us about the likelihood of certain future events. If you run the numbers, you find that certain events and/or tendencies tend to carry over from game (or season) to the next, while others just don’t.

So, for example, you could accurately say, “Michael Vick is prone to fumbling, and that hurts the Falcons’ chances of winning on Sunday.” However, no matter what happened last year or the last few weeks, you could not say, “the Falcons are great at recovering fumbles, and that helps their chances of winning on Sunday.” The facts just don’t support the latter statement.

8

Crap. Hey, Varlos. I just saw this thread. Don’t have too much time right this second but I hope if I post I’ll be more likely to remember to come back.

For now, real quick – their predictions: obviously they get some things right. My impression was that they didn’t really seem to get more right, than, say, I would. If I can scrounge up the book later on I’ll come up with some more examples. I know that they’ve been saying Hasselbeck would have Peyton Manning numbers for a couple years running, off the top of my head.

The numbers on Bailey: my problem is that, the way I look at it, the moment you say something like “we can’t explain this phenomenon which throws our methodology into serious doubt,” I have serious doubts. They acknowledged it, as you pointed out; so why should I trust their numbers at all? I feel like the first step towards a true statistical analysis of a game is to really trust that you’re evaluating the proper things to develop your metrics. But the Herndon thing… I mean, if they can’t tell me, what am I supposed to think, you know?

The Joyner thing: my problem isn’t with FO as such. The short of it is that I dispute that a statistical analysis of football will ever be able to develop a baseball-like set of metrics for player performance. I just don’t trust it. Football isn’t baseball, is what I think. Too much to account for, etc. This in combination with the admitted flaws discussed above means that I just don’t trust their numbers to tell me what they say the numbers are saying. Not that nothing they are doing is worthwhile – the fumble thing I’m totally on board with, for instance. But so I don’t trust Joyner either, and don’t hold that against FO, is the point.

Gotta go.

9

A few different issues here, I guess. FO makes two basic kinds of predictions: team success and individual stats. When it comes to team success, DVOA in Year X is a better predictor of success in Year X+1 than Wins, Point Differential, or Yardage Differential in Year X – and that’s before their projection system comes into play, which factors in schedule difficulty, player maturation, coaching turnover, roster turnover, and so on, which makes DVOA still more accurate as a predictor. Now, it’s certainly possible that you could do a better job of forecasting in the long run, but unless you have a nuanced understanding of something that the system fails to account for, I think it’s a tall order. They test and refine their formulas constantly, while we almost never test our half-informed subjective opinions.

That said, I’d be curious to compare their projection system’s “median wins” forecasts for 2006 with how other posters (or sports writers) saw the season unfolding. You didn’t happen to post complete speculative standings in a preseason predictions thread, did you?

Moving on, when it comes to indvivdual statistics forecasts, it will probably never approach the accuracy of, say, the PECOTA projections for baseball players. There are just too many hard-to-measure variables in football (injuries, usage, playcalling, etc.), not to mention many fewer opportunities per year for most football players. That said, their KUBIAK projection system will only get better as it assimilates more information, and it already does certain things quite well. In particular, its use of historical comparables gives it a better understanding of the effect of typical aging curves than most experts possess. It’s had a few noteworthy misses in predicting which young players will have a huge break-out season, but since its inception fantasy drafters would have done well to heed its advice when it projects an established, aging player to suffer a statistical collapse.

Somewhat off-topic, but you mentioned Matt Hasselbeck. Their 2006 projection was indeed off-base (partly due to injury, and partly due to the unexpected general collapse of the whole offense), but their 2005 projection was very accurate. He threw 15% fewer passes than they anticipated, which threw off his raw totals, but their rate-stat projections were effectively spot-on:

Projected – 62.2% Comp., 7.0 NtY/P, 19.6 Att/TD, 33 Att/INT, 38 Runs for 137 Yds.
Actual ------ 65.5% Comp., 7.0 NtY/P, 18.7 Att/TD, 50 Att/INT, 36 Runs for 124 Yds.

Basically, they predicted marked improvement across the board, and were proven correct. They were, in fact, slightly too pessimistic about his improvement as a passer.

If that were the case then nothing would ever get done. It’s much better to get to work on what seems to be the right path, and then use whatever lessons learned – and whatever criticisms received – to adjust course when necessary. “The best is the enemy of the better” they say over at FO: you shouldn’t let your inability to get perfect stats disuade you from developing better stats.

You’re supposed to think how interesting it is that the plays in which Kelly Herndon made a tackle or covered the intended receiver were, on balance, so insanely successful for his defense. Ideally, you’re supposed to wonder why that is, and what else could be done to better quantify a CB’s contribution to his team. Then, if you have some clever ideas, you could share them with FO and contribute to the project – or start your own project if you’re so inclined.

10

No, I’m pretty sure I wouldn’t have; but then I also wouldn’t hold it against anybody for getting predicted standings wrong. You could get everything “right” in your analysis and miss every team’s record by three games just because shit happens. I probably posted Eagles forecasts each of the last few years, and for shits and giggles I could probably scrounge them up. OK, I did NFC East predicted records this year:

Horrible. My Eagles call was spot on, at least. Conveniently I didn’t predict the 2005 season in the NFC East thread, but I imagine both me and FO would have been way the fuck off. 2004 prediction :

That’s all I got as far as predictions. But I got the Eagles’ record right both times I projected one! Take that, Outsiders!

You are more well-versed with this stuff than I am. I admit I haven’t seen much of their stuff since I bought the book two seasons ago and was disappointed. Didn’t you once take both Julius Jones and Kevin Jones at the turn in a fantasy draft because of FO advice, and didn’t that turn out poorly? I don’t have the book handy so if you tell me they’re more accurate than I thought I believe you. Could you give some examples of which breakouts they’ve predicted?

Maybe I’m misremembering. I thought they had him projected for like 4200 yards more than once. You’ve obviously got all the ammo here, so I submit to corrections wherever they’re warranted. And hell, if they’re good at it I’ll buy the next book.

Fair enough – from their perspective, and, I presume, from your perspective if you’re coming at it like a community project kind of deal. But from my perspective, as just a guy who watches games and really likes the talent evaluation aspect of it all, what’s in it for me while they learn these lessons? That’s really where I’m coming from; I’ll get to the gist of my problems in a second.

OK. I have absolutely no objection to the idea of FO’s projects as projects, like I said above. I think that’s really cool and absolutely worthwhile. In the Bailey thread, though, the question was how valuable is Bailey, and a peripherally FO-sponsored article was linked. So I was just kind of saying that I wasn’t all that comfortable with the idea of using those types of metrics to “prove” a corner’s value in the same way you might point to a VORP or something more esoteric to “prove” a starting pitcher’s worth. Does that make sense? If you tell me Charlie Hayes had an OPS of 1.200 last year, I trust that; I know that. It has a concrete significance to me in terms of how good he is as a hitter. I don’t think those numbers exist in football; if something ever happens where I say, hey, I was wrong about that, I’ll be pretty shocked. That has to do with what football is and what it isn’t – how could you possibly have a stat that can capably determine when a sideline route for a touchdown was the corner’s fault, and when the safety was late over the top? If it’s both, who got “beaten?” What if a tackle is supposed to be getting help from a back in pass protection, but the back misses on the chip and a sack results? And further, how many football plays could you really reduce to a statistical outcome if you didn’t resolve those kinds of questions first? In my opinion, since football is 11 on 11, with each guy having essentially a brand new and independent assignment on each play – an assignment he can fuck up and cause some other guy to look bad – numbers are always going to miss the boat in a way they don’t miss with baseball metrics, because of the head-to-head, discrete-outcome nature of an at-bat.* That’s why complicated individual stats, especially for positions like DB or linemen, just don’t do it for me. They describe exactly what it is they describe, but have no significance out of context.

It’s absolutely cool to me that there are people working on these numbers, and I seriously doubt that there could be any new stat that I wouldn’t at least be interested in checking out, but it’s when you make the jump from descriptive (of things that already happened, that is) to evaluative that I lose faith. I know FO even has a disclaimer, something to the effect of "our ratings don’t rank players by how good they are, they rank players by how good they are, on their team, at that time. Which is fine as far as it goes, but I’m not sure how far that is, is all. I can’t look at FO’s numbers and say, OK, this guy is better than I thought, this guy’s worse, etc., and that to me is the essence of statistical analysis.

  • Joyner is especially bad about this, with his “bad read” metrics and so forth, because he gets around this problem by just providing his own subjective determination of what happens on plays. If I’m going to trust some geek to watch tape and tell me what a “bad read” is and who’s making them, I don’t need the stats, do I? Just go whole hog – tell me who’s good and who isn’t and I’ll gobble it up and go on my merry way. FO doesn’t watch tape and make subjective evaluations, though, do they? I respect that.
11

Just to touch on this briefly, even though I know it wasn’t directly addressed to me and I think we’re generally on the same wavelength when it comes to football… I’m not going to try to pretend I have the stat chops to go toe-to-toe with Aaron Schatz or whatever. I’m not a statistician; I took AP Stat six years ago and another course in college. Maybe on the SDMB I’m not even of average mathematical ability. But the fact that these guys can run circles around me with probabilities and distributions doesn’t necessarily mean their numbers are football-valid. Good Will Hunting himself could draw me up a formula, but if it told me Kelly Herndon was better than Champ Bailey (hyperbolized by me), or that the Cincinnati Bengals were (rough paraphrase but not exaggerated) the next dynasty and were likely to win multiple championships in the very near future, I’d still have some questions to ask.

12

First off, as a Giants fan, I am unswayed by the following premise:

The Giants haven’t had even remotely the same roster in the first and second halves of a season since the Superbowl run in 2000. This past year was remarkably bad on the face of it, losing both DEs, both OLBs and both starting CBs, but the losses have been just as numerous and severe in each of the last several seasons. Seriously, I look at the Jets with their ONE injured starter all season and think to myself “what the fuckin’ fuck? They play on the same fucking field fer chrissakes!” I barely remember what it’s like to have a reasonably healthy Giants team.

I don’t see how or why you want to apply this to fantasy football. Losing a fumble – random or not – is quite detrimental to your team’s chances of winning. That, to me, is what fantasy football is all about.

And “offensive fumble return TD” is about as random as random gets, yet you seem to have no issue with those being worth 6 points. Why the hate for fumble recoveries? If you still would like to devalue fumble recoveries, then give me your reasoning for why you don’t have a problem with the following scoring categories:

  • offensive fumble return TD
  • return TD
  • defensive TD
  • safety
  • field goals

To my eyes, they (plus fumble recoveries) are all legitimate fantasy scores because they directly impact the team’s chances to win.

Finally, I concede that fumble recoveries are random. In the end, I am ignorantly unswayed by the statistics. Instead, it’s the shape of the ball, which was designed to bounce irregularly. Fumble recoveries being random was deliberately built into the game itself, and based on the FO analysis, it was built in quite effectively.

13

Good memory. I did indeed take the Jones ‘Brothers’ w/ the 10th and 11th picks in a fantasy draft in 2005. I don’t recall who I passed up to take them, but it wasn’t a huge reach or anything. Kevin Jones is almost certainly FO’s most memorable flop – their system loved his 2004 second half (911 yards in eight games) and his team’s upcoming easy schedule so much that it projected him to rush for more yards than anyone else in '05, so FO put him on the cover of their book that year. It liked Julius as well, to a slightly lesser extent. It whiffed on both – Kevin because he wasn’t as good as they thought and he was injured for much of the year, Julius mostly because they didn’t expect him to split carries with Marion Barber to the extent that he did. They projected 722 combined carries for the pair, but they actually had only 443 runs – at fours yards per carry, that’s over 1,100 missing yards.

They had him at about 4100 and 4300 yards in '05 and '06, respectively. Like I said, their '05 prediction was pretty much spot on except for the fact that they didn’t see the rest of the team being so good that they’d be running out the clock all the time. In '06 Seattle faced an inexplicable team-wide collapse, and Hasselbeck was out for like a third of the year.

I don’t want to comb through the books to come up with a list, but, off the top of my head: They fully expected Philip Rivers to have great success, they liked McNabb to bounce back, and expected Frank Gore to be a successful feature back (though not as successful has he actually was). Like I said before, they’ve had more success to this point predicting declines: they pegged Tony Gonzalez’s fall to the 2nd tier of TEs in ‘05, they figured Warrick Dunn would regress badly this year, and they nailed Edgerrin James’ complete collapse in Arizona (they said 1055, 3.7, 6 TDs, actual was 1159, 3.4, 6 TDs.

No doubt one could find a similar (or larger) list of failed predictions, but on balance I think they do about as well as your typical “expert” at picking out sleepers and busts. The difference is that KUBIAK will only get more accurate with every passing year, while the experts aren’t exactly making any great stides. Give it time.

Absolutely. In the paragraph following the above quote, I think you do a good job of explaining why assembling meaningful metrics for most of the players on the field is so difficult for football, and (relatively) so easy for baseball. That doesn’t mean the numbers can’t tell you anything, though.

Think about it: how do you know which CBs, Safeties, and LBs are really good? You probably have a pretty good sense of the players on whichever teams you watch every single week, though even that is touch & go (hard to see which DTs are always soaking up blockers on run plays, for example, or which safeties are locking down their quarter of the secondary on a regualr basis). Beyond that, you’re left with the group-think opinions from TV and print analysts. (And how did they form their opinions? Not by watching every game, that’s for sure.) Conventional individual defensive stats are often useless or worse – lots of Tackles often just means that opposing offenses liked to run right at that player. Ditto Interceptions. “Passes Defensed” is a joke – it’s based on the subjective impressions and definitions of the individual scorekeepers working at the various stadiums (Philadelphia is particularly wacky in this regard, as any defender who happens to be in the general vicinity of an incomplete pass gets credited with a defended pass). So there’s a lot of room for improvement in objective measurements, particularly in combination with some simple observation and scouting (e.g., Lance Briggs typically covers opposing TEs, and Chicago’s defense is phenomenal at preventing successful passing plays to the TE, therefore Lance Briggs is a great pass defender).

Sure you can. Example: think about Arizona Safety Adrian Wilson for a few moments. How good do you think he is? You’ve probably heard some favorable mentions about him, and his Madden rating is in the low 90s, so he’s good. But he’s not in the same class as Roy Williams, Ed Reed, Troy Polamalu, Bob Sanders, or Brian Dawkins, right?

Now consider this passage from FO’s 2006 book, regardin Wilson’s '05 campaign:

  • → “Stop Rate” refers to the percentage of a defender’s plays which prevent the offense from achieving a “success,” defined as at least 45% of needed yards on 1st Down, 60% on 2nd Down, and 100% on 3rd or 4th Down.

Unless you already thought Wilson was one of the best defenders in the league, how could this new information not improve your opinion of him? No doubt these numbers are helped by the fact that Wilson plays closer to the line than most Safeties, but he’s not a free-ranging extra-LB type, like Polamalu, and the fact that he’s in on so many plays and still has numbers that are leaps and bounds better than anyone else at his position is a pretty big clue that he’s a special player.

I don’t know, maybe nothing. It depends on what you enjoy about football. For instance, I enjoy being able to assign blame and praise more acurately – e.g. offensive lines tend to get too much blame for sacks and too little credit for rushing success. I like knowing how much of a team’s record has to do with luck – like, say, if other teams recover crazy fluke fumbles in the end zone or beat you with last-second 62 yard FGs. I like knowing I should be nervous if my QB throws a floater to Adrian Wilson’s territory. I like knowing that a team which beats the hell out of a bunch of awful clubs is more likely to win it all than a team with some close, gutsy wins over quality opponents. Mostly, I think, I like calling out people for relying on clichés and “common sense,” and being able to point to numbers that prove I’m right.

I’m with you here. It drives me crazy that he tracks all these complex metrics that are based entirely on subjective opinion. FO does do some film analysis, but they keep it separate from their stats (aside from keeping track of which defender is covering an intended receiver).

14

I don’t have any problem with the scoring categories you listed because they all represent something that a player (or defense) actually did. On the other hand, a RB doesn’t lose a fumble; he fumbles, and then the ball is either recovered or not. If you draft a player prone to fumbling, you shouldn’t get off the hook because his team gets lucky with the recoveries, nor should you get punished some additional amount because his team got unlucky.

No kidding. Off topic, but what did you think of the Tim Lewis firing? I’m of the opinion that he was not the problem – the Giants basically had average defensive performances these past two years in spite of some horrible injury problems and a total lack of talent in the starting secondary. I think he did a pretty good job with what he had.

15

But if he recovers his own fumble, which isn’t all that unusual, he shouldn’t be extra-penalized for what amounts to nothing happening. The other times, I chalk up the unfairness to the same fantasy gods as I do a deflected pass that gets intercepted. In the end, it’s the net effect to the team on the field that seems relevant to me, not what each player individually earns.

What is the correlation for 50+ yard field goal attempts? Do the kickers who excel at them in the first half of a season also excel in the second half and vice versa, or is there a similar negatively correlated regression to the mean as with fumble recoveries?

I was very unhappy with both coordinators, and am glad they are both gone. If there were a blue-chip coaching prospect out there, I’d want Coughlin gone as well. Sadly, there wasn’t, so I’m okay with keeping General Tom for another year. (Note that the last time a 1-year extension was offered to the head coach, Fassel brought the team to the Superbowl.)

My biggest gripe was that neither Lewis nor Hufnagel were capable of making adjustments. Granted, watching the Jets make real-time adjustments made it seem much worse by comparison, similar to the injury and penalties/discipline situations.

Think back to the beginning of the season, where the defense was getting gashed. It took Lewis three full games, a bye week and getting called out by the players to make his first adjustment, after which the defense stepped up and the team went on to an impressive five game winning streak. Four weeks is way too long to make an adjustment; something like one quarter is what I’d rather see.

About the best thing I can say about Tim Lewis is that he was better than Johnny Lynn. If that’s not damning with faint praise, I don’t know what is.

ETA: Thinking back on it, the first adjustment actually came in the second half of the Seahawks game. Still way too long, IMO.

16

Tiki’s latest bout of public whining shed a bit of light on this. First, a quick recap: One obvious reminder that the Giants were injury-ravaged even before the Coughlin era (when it’s been just as bad, if not worse) was Coughlin’s first press conference when he got the job. He was asked about all the injuries the Giants had been suffering, and if he had any plan to fix it. His response? “Injuries are state of mind.” Reporters jumped all over this, reading into it that he was implying the players were belly-aching fakers. At the time I read it differently; more along the lines of injuries being caused by a lack of mental discipline. ie: Keep your head on a swivel or you’ll get laid out.

Back to the present, Tiki recently whined about how he might have come back if he hadn’t gotten worn down so much the past three seasons. In addition to Coughlin’s constant yammering, Tiki noted that Coughlin has them practicing in full pads every single week, even into the playoffs.

WTF? My understanding is that nobody does this anymore, since it wears down players so much that injuries are almost guaranteed. I vaguely recall a story a few years back about some playoff-bound coach (maybe Andy Reid) being second-guessed because he was running full-pads practices into December. Coughlin was running these practices into January? Fuckin’ shit.

Looks like next year will be yet another injury-plagues year. Great. Hey, maybe we’ll have another star receiver get a major concussion during training camp. Who needs a guy like Shockey anyway, right?