Newton's Football

Forbes writer Allen St.John and materials scientist Ainissa Ramirez recently wrote a book titled Newton's Football on various intersections between science and our favorite sport. It's right up ANS' alley. Allen has an article at Forbes with an excerpt from the book:

"The theory that coaches were purely motivated by job security and didn’t want to go against the conventional wisdom, that didn’t quite satisfy me," says Brian Burke of Advanced NFL Stats. Week in and week out, Burke analyzed games and saw evidence that NFL coaches were costing their teams yardage, 1st downs, and, ultimately, games because of the questionable decisions they were making. He further noticed that those coaches almost invariably erred on the side of caution. But why?...

"In terms of team building, risk taking is good," Burke argues. He notes that in any given year the average NFL team enters a season with a Bayesian prior expectation of winning a Super Bowl that’s 1 in 32. "You’re a 31–1 underdog. You want to take chances," Burke explains. "It’s okay to miss the playoffs and win five games instead of seven. It doesn’t hurt you that much. But teams are very conservative. They'd rather win a few more games and avoid having a terrible record."

Cade Massey on Flipping Coins and the NFL Draft

Readers of this site will recall the name Case Massey. Along with fellow noted economist Richard Thaler, he co-authored the Massey-Thaler draft study titled The Loser's Curse. The paper found that, under the previous CBA, "surplus" draft value peaked with picks in the late first round and early second round. Surplus value was defined as the expected performance value above which a team could expect by spending an equivalent amount on a veteran free agent.

Massey has continued research into the draft. His presentation at the 2012 MIT Sloan Sports Analytics Conference outlines his recent findings. (I recommend using IE to view the presentation. Chrome didn't play nice with the video.) The slides from the brief can be viewed here.

If I understand things correctly, Massey has found that:

Roundup 9/23/10

Keyshawn Johnson thinks Romo is average, and he's holding the Cowboys back. Jason Lisk would beg to differ. This year, Romo sports a +30 EPA, meaning his play would be expected to generate 30 points of net point advantage over five games. But he has a negative WPA, meaning he hasn't played well in high-leverage situations.

Jason has been writing at a site called The Big Lead, which, except for JKL's stuff, is mostly pedestrian. The good news is that you can subscribe to Jason's stuff only using this feed link. Here's another neat post in which Jason looks how a few teams compare to historical teams with similar statistical profiles.

Say it ain't so, Phil!

More Phil on how mainstream reporters should report on sabermetric research. I've always had a very good experience working with reporters. They've been open-minded, inquisitive, and willing to accept that sometimes there's no statistical weight to the argument they're trying to make.

How amazing are the 2010 Chargers?

This seems to be a good idea.

Detecting momentum is harder than you think. I've been thinking of various ways to construct a research test for momentum in football games for a while. I have a few thoughts, but more ideas are always welcome.

Roundup 10/16/10

NFL Forecast has fresh playoff probabilities.

Peter King smartly uses efficiency stats (and not total yards or points--awesome!) to diagnose the Saints' troubles this year.

What are the odds?

I recently read this over at Cold Hard Football Facts: "Running is part of football. Always has been, and always will be. The problem from an analysis standpoint is that no one has come up with a way to demonstrate how running helps teams win. But that’s a shortcoming of NFL analysts, not a shortcoming of the running game." Hopefully, now there is one fewer shortcoming.

Platooning QBs? One year when I was at Navy we did that. We had one QB who was a passer who played between the 20s, and one guy who was the runner to played near the goal lines. I think we went 1-10 that year, with the only win against Army.

Last week was a bad one for #1 pick QBs.

Don't sleep on the Titans this season.

Over at the Community site, Chris Alan gives updates his analysis on surviving in a suicide league. Have your own research or analysis you'd like to share? Send it in to Advanced NFL Stats Community.

Roundup 9/25/10

Advanced individual player stats will be updated immediately after each game this season.


Game probabilities will start for week 4. They'll be featured at the NY Times again this season.

Wow.

TJ uses EP to analyze a big play call in the DEN-SEA game from last weekend.

NYT article on the 3-4. 

From the Advanced NFL Stats Community reboot:

Dan Schlauch on placekicker salary distribution. Leave some comments and give Dan some feedback.

John Candido is parsing weekly play-by-play for everyone this season. If you use John's data, be sure to leave behind a thanks. I'll be updating 2010 data periodically, but not every week.

Keep the submissions coming! Thanks again to Ed for tending to 'Community.'

Roundup 9/4/10

In basketball, don't call a timeout at the end of close games. You'll be more likely to draw a foul and win.

A salary cap also needs a salary floor.

The historical distribution of talent in the NFL.

A neuroscientist's struggle to understand regression to the mean.

The Sports Reference sites, including PFR, was named among Time's 50 best websites. Congrats, guys.

Tony Romo is not a mirage.

Advanced Public School Stats. What's next, Fantasy Teachers leagues? ...With his first round pick, Brian Burke selects Jaime Escalante, math teacher, Garfield High School... Stand and Deliver-great flick by the way. Escalante passed away earlier this year.

Roundup 8/21/10

The Roundup feature returns for the 2010 season. The twitter feed over there to the right has served as my voting ballot for interesting football links recently, but the Roundup lets me make comments and explain what I found interesting in each link. There's a backlog of links since the last roundup, so although some of these off-season links might be a little stale, they are still worthy of note.

If a free-agent who brings a certain number of wins comes to a big market team, he'll increase its revenues more than he would to a small market team. So why do teams tend to pay equal prices for free-agents? Phil Birnbaum tells us. Part II.

More Phil. This time a great essay on numeracy. Understanding the world around us requires and understanding of math, but sound logic is just as important if not more so.

Reviving fallen franchises.

What if the draft was an auction? (Off topic, here's a funny/disturbing auction. Apparently they have an edgy sense of humor in New Zealand. Helmet-knock: Mind Your Decisions.)

Player-specific win-loss records. (Part II) A very cool idea from Tango using Win Probability.

Pre-Season Predictions Are Still Worthless

Last year I started my stint at the NY Times by calling attention to just how bad NFL preseason predictions are. I compared the “advanced” projections for team win totals compiled by a fellow stats website called Football Outsiders to two benchmarks. They had predicted doom and gloom for the Jets last year, and my article was intended to relieve Jets fans of needless worry. As it happened, the Jets made the playoffs and went all the way to the AFC Championship game.

The first benchmark was a mindless 8-win prediction for every team. Let's call this the Constant Median Approximation system, or CoMA for short. This benchmark represents zero knowledge. It’s what you would guess if you had no information at all about any of the NFL teams except that they each play 16 games. Certainly anyone can out-predict a coma patient, right?

MIT/Sloan Panel

Here is the video of the panel at the MIT sports analytics conference I referred to in my post about how Bill Polian doesn't get it. The discussion is educational throughout, and despite my criticism of Bill Polian, he has some very wise things to share. The participants are Mark Cuban, Jonathan Kraft, Daryl Morey, Polian, and Bill Simmons, with Michael Lewis as the moderator. It's over an hour long and worth your time. I've embedded it below, but first let me share what I took to be the most interesting points:

Dave Berri Responds

Stumbling On Wins author Dave Berri responds to my analysis of the debate regarding whether top draft pick QBs are really any better than later picks. He has a good point about Elway.

Steven Pinker vs. Malcolm Gladwell and Drafting QBs

Last season you might recall a dust-up between Harvard evolutionary psychologist Steven Pinker and popular science author Malcolm Gladwell over whether teams really have any ability to predict which college QBs will pan out into good pros. You might be wondering what the heck a psychologist and a pop-science author have to do with NFL football.

In his book What the Dog Saw, Gladwell wrote about how hard it is for school administrators to discriminate the better teacher candidates from the lesser candidates. Gladwell used the NFL draft to illustrate how difficult it is for anyone to predict human performance, even in a sport where there is ample performance metrics and every step, throw, and catch is videotaped from 12 different angles. Gladwell was referring to what was reported by economists Dave Berri and Rob Simmons as a "very weak" correlation between draft order and per-play performance by QBs.

In an exchange of letters following Pinker's critical review of What the Dog Saw, Pinker took issue with Gladwell's claim that there was "no connection" between when a QB is taken in the draft and his per-play performance. Pinker wrote that this is "simply not the case."

As has been pointed previously, the problem with the weak correlation cited by Gladwell is that it excludes players who are not judged good enough by coaches during their development to warrant much if any playing time. At its core, the NFL draft is a process of selection, and we should expect  selection bias will taint most attempts at analysis. Gladwell looked at the draft process and (correctly) said:

"Coaches and GMs turn out to be good decision-makers when it comes to drafting quarterbacks when you consider the fact that the quarterbacks who never played aren’t any good. And how do we know that the quarterbacks who never play aren’t any good? Because coaches and GMs are good decision-makers!”

But Gladwell's argument cuts both ways. The only way to see that coaches and GMs aren't any good at drafting QBs is to assume they're no good at choosing which QB on their roster to play in games!

In this post I'll attempt to settle the question of whether NFL scouts really have any ability to identify the better QBs. Do the QBs picked higher in the draft turn out to be better performers on a per-play basis? Is Pinker correct that they do, or is Gladwell correct that they do not?

The CBA and Union Economics

Roger Goodell says veteran players want the rookie pay scale reduced. Even economist Richard Thaler is offended by the salaries of draft picks. He writes, "veteran players would probably agree with the principle that eight-figure salaries should be reserved for players who have already proved themselves on the field."

Of course they would!

People think of unions and corporations as adversaries in perpetual conflict. But there is one thing both groups can agree on.

Stumbling on Wins Giveaway

When I started to get really interested in sports statistics I thought, "Somebody should write a Freakonomics, but about sports." I was woefully uninformed about the vast library of research already done, particularly about baseball. At first I (foolishly) thought, "Maybe I could write that book." I did some clicking around and I was crushed to learn there already was one.

At my first opportunity I ran out and bought Wages of Wins. I thought," This is cool. I could do this too." So I'll always carry a debt of gratitude for authors Dave Berri, Martin Schmidt, and Stacey Brook. I didn't agree with everything in Wages, but it asked all the right questions and got me thinking about sports in a whole new way. Besides, if you're waiting for the sports book that you'll agree with 100%, you'll be waiting a long time.

Rethinking the Massey-Thaler Draft Study

Economists Richard Thaler and Cade Massey authored a widely-read research paper analyzing the value of NFL draft picks, and they've recently published an updated version. The paper's primary finding was that teams are overconfident in their ability to choose the best players. In essence, the very top picks are overvalued relative to later picks, both in terms of what teams are willing to trade to move up in the draft and in terms of salary.

Recently Richard Thaler penned an article for the NYT discussing his paper. But puzzlingly, he goes on to make a claim that clearly contradicts his own research. He writes, "it makes absolutely no sense to be giving so much money to unproven rookies, many of whom turn out to be busts." Further, he writes, "veteran players would probably agree with the principle that eight-figure salaries should be reserved for players who have already proved themselves on the field."

Roundup 3/20

More on advanced stats for golf.

Here is a great application of the Expected Points values I released into the wild last season. This is exactly the kind of stuff I hoped would start happening. It's a look at the Cowboys running game through the lens of EPA. I'd love to see more people take advantage of the EP model.

Here's another article in the same vein. It implores Broncos coach Josh McDaniels to use the EP model to become more aggressive on 4th down.

Sean McCormick at Football Outsiders has a great write-up on the much-hyped Draft class of 2004.

The geometry of free-throw shooting. (Hat tip: M/R)

A strange quirk in how seeding in the NCAA basketball tournament can allow lower seeds to advance more easily. #10 seeds are actually more likely to advance than #9 or #8 seeds. #12 seeds are twice as likely to make the Sweet 16 than #8 seeds.

Turnovers are the key to predicting upsets.

Mathletics

With the football calendar at its darkest nadir--no games, no signings, no draft, just some off-season workouts--and basketball season getting into full swing, maybe it's a good time to broaden our statistical horizons. If you're new to sports analytics or want to become more familiar with the methods used in other sports than football, I recommend Wayne Wilson's book Mathletics.

Wayne is a professor of decision science at Indiana University and has consulted for the Dallas Mavericks for several of the last few seasons. Basketball is his wheelhouse, but Mathletics covers baseball and football analytics as well. The book is really two things. It's a primer on the various principles and techniques used in sports analysis, and it's a how-to book on how to use Excel to do the actual computations. It's perfect for the guy who wants to grab some data off the Web and get his feet wet crunching numbers.

Roundup 3/5

Scientists find the part of the brain that causes coaches to punt too often. (Hat tip: Marginal Revolution)

The Coase theorem says that dynamic ticket pricing is a good idea.

The Chicago Sun-Times says that Jay Cutler's interceptions were all due to poor decisions and not to pass protection deficiencies. Seems KC Joyner was right and I was wrong about Cutler going into last season. The stats show Cutler was a still minor upgrade over Orton, even accounting for all the interceptions. Whether the trade was worth it is another matter.

Jason Lisk defends Joe Namath from the charge he may not be a legitimate Hall of Fame QB. My take--It's the Hall of Fame, not the Hall of Passing Efficiency. He definitely belongs.

Lisk also tells us all about real football. If you read one link from this post, read this one.

What tends to happen to teams that lose the Super Bowl?

Roundup 1/16

Does a playoff bye increase a team's chances of winning in the divisional round? Jason Lisk reminds us that, accounting for team strength, probably not.

Jason also shows that Vinny Testaverde was better than most people think if you consider how bad the teams around him were.

More on QB comebacks from Scott Kacsmar.

Last week we saw the Ravens stomp the Patriots throwing only 10 times for 34 yards. Chase Stuart looks at the reverse. How close have NFL teams come to never running the ball in a game? Could we see something like that in today's Cardinals-Saints game?

Roundup 12/19

If you're snowed in like me this Saturday afternoon, you've got plenty of time on your hands. Early season college basketball doesn't really grab my attention. Neither does the apparently sponsor-less "New Mexico Bowl." (Actually, I often root for Fresno State due to being stationed nearby at Naval Air Station Lemoore, CA for many years.) Here are some links to help pass the snowy Saturday.

Chase Stuart from PFR looks at Steven Jackson's huge year for the struggling Rams. That's very unusual for a RB to have such a big year on a losing team. In a contribution to the Fifth Down Blog, Chase looks at the five least likely playoff teams in history based on their early season struggles.

Another Run-Pass Balance Study

Benjamin Alamar, author of the Passing Premium paper (critiqued here), takes a second stab at comparing the values of running and passing with new research. A brief presenting his methodology and findings was presented at a recent symposium on sports statistics. This time Alamar uses expected points as a measure of value, and compares the EP values gained by passes with those of runs. He defines risk as the probability each play type will result in a negative EP change, and finds that running is both less productive and "riskier" than passing. There are three fairly big problems in the methodology. Fortunately, all three can be rectified.

First, Alamar creates his EP values using a simple linear regression (see slide 10). Using down, distance, and yardline (plus other variables controlling for quarter and other effects), he produces an EP equation. This is a really bad way to estimate EP values. They're not linear, and there are any number of interacting effects within. The Levitt-Kovash paper, in contrast, uses a regression with quintic terms and full interactions to create their EP values. (I prefer a  more direct method--looking at the data empirically and smoothing it using a method called LOESS.)