Doing the math on Belichick’s decision

When the New England Patriots and Indianapolis Colts met this NFL season it was obvious beforehand that the game would be a good one. Winners of 2 of the past 3 Super Bowls, the teams sported future Hall of Fame quarterbacks and at least one future Hall of Fame coach: Bill Belichick, the most successful NFL coach of the past decade. What was unexpected, perhaps, was that the game would essentially hinge on one play and one decision by Belichick. That decision was to “go for it” on a fourth down play deep in the Patriots own territory rather than punt the ball far down the field, a decision that was virtually unprecedented.

Unprecedented, maybe, but not unadvised. For years statistical analysts have been crunching the numbers on the various league-wide probabilities of success and failure and suggesting that the alternative strategy of never punting is the better one. Despite these conclusions, NFL coaches had–up until Belichick’s decision–continued business as usual and routinely punted when in similar situations. And as expected Belichick was pretty much excoriated by his peers for even contemplating going for it.

Belichick’s decision brought the aforementioned analysis to the fore, and for a few days statistics took center stage on the sports pages. Calculations and debate appeared on ESPN (including at their online site here), as well as at the NY Times, the Boston Globe, and countless blogs and discussion boards. Even the Freakonomics site weighed in.

(And in case you did not know, the Patriots failed on their fourth down try, and ended up losing the game.)

Q&A: Author Malcolm Gladwell

A very short interview, and so a very short note here. But Malcolm Gladwell, the mega-bestselling author of The Tipping Point, Blink, and Outliers, has come down firmly for more math in the media. (Not more of this blog specifically, that is, but actually more math in the media.) Specifically, in this Time magazine interview, he advocates:

Aspiring journalists should stop going to journalism programs and go to some other kind of grad school. If I was studying today, I would go get a master’s in statistics, and maybe do a bunch of accounting courses and then write from that perspective…. Journalism has to get smarter.

Statistics is getting some good press lately. See also the earlier NY Times article “For Today’s Graduate, Just One Word: Statistics.”

Some See Numerical Oddity in Pollster’s Election Surveys

Carl Bialik, the “numbers guy” at the Wall Street Journal, writes here about a polling firm that has come under fire for the results it reported concerning the 2008 presidential election. Strategic Vision LLC, a polling firm based in Georgia, was the only polling firm that declined to release any information regarding where its polling numbers came from to the American Association for Public Opinion Research, the industry’s professional organization. Nate Silver, a statistician who writes a political blog, then examined the numbers that made up Strategic Vision’s polls and found extremely unlikely statistical anomalies. Subsequently, Strategic Vision was censured by the AAPOC and appears to have dropped the polling aspect of its business. Other anomalies have since surfaced in the firm’s polls, including one done for the Oklahoma Council of Public Affairs which claimed to show (among other things) that only 23% of the high schoolers there knew who America’s first president was, and that about 10% of the students had listed the two major political parties as ‘Republican and Communist.’

Plugging Holes in the Science of Forensics

This NY Times article discusses the recent National Academy of Sciences (NAS) report that found “serious problems” with many of the forensic ‘sciences’ practiced in crime labs today. Most of these problems are of the mathematical variety: with the exception of DNA evidence, much of the type of research done on the forensic sciences is geared towards finding information at the crime scene, and not geared towards statistically checking how valid and reliable these finds really are in identifying the perpetrator. A foretaste of this came a few years back, when the forensic science of bullet lead analysis (first used to convict Lee Harvey Oswald of the murder of John F. Kennedy) was abandoned by the FBI after an NAS report found it lacking as well.

The article goes on to describe the efforts of various scientists to “refine [the] mathematical tools” being used to put the other forensic sciences on the same level as DNA.

A $1 Million Research Bargain for Netflix, and Maybe a Model for Others

Netflix is a company that (for a monthly fee) allows you to check out and return movies through the mail. A large number of their customers request movies based on Netflix’s “Cinematch” suggestions, which recommends other movies the customer might like based on the movies he or she has previously ordered. In October 2006, Netflix offered a $1,000,000 prize to anybody who could improve the company’s recommendation algorithm by 10% or more. It proved to be an very difficult and interesting contest, with various mathematical (and other) twists and turns along the way, some of which are sketched here. The prize was recently claimed by a seven-person team of statisticians, machine-learning experts and computer engineers from the United States, Austria, Canada and Israel.

Amazingly, the winning team’s algorithm was mathematically identical to another team’s algorithm, which was submitted only 20 minutes after the winning team’s entry. Still, the second place team is not crying about it:

Yet the scientists and engineers on the second-place team, and the employers who gave many of them the time and freedom to compete in the contest, were hardly despairing. Arnab Gupta, chief executive of Opera Solutions, a consulting company that specializes in data analytics, based in New York, took a small group of his leading researchers off other work for two years. “We’ve already had a $10 million payoff internally from what we’ve learned,” Mr. Gupta said….”So for us, the $1 million prize was secondary, almost trivial.”

The article from which that second quote is taken notes that this type of corporate-led open competition could become a new model for improving business, and mentions a number of other such competitions that have cropped up.

Are Your Friends Making You Fat?

This story by Clive Thompson from the NY Times Magazine examines networks and the signals that travel over them–only the network and the signals aren’t made of wires and electric impulses, the network is the social network of friends and relations and the signals are behaviors. A number of scientists have been using the mathematics of networks to analyze the impact that the behaviors of peoples’ friends and relations have on their own behaviors, and how certain behaviors sometimes get “transmited” through the network in ways startlingly like more physical networks. This article (free registration required) describes some of those scientists and their findings.

Credit Card Use Is Ripe for Data Mining

Credit card companies are increasingly using statistical models to guess which of their customers are most likely to default, to not care about rate increases, to incur late fees, and things like that. This radio segment, which appeared on the daily “Marketplace” segment of American Public Radio, describes some of the issues surrounding the practice.

Math Backs Limited Profiling in Airport Screening

William Press, a computational biologist and computer scientist at the University of Texas, Austin, applied a statistical model for studying rare events to the problem of determining how often to select people for searches in airport screenings. This NY Times article (free registration required) by Sandra Blakeslee describes his findings.

For Today’s Graduate, Just One Word: Statistics

This title of this Aug. 5, 2009 NY Times article (free registration required) says it all. I’ll include on more snippet from the article by Steve Lohr:

Statisticians … are finding themselves increasingly in demand — and even cool. “I keep saying that the sexy job in the next 10 years will be statisticians,” said Hal Varian, chief economist at Google. “And I’m not kidding.”

Super Crunchers: Why Thinking-By-Numbers is the New Way To Be Smart

This 2008 book by Ian Ayres landed on the NY Times business bestseller list. Ayers, who has a Ph.D. in economics, is a professor at both the law and management schools at Yale University. Super Crunchers describes the many ways in which two statistical tools—random sampling and regression—are changing the ways in which corporations and government are doing business.