National Football League

The National Football League (NFL) is presented as both a forecasting arena and a setting where conventional expertise can resist evidence. Its games generate point spreads that summarize bettors’ expectations: although spreads only weakly predict the final scores of most games, the market is difficult to beat consistently, and its final line is often a more reliable forecast than alternatives. The league’s coaches, however, do not necessarily make the most effective decisions. A statistical analysis of fourth downs challenged their customary caution, and the Rams’ loss in Super Bowl XXXVI made the contrast concrete: a crowd can aggregate dispersed judgments effectively, while a professional group can continue to follow a strategy that the evidence calls into question.

A Forecasting Market

The NFL’s weekly games provide a recurring test of judgment under uncertainty. At the Mirage Hotel and Casino in Las Vegas, sports book director Robert Walker offers lines on everything from professional football to Ivy League basketball. A point spread indicates how much a favored team is expected to win by; a bettor backing the favorite must see it win by more than the spread, while a bettor backing the underdog wins if it loses by less than that margin or wins outright. Odds can instead set the amount wagered against a potential payout.

The bookmaker’s purpose is not to predict the winner personally. Walker aims to draw roughly equal amounts of money to each side. The book then keeps the vigorish, or vig: in the book’s example, two bettors each put up $11, the winner receives $21, and the bookmaker retains $1. If betting concentrates on one team, the line shifts to encourage wagers on the other. Before the 2001 Super Bowl, for example, the Mirage opened with the Baltimore Ravens favored by two and a half points; after two early $3,000 bets on Baltimore, Walker raised the spread to three. The opening line begins with the bookmaker, but subsequent movement incorporates bettors’ judgments. Once posted, the number becomes public property, and the final spread reflects their collective assessment.

What the Spread Can and Cannot Predict

The NFL betting public is not a crystal ball. Point spreads only weakly predict the final scores of most games. But consistently beating the final spread is difficult even for well-informed gamblers: favorites cover in about half of games, underdogs beat the spread in the other half, and there are no obvious recurring errors such as home teams winning more often than predicted or road underdogs being systematically undervalued. Occasional patterns can be found—for instance, one study identified home underdogs in weeks fifteen through seventeen as historically good bets—but such exceptions require searching rather than offering an easy route to outperform the crowd. Roughly three-quarters of the time, the Mirage’s final line is described as the most reliable forecast available for NFL game outcomes.

This record makes sports betting a useful laboratory for examining prediction, not proof that markets always know best. The spread’s strength lies in its ability to incorporate information through many wagers, while its limits are visible in the distinction between predicting a score and identifying which side will cover. The market can be well calibrated in aggregate without providing precise foresight about each game. That combination—imperfect individual predictions, but a difficult-to-beat collective benchmark—is central to the book’s account of crowd judgment.

The Fourth-Down Problem

A different kind of judgment appears on the field, where coaches make decisions under pressure and within established professional norms. Before Super Bowl XXXVI, the St. Louis Rams were fourteen-point favorites over the New England Patriots. The Rams had one of the most potent offenses in NFL history, led the league in eighteen statistical categories, and had outscored opponents 503 to 273 during the regular season. Their apparent advantage made the choices Mike Martz faced during the opening quarter especially consequential.

On the Rams’ first major drive, they moved from their own twenty-yard line to the Patriots’ thirty-two. Facing fourth down and three yards to go, Martz declined to try for a first down and sent in kicker Jeff Wilkins, who made the field goal for a 3–0 lead. Six minutes later, after the Rams’ drive stalled at the Patriots’ thirty-four, Martz faced fourth down with five yards needed. He again chose a field goal attempt; this time Wilkins missed wide left, and the Rams scored nothing. The two decisions exemplified a familiar preference for taking the apparently safer points rather than risking a failed attempt to keep possession.

Economist David Romer’s analysis compared the expected value of those alternatives by accounting for what could happen after either success or failure. On fourth-and-three at an opponent’s thirty-two, if a team had a 60 percent chance of gaining a first down, the expected value of doing so was 1.8 points, while a field goal attempt from the thirty-one was worth barely more than one point. The calculation also included the cost of a failed kick, which would give the opponent possession at its own thirty-five. By translating different outcomes into points, Romer’s method put the options on a common scale instead of treating the kick as the default safe choice.

Analysis Versus Convention

Romer’s recommendations were markedly more aggressive than standard NFL practice. He argued that teams should usually go for a first down on fourth down when they needed three or fewer yards, and should be even more willing to do so between midfield and the opponent’s thirty-yard line. Inside the opponent’s five, his analysis favored going for a touchdown. At the opponent’s two, for example, a field goal offered a near-guaranteed three points, while a touchdown attempt succeeding 43 percent of the time had an expected value of about three points as well. But a failed attempt would leave the opponent pinned at its own two-yard line, adding a field-position benefit that the simple points comparison misses. At midfield, a fourth-and-three attempt with a 50 percent chance of success also favored aggression: success would yield an average gain of six yards, while failure would leave the teams with the ball at roughly the same field position.

Romer acknowledged that his calculations could not capture every consideration, including momentum, and that his league-wide averages might not match a particular team’s abilities. Even with those limits, the difference he found was substantial. For fourth-down plays between midfield and the opponent’s thirty, he considered it worthwhile to go for a first down even when nine yards were needed, given the greater chance of a missed field goal and the relatively small cost of losing possession. With the Rams’ unusually strong offense, the potential value of trying would have been higher still. The evidence does not establish that Martz’s two choices caused the Super Bowl loss; it does make their conservatism look striking in retrospect.

The pattern extended beyond one coach or one game. In the first quarter of the seasons Romer studied, he identified 1,100 plays on which teams would have been better off going for it; instead, teams kicked 992 times. Coaches had strong incentives to win, and aggressive teams adopting the more profitable strategy could gain an edge. Yet the conventional choice persisted. The book suggests that imitation and social proof help explain why: when no one else goes for it, that absence can be taken as evidence that going for it does not make sense. In the NFL, then, the market’s collective forecast and the coach’s customary judgment stand in revealing contrast. One mechanism gathers many independent assessments; the other can reproduce a shared assumption even when quantitative analysis points toward a different play.