For many sports bettors, the most serious impediment to monetizing the hot hand is the bookie commission. As you may have gathered, sports books are not charities. They take a cut of each bet placed. For a standard point-spread bet, you need to be right 52.38 percent of the time just to break even.
Spread betting was invented by math-teacher-turned-bookie Charles K. McNeil in the 1940s. McNeil’s intention was to manage risk. Though bookies are in the business of selling risk, they don’t much care for it themselves. The point spread is designed to eliminate unhappy surprises—for the bookie, not his customers.
Most gamblers like to bet on a favorite. To attract bets to the underdog, bookies have to offer long odds. This inducement is often inadequate. Bettors still want to wager on the team that’s more likely to win, requiring the bookie to offer longer odds yet on the underdog. That worries bookies, for an upset win could be bankrupting.
McNeil’s solution was to convert every bet into a fifty-fifty proposition. For a wager on the favorite to win, the favorite has to beat the underdog by a specified number of points. Say that Cincinnati is favored over Baltimore with a spread of 4.5 points. (The spread is often quoted in half points to avoid “ties.”) Cincinnati wins the game narrowly, 22 to 20. But since it didn’t “cover the spread”—win by at least 4.5 points—those who wager on Cincinnati lose, and those who bet on Baltimore win.
In McNeil’s conception, the point spread is chosen so that half the money wagered will be on the favorite and the other half on the underdog. The result is a bulletproof business plan. When the bets are balanced, the bookie can skim his vigorish off the top and use the losers’ wagers to pay off the winners. No matter who wins, the bookie never has to dip into his own pocket.
The media often refers to point-spread betting as a prediction market, a term that has lately taken on an aura of infallibility. A spread of five points supposedly means that the median bettor believes the favorite will beat the underdog by five points. It’s the crowd’s prediction.
This is not quite true, and it’s important to understand why. In 2004 economist Steven Levitt (the coauthor of Freakonomics) advanced the theory that bookies set the point spread to maximize their profits—not to run a fancy-schmancy prediction market. They don’t always balance bets exactly, and the point spread might not reflect the crowd’s average opinion. This theory is now called the Levitt model.
I’ll give an extreme example to illustrate Levitt’s point. A crooked bookie knows that the big fight is fixed and the crowd’s favorite is sure to take a dive. The bookie would prefer that as many people as possible bet on the favorite. He would have no interest in balancing bets. The bookie might even want to move the point spread in the “wrong” direction to encourage more bets on the favorite. He knows that every bet on the favorite will lose and be money in his pocket.
Bookies usually aren’t that certain, but the same principle applies whenever they know more than the bettors do. Bookmakers are presumably aware of all the common bettor biases (including hot hand beliefs) and bake them into their point spreads. Levitt showed that there are limits to how far bookies can press their advantage. They have to make sure that a bettor who’s just as smart as the bookies cannot profit from these intentionally misplaced point spreads.
Consider the bet-the-underdog strategy. From 2001 to 2008, NFL underdogs—the teams expected to lose, according to the point spread—won 51.1 percent of resolved bets. (This excludes “tie” cases where no money changed hands because the teams just made the point spread.) A bettor who knows zilch about football and simply wagers on every underdog will win more often than not. He just won’t make any money. The win rate isn’t good enough to beat the bookie commission—and the bookies make sure it’s not.
Studies of real-world betting have confirmed that Levitt was right. Bookies do not stand in awe of the wisdom of bettor crowds. They believe they’re smarter than the crowd, set point spreads accordingly, and make more profit than would be possible with balanced bets. To find out what the betting crowd thinks, then, you can’t just look at the point spread. You’ve also got to look at how much money was bet on each side. Many researchers have done that, and some have examined the hot hand specifically.
One 2011 paper found that betting on an NFL home team increased 3.15 percent when the team had won the past two games. There was no evidence that such hot streaks budged the likelihood of winning the next game. They just increased the willingness to bet hard-earned money.
There is some reason to believe that bet-the-underdog can be made profitable by choosing wagers selectively. Levitt examined bets from a Las Vegas sports book’s promotional tournament. He found that point-spread bets on an NFL underdog playing at home won an amazing 57.7 percent of the time. That’s well over the 52.38 percent needed for a profit net of commission.
To make that absolutely clear: Find an NFL game where the point-spread underdog is playing in its home stadium. Bet $110 on that underdog. When you win, you gain $100 and also get your $110 wager back. If Levitt’s 57.7 percent win rate is typical, you’ll be looking at a 10 percent average profit on every bet.
Why should it be good to bet on a team playing at home? Most likely, bettors forget to factor in the not-entirely-mythic home court advantage. The home team sleeps in their own beds, knows the stadium, and has the hometown media’s support. The visiting team is battling jet lag and the temptation to party in an exciting new city; they aren’t so motivated to perform for a crowd that’s jeering them. This isn’t news to any real gambler, but someone who’s sure the Giants are superior to the Cowboys may not always check the travel schedule before placing a bet.
It’s tough to say whether the casino tournament that Levitt studied was typical. Research on regular betting points in the same direction but is usually less optimistic. A wager on home underdogs would have won 53.3 percent of the time in the NFL (over twenty-one seasons, from 1980 to 2001); 53.2 percent of the time in NCAA college football games (2002 season); 53.0 percent in the National Basketball Association (2002 season). All would have narrowly beaten the bookie commission.
In a 2010 article, Sean Wever and David Aadland, two economists at the University of Wyoming, looked at all the NFL games from 1985 through 2008, some 5,976 matches in all. They found that the higher the spread, the more profitable the underdog strategy was. One likely reason is that teams that are way ahead don’t bother to run up the score in the game’s final minutes. Coaches hate to risk injuries for points the team doesn’t need. This would yield more games where the favorite won but not by enough to beat the spread. That would work to the advantage of those betting on the underdog.
Wever and Aadland describe one strategy that they project to have about a 53 percent chance of winning. You bet on an underdog home team when the spread is at least 6.5 points, or an underdog visiting team when the spread is at least 10.5 points.
For example, the Indianapolis Colts began the 2006 season by winning their first nine games and ended it by winning the Super Bowl. Throughout the season, the media fawned over them. In six games the Colts were a seven-point-or-more home favorite or a ten-point-or-more visiting favorite. Despite that, they covered the spread in only one of these six games. In three games Indianapolis won, but not by the needed margin. Had you been betting on Indianapolis in these high-spread games, you would have lost five games out of the six. Had you been betting against the Colts, you would have won five out of six.
One problem with sports betting systems is that anyone who looks hard enough for a pattern tends to find one. It might be that the Denver Broncos always won on odd-numbered days in February when playing teams named after animals. Analytics is good at finding such patterns. It’s not so good at saying whether you should believe in them.
Wever and Aadland derived their strategy from the period 1985 to 1999 and then tested it on the years 2000 to 2008. In six of those eight seasons, their strategy would have had a profit over the bookie commission. Another season would have been virtually break-even, with a 52.0 win percentage. That bolsters the case that the system’s success is not just a fluke.
On average, about two NFL games a week meet the criteria for the Wever-Aadland strategy. Had it been applied over the entire twenty-three-year time frame, it would have won 53 percent of bets, enough to secure a 1 percent profit net of commission. It’s possible to do better yet by being more selective. By focusing on underdogs with even larger point spreads (and making fewer bets), Wever and Aadland believe it’s practical to achieve a win rate as high as 60 percent.
Publicizing a winning strategy may cause too many people to adopt it, thereby making it useless. That’s worth considering. Ironically, one point in favor of the Wever-Aadland system is that it isn’t all that new. The home and underdog advantages have been known for decades, and analyses of them turn up with some regularly in the Journal of Sports Economics. Most sports bettors aren’t influenced by that discourse. They already “know” how to pick a winning team, from gut instincts and what they see on TV.
Bookies, of all people, must be aware of the home underdog advantage. They seem to have a hard time policing a handful of contrarians while profiting from the great mass of think-alike bettors. Levitt found that three-quarters of bettors pick favorites more than underdogs.
For those who like to gamble for recreation, betting the underdog—especially the long-shot home underdog—can shrink the bookie advantage to the vanishing point. Don’t expect to get rich. Think of it as a way to gamble for free.
The title of this chapter promises advice on outguessing an office football pool. As with a March Madness pool, beating coworkers is a lot easier than beating bookies.
A typical football pool has a weekly list of about ten games. Each bettor must guess the point-spread winner for every match. The point spreads are lifted from the websites of the big Las Vegas sports books, offering the Las Vegas line without the bookie’s cut. The player who predicts the most games correctly wins the pot. The pot is split equally in case of a tie, and some pools give lesser prizes to runners-up.
One drawback is that you can’t cherry-pick the games where you believe you have an advantage. You have to guess the winners of all the games on the list.
In the early 1990s, Emmy’s, a Dartmouth tavern frequented by the math faculty, was running a football pool. Grad student Joseph DeStefano noticed that many players were betting randomly. Some literally tossed a coin to decide which teams to pick. Why not? The point spread is supposed to offer fifty-fifty odds, and it comes close to that. It might seem there’s no way to wring a profit from fair bets. DeStefano realized that there is.
He called his scheme the Evil Twin. It repurposes an idea from the self-correcting codes that cell phones use to communicate with the tower. Signals often get garbled, but the code allows software to guess the original data with amazing accuracy. DeStefano applied that principle to guessing which teams would win.
You don’t have to know the theory to use DeStefano’s system. With the Evil Twin strategy, you make two bets, each the mirror image of the other. If the first bet is for Green Bay to win, the second, “Evil Twin” wager would be for Green Bay to lose—and so on, for every other game in the pool’s list.
It sounds crazy. Normally there is no advantage in placing competing wagers. You may bet on two horses in a race, or on every horse running, and it won’t do you a bit of good. It’ll just draw your bankroll down all the more quickly.
Football pools are different because the objective is to come closest to a hard-to-attain goal. The goal in this case is to guess every game right. With ten games, there are 210 or 1,024 possible outcomes. In a midsize office pool, it’s unusual for anyone to get all the predictions right. Ties are common. When three bettors each guess seven games correctly out of ten, and no one else does better, they split the pot three ways. The advantage of the Evil Twin strategy comes from minimizing ties.
Look at the simplest possible case. There is just one game on the list (Bengals versus Ravens), and just two bettors. Call them Joe and Jane. Each wager costs $1. Joe places two bets (one for himself and one for his imaginary twin) and Jane places one, so the pot comes to $3.
Joe bets on the Bengals and, separately, on the Ravens. Jane picks whichever team she wants. Assuming the point spread is right, Jane has a 50 percent chance of winning. If Jane wins, she will share the $3 pot with Joe. That’s because one or the other of Joe’s two bets has to win. Jane’s share would be $1.50. Jane can expect to win only half the time. Her average winning is half of $1.50, or $0.75.
That’s not so good, considering that she wagered $1. In the long run, Jane can expect to lose $.25 every time she wagers.
That money is vacuumed up by Joe. It can’t go anywhere else. Joe guarantees himself a win by betting both sides. Should Jane pick incorrectly, Joe gets the $3 pot to himself. Otherwise, Joe shares the pot with Jane and gets $1.50. Both scenarios are equally likely. Joe’s expectation is half of $3 plus half of $1.50, or $2.25. Deduct the $2 Joe wagered, and he’s got a $0.25 profit—which of course comes from Jane.
DeStefano used math to show that the Evil Twin scheme always delivers an expectation of profit, no matter how many players are in the pool or how many games they’re betting on. The strategy’s profit peaks when there are four players (five bets, counting the Evil Twin one) predicting an odd number of games. Then the strategy player can expect a 17.2 percent profit on every dollar wagered.
As the number of players in the pool increases, the profit slowly decreases. The expected gain is close to one divided by the number of players in the pool. An easy-to-remember rule: With ten players betting on ten games, the system’s profit is about 10 percent.
These profits are possible because of clustering. When bettors pick winning teams randomly, their choices tend to cluster (like the random bomb hits in the London Blitz). Many bettors will pick the same teams purely by coincidence. By making a bet and its complete opposite, you increase the chance that one will be a contrarian pick, with less competition and less likelihood of a shared jackpot.
On top of the random clustering is a psychological clustering. In general, bettors think alike. They pick favorites, local teams, and teams on winning streaks. None of these have much or any predictive power in a point-spread game, but they do nudge bettors into making similar picks. Office pools are further subject to groupthink. Water-cooler theorizing will enforce shared beliefs about which teams are likely to win. This makes it all the more profitable to challenge the crowd. The Evil Twin works when bettors pick randomly, and it works even better when they don’t.
The Evil Twin strategy applies only to pools where the goal is to pick the most winners in point-spread games (or games where the chances are fifty-fifty). In practice, that limits it mainly to football pools. At Dartmouth, DeStefano’s strategy raised ethical questions as well as mathematical ones. Is it cheating? Most football pools don’t have complicated rules about what’s legal and what’s not. Normally, each player places one bet only. It’s easy to get around that by teaming with another player. You and your partner coordinate bets and agree to split winnings. It would be difficult to prohibit that. Of course, it’s also difficult to prevent coworkers from stealing staplers.
While mulling that over, consider this: The Evil Twin strategy is counterintuitive because, deep down, most serious bettors think they can predict winners. Why stake half your money on a bet that’s the exact opposite of what you believe will happen?
You shouldn’t do that, provided your predictions are sufficiently accurate. Were you 100 percent certain your picks were right, you should stick with them and forget the Evil Twin. But no one should be that confident in football, least of all with point-spread bets.
There is a threshold probability beyond which you don’t need the Evil Twin. This threshold isn’t a very high bar, about 52 percent. Anyone who can pick the winners of games in the pool with much greater than 52 percent accuracy is better off simply placing his own bets and forgetting about the Evil Twin.
That may sound tempting, but remember, you can’t select the games to bet on. Also, the 52-percent threshold is a theoretical value for bettors picking winners randomly from among all the possible permutations. When you factor in psychology, the Evil Twin system probably performs better than the math says.
• Bet-the-underdog strategies appear to offer profits net of bookie commission. One system is to wager only on underdogs playing at home when the spread is 6.5 points or more, or on underdog visitors when the spread is at least 10.5 points.
• The Evil Twin strategy can offer a substantial edge in small-to-medium office pools. Place two bets: one your usual set of picks, and the other an Evil Twin wager picking each of the “losing” teams of the first wager. To avoid suspicion, you might partner with another player and split winnings.