The Perfect Bet

Home > Nonfiction > The Perfect Bet > Page 11
The Perfect Bet Page 11

by Adam Kucharski


  Betting might seem somewhat distant from other types of investment, but that is one of its selling points. During the 2008 financial crisis, many asset prices fell sharply. Investors often try to build a diverse portfolio to protect against such shocks; for example, they might hold stocks in several different companies in a range of industries. But when markets run into trouble, this diversity is not always enough. According to Tobias Preis, a researcher in complex systems at the University of Warwick, stocks can behave in a similar way when a financial market hits a rough period. Preis and colleagues analyzed share prices in the Dow Jones Industrial Average between 1939 and 2010 and found that stocks would go down together as the market came under more stress. “The diversification effect which should protect a portfolio melts away in times of market losses,” they noted, “just when it would most urgently be needed.”

  The problem isn’t limited to stocks. In the run-up to the 2008 crisis, more and more investors began to trade “collateralized debt obligations.” These financial products gathered together outstanding loans such as home mortgages, making it possible for investors to earn money by taking on some of the lenders’ risk. Although there might have been a high probability that a single person would default on a loan, investors assumed it was extremely unlikely everyone would default at the same time. Unfortunately, this assumption turned out to be incorrect. When one home lost its value during the crisis, others followed.

  Advocates of sports betting point out that wagers are generally unaffected by the financial world. Games will still go ahead if the stock market takes a dive; betting exchanges will still accept wagers. A hedge fund that concentrates on sports betting should therefore be an attractive investment, because it provides diversification. It was this idea that persuaded Brendan Poots to set up a sports-focused hedge fund in 2010. Based in Melbourne, Australia, Priomha Capital aimed to give the public investors access to the traditionally private world of sports prediction.

  Creating good forecasts can require additional expertise, so Priomha linked up with researchers at the Royal Melbourne Institute of Technology. To some extent, the approach is a twenty-first-century version of the Computer Group’s strategy. Priomha creates a model for a particular sport, runs simulations to predict the likelihood of each result, and then compares the predictions with the current odds on betting exchanges such as Betfair.

  The big difference is that investors are not restricted to betting before a game starts. Which is good news, because Poots has found that odds generally settle down to a fair value in the run-up to a fixture. “Come kick off, the market’s pretty efficient,” he said. “But once play starts, that’s where we have a huge opportunity.”

  When it comes to soccer prediction, “in-play” analysis was always the natural next step. After working on final score predictions in 1997, Mark Dixon turned his attention to what happens during a soccer match. Along with fellow statistician Michael Robinson, he simulated matches using a similar model to the one he’d published with Stuart Coles, but with some important new modifications. As well as accounting for each teams’ attacking strength and defensive weakness, the model included factors based on the current score and time left to play. It turned out that including in-play information led to more accurate predictions than the original Dixon-Coles model.

  The model also made it possible to test popular soccer “wisdom.” Dixon and Robinson noted that commentators would often tell viewers that teams were more vulnerable after they scored a goal. The researchers referred to this cliché as “immediate strike back.” The idea is that after a goal goes in, attackers’ concentration wobbles, which can allow the opposition back into the game. But the cliché turned out to be misguided. Dixon and Robinson found that teams weren’t especially vulnerable after they have scored a goal. So, why did commentators often claim that they were?

  If we come across something unusual or shocking, it stands out in our mind. According to Dixon and Robinson, “People have a tendency to overestimate the frequency of surprising events.” This doesn’t just happen in sports. Many worry more about terrorist attacks than bathtub accidents, despite the fact that—in the United States at least—you’re far more likely to die in a bathtub than at the hands of a terrorist. Unusual events are more memorable, which also explains why people think it’s easier to become a millionaire with a one-dollar lottery ticket than by playing roulette repeatedly. Although both are terrible ideas, in terms of raw probability, playing roulette again and again is more likely to generate a lucky million dollars in profit.

  Betting successfully during a soccer match means identifying human biases like these. Are there certain aspects of the game that gamblers consistently misjudge? Poots has found that a few things stand out. One is the effect of goals. Just as Dixon and Robinson noted, the popular view is not always the correct one: a goal doesn’t always create the shock people think that it does. Gamblers also tend to overestimate the impact of red cards. That isn’t to say they don’t have any effect. A team playing against an opponent with ten men will probably score at a higher rate (one 2014 study reckoned the rate would be 60 percent higher on average). But the odds often move too far, suggesting gamblers mistake a difficult situation for a hopeless one.

  Following a dramatic event, the odds available on a betting exchange gradually adjust to the new situation. When things have settled down, Priomha can hedge its bets by taking the opposite position. If it backed the home team to win at long odds, perhaps after a red card, it will bet against them when the odds decrease. This way, it doesn’t matter how the game ends. Like a trader who buys an item from a panicked seller and later sells it back at a higher price, the team closed their position and offloaded any remaining risk.

  There are plenty of opportunities to capitalize on inaccurate odds during a game. Unfortunately, there are also fewer bets on offer, which means Priomha has to be careful not to disrupt the market with large wagers. “During play, you have to drip feed your money in,” Poots said. In fact, the size of the market is one of the biggest obstacles facing funds like Priomha. Because it makes money by identifying incorrect sports odds, the more money it has to invest, the more erroneous odds it has to find.

  The current plan is to manage up to $20 million of investors’ money. Poots pointed out that if they tried to handle a much larger amount—such as $100 million—it would be a struggle to make reasonable returns. They might be able to find enough opportunities to bring in a 5 percent annual return, but as a hedge fund, they really want to be making double figures for their investors, and they are more likely to achieve this if they constrain the size of the fund.

  Although Priomha has not reached its limit yet, as the fund grows Poots is noticing a change in who is buying into the strategy. “Our investor profile used to be someone who likes sport, and liked to have a bet,” he said. “It’s now becoming people who have got their pension or other funds to invest.”

  Priomha is not the only sports betting fund to have appeared in recent years. The London-based Fidens syndicate opened its fund to investors in 2013; two years later it was managing more than £5 million. Mathematics graduate Will Wilde heads up Fidens’s trading strategy. This involves betting on ten soccer leagues around the globe, placing around three thousand wagers per year.

  Stock market investing has often been compared to gambling, especially when shares are held for only a short period of time. There is a certain irony then, that gambling is increasingly seen as a viable option for investors. Not all sports betting funds have been successful, however. In 2010, investment firm Centaur launched the Galileo fund, which was designed to allow investors to profit from sports betting. The plan was to attract $100 million of investment and generate an annual return of 15 to 25 percent. The finance community watched with interest, but two years later the fund folded.

  Although the ambitions of funds like Priomha are currently constrained by the size of the betting market, things could be very different if sports betting were to expand in t
he United States. “If America was to open up,” Poots said, “the whole game changes completely.” The first major hints of change came soon after Priomha was founded. Following a referendum in 2011, New Jersey governor Chris Christie signed a bill legalizing sports betting in the state. For the first time, gamblers in Atlantic City would be able to bet on games like the Super Bowl. That was the theory at least. It did not take long for professional sports leagues to bring in lawyers to halt the expansion. The case has been bouncing through the court system ever since, the main obstacle being a federal law from 1992 that prohibits sports betting in all but four states. Opponents say gambling should be limited to places like Las Vegas; New Jersey claims the law is unconstitutional and that the public supports legalized betting. Indeed, many sports leagues already allow people to put money on their predictions coming true. Every year people pay to take part in fantasy sports leagues, even though betting on a specific match outcome is still illegal.

  Advocates for law changes say there are two main advantages to legalized gambling. First, it would generate more tax. It’s been estimated that less than 1 percent of sports bets in the United States are placed legally. The remaining 99 percent of wagers, made through unlicensed bookmakers or offshore websites, probably run into hundreds of billions of dollars. If these bets were legal, the tax revenue would be enormous. Second, legalization means regulation, and regulation means transparency. Bookmakers and betting exchanges keep records of customers, and online firms also have bank details. According to NBA commissioner Adam Silver, legalizing gambling would bring the activity into view of government scrutiny. “I believe that sports betting should be brought out of the underground and into the sunlight where it can be appropriately monitored and regulated,” he wrote in the New York Times in 2014.

  Betting syndicates would also stand to benefit from legalization. With more bookmakers taking wagers, syndicates could place bets on a much grander scale. There is also a chance that new laws will allow syndicates to bet in Las Vegas. Currently, if gamblers want to bet on sports in the city, they still need to turn up at a casino with a handful of cash, which makes it difficult to systematically place large bets. In 2015, the Nevada senate passed a bill that would allow a group of investors to back a bettor, which is essentially what Priomha already does outside of the United States. If the bill gets through the state assembly and becomes law, many more sports hedge funds could appear. Other countries are also debating new gambling laws. In Japan, sports bettors can currently put money only on horse, boat, or cycle races. A new bill, submitted in April 2015, and supported by the prime minister, proposes to change that. New opportunities will also arise in India and China, as informal betting markets become more regulated.

  According to sports journalist Chad Millman, it is not just established gamblers who would be well positioned to profit from law changes. During a visit to MIT in March 2013, Millman got talking to Mike Wohl, an MBA student at the university’s business school. For his study project, Wohl had considered gambling as “the missing asset class.” Wohl had a background in finance, and his analysis—along with his personal experience of betting—suggested that sports wagers could produce as good a trade-off between risk and return as investing in stocks could.

  Millman pointed out that there are two extremes to the gambling spectrum. At one end are professional sports bettors, the so-called sharps who regularly place successful bets. At the other are the everyday gamblers, who don’t have predictive tools or reliable strategies. In between, Millman says, are a number of people like Wohl who have the necessary skills to bet successfully but haven’t yet chosen to use them. They might work in finance or research; perhaps they have MBAs or PhDs. If sports betting was to expand in the United States, these small-scale bettors would be in a good position to profit. With their quantitative backgrounds, they are already familiar with the crucial methods. They also have the necessary tools, thanks to increases in computing power and data availability. All they need now is the access.

  THERE ARE CERTAIN ADVANTAGES to being a betting start-up. For one, it means more flexibility. But should new syndicates follow sports betting strategies that have already been successful? Or should they exploit their flexibility and try something else?

  In retrospect, Michael Kent would look at matches in far more detail. “If I was starting over right now,” he said, “I would want to have play-by-play data.” The additional information would make it possible to measure individual contributions. This would be a stark contrast with his previous analysis: in his models, Kent has always treated teams as a single entity. “I have no knowledge of players,” he said. “I know what the team did, but I don’t know the name of the quarterback.”

  Some modern betting syndicates go to great lengths to measure individual performances. “We do analysis on the effect of every player in every team,” Will Wilde said. “Every player has a rating that goes up or down, regardless of whether they play or not.” In Hong Kong, Bill Benter’s syndicate even employs people to sift through videos of races. They might look at how a horse’s speed changes during the race or how well it recovers after a bump. These “video variables” make up a relatively small part of the model—about 3 percent—but they all help nudge the predictions a little closer to reality.

  It is not always just a matter of collecting more data. In soccer, successful defenders can be a nightmare for statisticians. During his years playing for Milan and Italy, Paolo Maldini averaged one tackle every other game. It wasn’t because he was a lazy player; it was because he didn’t need to make many tackles. He held back the opposition by getting into the right positions. Raw statistics such as number of tackles can therefore be misleading. If a defender makes fewer tackles, it doesn’t always mean he’s getting worse. It could mean he’s improving.

  A similar problem crops up with cornerbacks in US football. Their job is to patrol the edges of the field, defending against attacking passes by the opposition. Good cornerbacks intercept lots of passes, but great ones won’t need to: the other team will be trying to avoid them. As a result, the best cornerbacks in the NFL might touch the ball only a handful of times per season.

  How can we measure a player’s ability if they rarely do anything that can be measured? One option is to compare the overall team performance when a player is and isn’t on the field. At the simplest level, we could look at how often a team wins when a certain individual is playing. Sometimes it is clear that a player is valuable to a team. For example, when striker Thierry Henry played for Arsenal soccer club between 1999 and 2007, the team won 61 percent of matches he appeared in. On the other hand, they won only 52 percent of the games he missed.

  Counting wins is simple enough, but measuring players in this way can raise some unexpected results. In some cases, it might even appear that fan favorites are not actually that important to the team. Since Steven Gerrard made his first appearance for Liverpool in 1998, they have won half the games he’s played in. Yet they have also won half the games he hasn’t had a role in. Brendan Poots points out that the best clubs have strong squads, so can often cope with losing a star individual. When top players go off injured, teams adjust. “In the sum of the parts,” Poots said, “the effect that they have—or their absence has—is not as great as people think.”

  The problem with simply tallying up wins with and without a certain player, however, is that the calculation doesn’t account for the importance of those games or the strength of the opposition. Teams often field more big-name players in crucial matches, for example. One way to get around these issues is to use a predictive model. Sports statisticians often assess the importance of a particular player by comparing the predicted scores for the games the player played in with the actual results of those games. If the team performs better than expected when that player is on the field, it suggests the player is especially important to the team.

  Again, it’s not always the best-known players who come out on top. This is because identifying the most important
player is not the same as finding the best player. The most important player—as judged by the model—might be someone without an obvious replacement or a player whose style suits the team particularly well.

  To interpret the results of their predictive models, firms working on sports forecasts employ analysts with a detailed knowledge of each team. These experts can suggest why a certain player appears to be so important and what that might mean for upcoming matches. Such information is not always easy to quantify, but it might have a big effect on results. The trick is to know what the model doesn’t capture and to account for such features when making predictions. Sports statistician David Hastie points out that this goes against many people’s idea of a scientific betting strategy. “There is a common perception that betting is all about models,” he said. “People expect a magic formula.”

  GAMBLERS NEED TO KNOW how to get at crucial information, whether it is quantitative, as is the case with model predictions, or of a more qualitative nature, as with human insights. Although well known for his computer models, Kent knew the importance of human experts when making predictions. He received regular updates from people with in-depth knowledge of certain sports, people whose job it was to know things that the model might not capture. “We had a guy in New York City who could tell you the starting lineup for 200 college basketball teams,” he said.

  Making better predictions about individual players doesn’t just benefit gamblers. As techniques improve, bettors and sports teams are finding more common ground, drawn together by a common desire to anticipate what will happen in the next season, or the next game, or even the next quarter. Every spring, team managers chat with statisticians and modelers at the MIT Sloan Sports Analytics Conference. Prediction methods can be particularly useful when teams go scouting for new signings. Historically, assessing a player’s value has been difficult because performances are subject to chance. A player might have an impressive—and lucky—season one year, and then have a less successful time the next.

 

‹ Prev