About the Blog

The FICO Labs blog focuses on innovation, technology and the role of analytic sciences in today’s business world. We’ll share our perspectives, vision, successes and challenges in the areas of predictive analytics, Big Data, decision optimization, cloud computing and Big Marketing.

« Understanding Individual Consumers Goes “Back to the Future” | Main | Location-Based Marketing Using GPS Data »

Covering the Spread

Football 2

By Mike Farrar

A few weeks ago, we wrote about why analysts can’t predict the outcome of the Super Bowl. One related note that I wanted to cover: the odds makers. 

Markets don’t only exist to trade goods or services. Markets also exist to exchange information and express it in prices. Sports betting is an information market, exchanging information about what the general betting population believes will be the outcome of a game. It probably gets closer to a consensus 50-50 prediction on sporting events than anything else.

It works in football because betting on a game involves a point spread.  A team that is favored to win a game will need to “cover the spread,” i.e.  outscore the underdog by more points than what’s expected, before a bet on the favored team will pay out. If the underdog wins, or if the favored team doesn’t cover the spread, bets on the underdog pay out. 

Bookies try to hedge risk by balancing bets on one team with counter bets on the other team. If the point spread is out of whack as far as bettors are concerned, everyone piles in on one side of the bet, and the bookie has to change the spread to get anyone to be willing to take the other side. 

For example, the point spread on this year’s Super Bowl opened with the Ravens as 5-point underdogs. Enough bettors believed that the Ravens would stay within 5 points of the 49ers that bookies adjusted their point spread from 5 points to 4.5 points. As time went on, the betting line continued to shift in Baltimore's favor. Today many bookies are setting the spread at 3.5 or 4 points.

There’s a bit of inefficiency in the market. The bookie has to take her cut to be willing to do that job, but she’s in it for the vigorish, not for winning bets. She just wants her bets hedged evenly on both sides, and plenty of them – the more bets, the more vig she earns.

It’s exactly the same role that specialists play on the floor of the NYSE. Specialists match buy and sell orders until nobody piles unduly onto one side of the trades. If markets trade information, specialists (and bookies) provide liquidity of information, without which nobody would know where to place their bets.

Thus, just as in the stock market, the line on a game represents the betting market’s current best guess at a 50-50 chance that either team will win.

This has implications for predicting all sorts of uncertain outcomes. For more than twenty years, the University of Iowa has run the Iowa Electronic Market (IEM).  Intrade is another with even broader scope. These markets allow participants to trade futures contracts on the outcome of political contests. You can make a bet on a candidate by buying a contract at a certain price. If you think the market is overestimating a candidate’s chances, you sell contracts on that candidate, and vice-versa if you think the market is misunderestimating her. The market price shows the general consensus of an election’s outcome. 

It’s even possible to arbitrage information between such markets! There are multiple other sites that accept bets on elections, and their prices don’t always agree with each other.  A savvy trader can buy low in one market and sell high in another. Until prices synchronize, it’s free money for the trader.

For more on the dynamics of markets such as the IEM, you might want to read frequent Freakonomics blogger Justin Wolfers. Nate Silver also has a very good chapter on the power of prediction markets in his recent book The Signal and the Noise. For the truly geeky among us, it’s a short trip from there to a better understanding of the efficient market hypothesis and why casinos can afford to build billion-dollar facilities. I recommend Burton Malkiel’s classic A Random Walk Down Wall Street. For a more lighthearted treatment of efficient markets, see The Money Game by Adam Smith. Look for the chapter, “What the Hell is a Random Walk?”

So why can’t an individual analyst or individual bettor predict football games particularly well? Much of it comes from the unexpected things that affect the game: weather, matchups, injuries, and so forth. There’s a lot of noise, and bettors and analysts will inevitably have different estimates of their likely effect on the game.

Make no mistake – each bettor is trying his hardest to win. He makes money by being right. Each bettor has his own personal opinion, but it’s only his opinion, and every other bettor has his own opinion, too. But lots of bettors create lots of opinions, and their combination is what makes bookies set the spread on a game.

I’m not going to go out on a limb and predict the outcome of this year’s Super Bowl. I’m certainly not going to place any wagers on it. Once you pay the bookie her share, you lose money even if you’re right half the time, and the point spread only gives me a 50-50 chance anyway.

Plus, the Bears aren’t playing, so who cares?

We’ll get ‘em next year.

First time on the FICO Labs Blog?
Subscribe to the FICO Labs Blog Feed or check out some other recent posts:

Comments

Post a comment

If you have a TypeKey or TypePad account, please Sign In.

Search Site


FICOLabsBlog.fico.com

Subscribe

Enter your Email

Preview | Powered by FeedBlitz