Mathematicial inefficiencies of sports betting markets

December 17, 2006 11:59 AM

Mathematicians, I've always suspected there is a real opportunity to exploit the inefficiencies of the online sports betting markets and amateur bettors and make serious money. Have you felt the same way or am I kidding myself?

For example, the odds across different sites for the same event are often substantially different. Is there some sort of arbitrage opportunity there? Also, I have a feeling that bettors (I assume bettors dictate how the odds are created?) overreact to events. By studying the longterm relationship between events and final outcomes and comparing to the odds at the time of the events, couldn't some pattern of inefficiency be identified which could be exploited? Any other statistically-sound ideas you've had about this? (FYI, the three sites I have my eye on are betfred.com, betfair.com and centrebet.com.)

posted by vizsla to sports, hobbies, & recreation (7 comments total)

I realised I unfortunately used "events" in two different contexts in the above. For the arbitrage question, I meant "event" as a sporting occasion (e.g. a particular current or future contest e.g. a boat race, cricket match, football game etc). For the overreaction question, I meant an "event" as a particular occurrence in a contest e.g. points are scored, a competitor is knocked out, a wicket falls, someone is injured etc. Hope that helps clarify.
posted by vizsla at 12:07 PM on December 17, 2006

By studying the longterm relationship between events and final outcomes and comparing to the odds at the time of the events, couldn't some pattern of inefficiency be identified which could be exploited

What you are describing can be studied with hidden Markov models (HMMs).

There is a probability of going from hidden state to hidden state (e.g., "spreads") which yield particular outcomes ("scores").

By modeling how bettors determine their spreads and comparing this with actual empirical outcomes, you might be able to arbitrage an advantage on future outcomes, based on which bettor might overstep its spread.

Some hedge funds use HMMs to place informed bets on various financial markets.
posted by Blazecock Pileon at 12:29 PM on December 17, 2006

Calculated Bets : Computers, Gambling, and Mathematical Modeling to Win
by Steven Skiena is a pretty fascinating book on the subject. The author applied himself to Jai-Lai (he likes Jai-Lai), and made a little money.
posted by hexatron at 12:50 PM on December 17, 2006

There's already services that do this and send text messages to subscribers to inform them of the opportunity (google "sports arbitrage"). There was a paper written on the size and duration of mispricing across different sports ("these arbitrage opportunities persist for a median of 15 minutes").
posted by milkrate at 1:23 PM on December 17, 2006

I wouldn't try this unless you have a CS Masters thesis to write.
posted by delmoi at 3:44 PM on December 17, 2006

The easiest way to do something along the lines you describe is to find a widely read, influential tipster. Grab the best odds available then lay the horse/team at slightly lower odds on a site like Betfair.

In the UK for instance, "Pricewise" from the Racing Post for horses is very influential. Almost all his tips will reduce dramatically in price.
posted by selton at 4:41 PM on December 17, 2006

Take a look at this very interesting Google Techtalk video on the subject. It's called Decision Making and Chance and gives some insight into automatic programs that discover these opportunities. I'm not sure that one could make any money, however.
posted by ontic at 5:20 PM on December 17, 2006