Ů��С��Ƶ

Fancy a flutter? The chances of making a profit by betting on football matches are extremely low, but a trio of researchers has managed to beat the odds with a simple formula.

Mathematicians had already developed bookie-beating models that attempt to predict the outcome of sports matches, but they are difficult to devise and don’t always perform consistently. Lisandro Kaunitz at the University of Tokyo and his colleagues wanted to know if a more direct approach would work: using the bookmakers’ own odds against them.

The team studied 10 years’ worth of data on nearly half a million football matches and the associated odds offered by 32 bookmakers between January 2005 and June 2015. For every game, they looked for odds that might yield a better return than the average offered by bookies – say, 5 to 1 versus a mean of 2 to 1.

“Using the odds that are published, you can get a very good estimate of an event occurring,” says team member Javier Kreiner, a data scientist at transport start-up CargoX in São Paulo, Brazil. “What’s the probability of Barcelona winning against Real Madrid, for example.”

Mean odds of 2 to 1 suggest the bookies collectively think this reflects fair odds for that outcome. But 5 to 1 offers higher returns should the outcome materialise. The team used the historical data to work out the optimal distance from the mean odds – the one that would give a positive payout for the largest number of games.

When they applied their strategy in a simulation, they made a return of 3.5 per cent. Making bets randomly resulted in a loss of 3.32 per cent.

Into real time

Then the team decided to try betting for real. They developed an online tool that would apply their odds-averaging formula to upcoming football matches. When a favourable opportunity arose, a member of the team would email Kaunitz and his wife, one of whom then placed a bet. They kept this up for five months, placing $50 bets around 30 times a week.

And they were winning. After five months the team had made a profit of $957.50 – a return of 8.5 per cent. There were some administrative costs, such as the expense of running servers to keep track of the odds in real time. But Kaunitz and his wife did get to splash out on a nice dinner in Tokyo, he says. “We were excited but it’s worth mentioning – you need to spend a lot of time to do it.”

But their streak was cut short. Following a series of several small wins, the trio were surprised to find that their accounts had been limited, restricting how much they could bet to as little as $1.25.

A spokesman for William Hill, one of the bookmakers used by the team, said betting is sometimes restricted “in a small number of cases”.

“This can be for a number of reasons, including bonus abuse and taking proportionately more than their fair share of special offers and enhanced prices, which are designed for the many rather than a few,” he says.

The gambling industry has long restricted players who appear to show an edge over the house, says Mark Griffiths at Nottingham Trent University, UK. A classic example is card-counting, which can help players win at blackjack. Casinos are quick to expel those who try it and it is sometimes flagged as cheating. But it’s not illegal in the US or the UK.

Trapping bots

Griffiths compares the team’s approach to card-counting. “It’s not cheating at all – it’s using mathematics to try and beat a particular system,” he says.

Online bookmakers have very tight profit margins, so some offer generous odds for very short periods of time on certain matches to lure automated or expert betting systems out into the open, says economist David Forrest at the University of Liverpool, UK. “They will try to trap the robots,” he says. “Anyone who responds has their account closed.”

Some of the researchers’ accounts may have fallen foul of this system. In some cases they found they could not bet at all after signing up with certain bookies. If so, then this suggests their technique really was finding the best odds out there. But the house still always wins in the end.

Reference: