The World Cup is one of the most bet-upon sporting events in human history, but as fans agonize over whether Brazil has the edge over Spain, or if England can reach the quarter finals, a super computer might have the whole thing locked up already — and its prediction does not match those of bookmakers. Spoiler alert: look away now if you don’t want to know what he came up with, because we’re about to reveal his conclusion…
Dr. Andreas Groll of the Technical University of Dortmund in Germany — who programmed the sports betting bot — believes employing the latest artificial intelligence (AI) machine-learning techniques could be the most accurate way to predict winners of sporting events.
But how did the professor’s AI machine-learning genius get there?
100,000 World Cups
Groll programed in a number of potential factors that might decide the outcome of a World Cup. It was then up to the computer to figure out which of these were the most important by analyzing past results.
These factors included things such as a country’s GDP, population, and FIFA rankings, as well as the properties of the teams themselves, including their average age and the number of Champions League players they have.
With all these variables plugged into the machine, Groll simulated the World Cup 100,000 times, and found that Spain was most likely to win the thing, with a 17.8 percent chance.
Groll noted that one of the biggest factors is the structure of the tournament itself. If Germany clears the group phase, it is more likely to face strong opposition in the next game, which means it only has a 58 percent chance of reaching the quarter finals, as opposed to Spain’s 73 percent chance, because Spain is unlikely to face strong opposition in the Round of 16.
If Germany does reach the quarter finals, however, it suddenly becomes the slight favorite over Spain. That means Spain is worth a bet now, but if Germany makes it to the final eight, lump some money on them, too.
And if all this seems a little confusing, there are always clairvoyant animals to fall back on. This year, it’s a deaf cat from St. Petersburg in Russia called Achilles, who chooses winners by selecting bowls of cat food marked with national flags.
He successfully predicted that Russia would win its opening game, but Achilles’ results since then have been spotty at best.
Before Achilles, there was Paul the Octopus, the psychic cephalopod who amazed the world with his unerring World Cup predictions in 2010. Paul scored an 85 percent success rate when he was asked to predict the outcome of games by embracing the winning country’s flag in his rubbery tentacles.
For those unfamiliar with machine learning, it’s a branch of AI that usually uses statistics to give a computer the ability to “learn,” through trial and error, by simulating possible eventualities hundreds of thousands — or in some cases, billions or trillions — of times to come up with the likeliest answer.
The machine learning super computer Libratus, for example, was able to teach itself heads-up no limit hold’em poker, armed with nothing but the rules and arriving at an almost-perfect strategy after playing trillions of hands against itself.
Last year, it became the first computer to beat human professional players over the course of180,000 hands at the heads-up no limit hold’em format.