The 2018 FIFA World Cup began a few days ago. It might be easy to know which teams are more likely to advance over others right now. But as the tournament reaches the semifinals and finals, the winning team could be anyone’s guess.
Fortunately, now we have artificial intelligence (AI) to give us a clearer picture of the future.
Place Your Bets on Man or Machine
Usually, those betting on the games turn to professional firms who hire statisticians that analyze databases of past results. These mathematicians return with odds on early games and the final winner. Looking at multiple professional analysis firms, the numbers point to a 16.6% chance for Brazil winning the World Cup, 12.8% on Germany, and 12.5% for Spain. These numbers aren’t very confident, but these three countries have historically taken the trophy many times.
Now, with machine learning, a subset of AI, we can examine the potential winners with literal machine efficiency. And these algorithms don’t always align with what professional statisticians are predicting. Andreas Groll is a researcher at Germany’s Technical University of Dortmund. He and his team developed an AI using machine learning and classical statistics to create a random-forest tree to answer the problem.
Random-forest trees are nothing new, but recently they’ve become very powerful in analyzing huge amounts of data. Unfortunately, as with most algorithms, feeding less data to the system results in strange results. Random-forest trees are no exception. But when given enough info, this method outputs what it thinks are the top factors that will determine the end result.
Two Different Perspectives
Groll’s team simulated each game, found a probability of the winner, and continued this method until they reached the final round. By feeding the algorithm a preliminary set of factors that might affect who wins, like the country’s GDP, population, national teams’ FIFA rankings, average age, and other variables, the algorithm had plenty of data to work with.
The AI was even fed rankings from professional analysis firms, FIFA, and other parties. These rankings turned out to be the algorithm’s most trusted factors for determining the winner. What really didn’t make a difference for the algorithm were factors like population, coach’s nationality, and other random figures.
Game-by-game, this led to the algorithm picking Spain as the forefront choice for winning this year’s trophy, at a probability of 17.8%. Germany, according to the AI, will face a steep uphill battle to make it to the finals, especially compared to Spain. However, if both Spain and Germany reach the finals, the odds are almost 50/50 on the winner.
Probability vs Possibility
When the AI is asked to provide probabilities based on the entire tournament, rather than game-by-game, it produces a different winner. Groll’s team ran the AI through the entire tournament 100,000 times. “According to the most probable tournament course, instead of the Spanish, the German team would win the World Cup,” they conclude.
But this course of events is more unlikely than the game-by-game conclusion because of the number of game permutations the AI ran when it looked at the overall tournament. According to the researchers, the chance of Germany winning is 1 in 100,000.
Soccer buffs know that those types of figures have never stopped Germany from winning the trophy in the past, and rightfully so. The AI doesn’t take into account events like player injuries, which can derail entire teams from winning their round, especially if the injured player is the team leader, like Lionel Messi or Neymar Jr. We personally can’t wait to see what unfolds during this World Cup, especially with these AI predictions in tow.
We’ll Be Back!
It’s a bummer that the U.S. team didn’t qualify for this World Cup, but we have high hopes our team will make us proud when the 2026 World Cup comes around. This one will be hosted by Canada, Mexico, and the U.S.
While we still don’t know which venues will be enlisted to host the 2026 games, San Francisco, Dallas, New York, and Denver are among the 17 cities finalized. Only 10 will be picked in the U.S., while Mexico and Canada will contribute three venues each.
We also can’t help but wonder what AI will be capable of by then, either!Tags: AI, AI and ML, AI App Developer, AI App Development, artificial intelligence, machine learning, ML, San Francisco AI app developers, san francisco mobile app developers, San Francisco mobile app development, sports, World Cup, World Cup 2018