AI is getting bigger and more visible by the minute, and its capabilities are becoming much clearer: in the past few years, we’ve seen AI beat the best Go player in the world, as well as a selection of the best No Limit Hold ‘em cash game players in the world (although one guy made it through all right). This year, AI seems to be edging in on the bookies’ territory, with multiple reports of hundreds of thousands of AI being run to predict, among other things, the World Cup.
Investment bank Goldman Sachs ran a 1 million iteration World Cup AI simulation, featuring 200,000 models of team and player data as information sources to draw from. Match scores were predicted in terms of raw average number of goals scored. Their models predicted that Brazil will beat Germany in the final, 1.70 – 1.41. The AI also has England’s Cup run set to end vs. Germany in the quarter finals, a stage where the AI predicts we will also likely lose Spain and Argentina. Saudi Arabia, rather than Russia, is picked as the outlier, with a chance to go beyond the group stages.
“We capture the stochastic nature of the tournament carefully using state-of-the-art statistical methods and we consider a lot of information in doing so,” a spokesman for Goldman Sachs said. “But the forecasts remain highly uncertain, even with the fanciest statistical techniques, simply because football is quite an unpredictable game. This is, of course, precisely why the World Cup will be so exciting to watch.”
(Editor’s note: Neither Saudi Arabia nor Germany made it out of the round of 32, Argentina lost to France and Spain got knocked out by Russia in the round of 16.)
The bank has some stiff competition from Andreas Groll and his colleagues at the Technical University of Dortmund in Germany. Groll has put together a so-called “random forest approach” which uses a combination of machine learning and statistics to arrive at a different favourite to win the Cup: Spain, with a probability of 17.8%. The researchers concluded that Germany would face tougher opposition at an earlier stage, making them more likely to fall in the quarterfinals despite being the stronger team. However, using a full simulation derived from the random forest approach yielded a different answer – Germany to win.
The tournament is also making history by featuring the first use in the World Cup of the Video Assistant Referee – a method of real-time video review that can be triggered by either a field ref or the VAR – but at some stage AI is going to break into that domain as well. IBM Watson, meanwhile, recently cut a deal with Fox Sports to provide AI-based video editing techniques in order to help fans prepare and access thousands of hours of football coverage.
AI isn’t about to edge out the refs and the bookies just yet, but don’t blink – you might miss the singularity.
