As World Cup 2018 continues, we look at whether a cat called Achilles can predict the outcome better than predictive analytics applications.
You may or may not have heard of Achilles the cat. If you have then you’ll know that has at least something in common with IBM’s Watson supercomputer. Achilles, it seems, who lives in St. Petersburg's Hermitage Museum in Moscow is currently serving as the World Cup 2018 oracle. At the time of writing, Achilles was predicting that Switzerland would beat victory Sweden in the last 16 of the competition, but in the end that match went the other way.
Nevertheless, in earlier matches, Achilles correctly predicted Russia's wins over Saudi Arabia and Egypt, Iran's victory over Morocco and Brazil's defeat of Costa Rica. To date he has only missed one other match, notably when he chose Nigeria as the winner of their Group D match against Argentina. This makes his success rate close to 90 percent, an impressive run in anyone’s book and one that most data scientists would relish.
Why Predictive Analytics Is A Problem
For clarity, and according to Cary, North Carolina-based SAS, predictive analytics is the use of data, statistical algorithms and machine learning techniques to identify the likelihood of future outcomes based on historical data. The goal is to go beyond knowing what has happened to providing a best assessment of what will happen in the future.
Predictive analytics at the best of times is difficult — unless you have a cat gifted with foresight — and enterprises are increasingly turning to predictive analytics applications to predict the likely behavior of customers, the possible success of digital marketing campaigns and even what content is likely to work best in those campaigns. But can it predict the outcome of something like the World Cup that has so many variables in the mix?
While many companies are already using some kind of analytics applications, fewer are using predictive analytics. However, it is already gaining traction in banking and financial sectors, retail, health, insurance and manufacturing — in fact any industry where future risk assessment can impact business.
In the recent Forrester Wave: Customer Analytics Solutions, Q2 2018 (paid for report), SAS was named as leader in the Wave. Wilson Raj, Global Director of Customer Intelligence at SAS, reacting to the company’s iinclusion pointed out that the sheer complexity and volume of customer data require analytics that go beyond traditional reporting to predicting customer intent and behavior at the moment of engagement. “Customer analytics will optimize a brand’s customer experience management activities and enable marketing leaders to drive growth strategies and generate revenue.”
So much for its use, but can it give us the winner of this year’s World Cup, or any other sporting event that has a bunch of variables. Venkat Viswanathan, founder and chairman of San Jose Latent View Analytics argues the World Cup, or other sporting events, are the perfect occasion to demonstrate the power of these applications particularly as the complexity of the different permutations rises. "Predictive algorithms powered by AI or machine learning can certainly be used to predict things like sporting events. After all, people have been crunching sports stats for ages in order to build winning teams and devise game strategy,” he said. “As with all predictive analytics, accuracy boils down to the quality and quantity of data available. The richer the data, the greater the potential for making accurate predictions. AI has the potential to identify patterns in more nuanced data that humans can't typically see.”
He added that some sports inherently provide more data than others based on the number of players and other variables (football or baseball for example vs. tennis). However, despite all of the data in the world, there are always some intangibles — team and player dynamics — and other things that lead to upsets as we've seen in this World Cup with the early elimination of Germany. Here’s how it works. Once the tournament schedule is fed in, ML algorithms are used to create models that analyze past and present data to predict winners of every match. They further use AI to recognize patterns in the algorithms to reiterate the process to ensure a more accurate prediction. They were able to achieve an 88 percent accuracy rate for the group stages using this method.
Humans vs. Analytics
Dr. Siwei Lyu, associate professor at the College of Engineering and Applied Sciences at the New York-based University at Albany, is not sure that analytics applications can predict sport events any better than humans can, let alone cats like Achilles. He said that although the current technology development in AI has achieved some very impressive progress it has yet to be shown that an AI algorithm can make predictions on complex events like the world cup any significantly better than a human can do. “What the current AI technology is good at is to extract regular repetitive patterns from past historic data. This works well for recognizing faces and voices, but a complex social event such as world cup has a lot more factors beyond simple numerical data,” he said.
He added that even taking all information available together may not warrant a reliable prediction, as uncertainty and unexpected is a norm there. However, the next decade may see more exciting development in AI, and this may become reachable with better prediction algorithms, more powerful hardware and bigger knowledge base.
However, if in theory AI can predict the result, the practice is different, according to Gonzalo Rios, Chief Scientist at Chile-based NoiseGrasp. It would only be possible, he said, to create an AI model to predict the winner of the World Cup if we could obtain the historical data for all of the following variables:
The World Cup is a schedule of matches between two soccer teams, in different locations, stadiums, weather, public and arbitration. All these geographic and temporal variables influence the performance of the teams.
Each team is a configuration of different players, a goalkeeper, some defenders, midfielders and strikers, all led by a captain and a technical director.
Each player has a historical performance and its real performance, which can be affected by personal problems, family, illness and psychological pressure.
The technical director makes decisions before and during each game, which may have logical, emotional or political reasons.
The relations between the players determines the synergy of the team, which can produce that the total is more or less than the sum of its parts.
The history of past encounters with a certain team or country can mark a trend or collective psychological pressure that can positively or negatively affect the performance of the game. He did concede, however, that there is always the hidden variable of luck, which can beat any team and any model in the grand finale
One of Europe’s largest online gambling operators, Kindred Futures, has not let all this stop it, though, and recently announced the launch of an AI chatbot developed by Artificial Solutions that goes beyond interacting with keywords to providing a new dimension of gambling. The chatbot built on Artificial Solutions software was launched on Unibet’s Facebook page so folks can place bets and understand the complexities and nuances of gambling language while learning and adapting to provide better answers faster. This is especially interesting because of the World Cup games. On the surface, this may seem like another chatbot that can provide stats on game times and scores with integrations like Amazon Echo or Alexa, but according to a statement the company sent CMSWire, it goes beyond that. It's sais that the chabot can do the following.
Uses AI conversational abilities to understand natural gambling language, an increasing trend in the US and globally
It enables customers to find odds and place bets — a more personalized data-heavy experiences for fans.
The Kindred Futures is testing the AI Chatbot within its Unibet World Cup Russia 2018 inventory so we will see after the competition what it managed to do.
Predict Odds Not Outcomes
Lakshmi Subramanian, an associate professor at NYU and co-founder of Entrupy, creators of an AI solution for product authentication. He points out that when it comes to sporting events like the World Cup, artificial intelligence algorithms are very good at giving odds. They’re less effective at predicting outcomes. “This year’s World Cup is great case in point — so many of the results seem to defy the odds,” he said. “We’ve seen teams with better odds lose, even though they had better stats in almost every other category, including possession, shots on goals, corners etc. Still, they fell short when it comes to the only number that really matters in the end - the final number of goals."
In the light of all this it may be that, at least for sporting events, if predictive analytics can’t predict the outcome then maybe they should look at finding a cat that can. For customer intent and marketing campaign success likelihood, it’s probably best to stick with one of the many predictive analytics applications that are on the market at the moment.
About the Author
David is a full-time journalist based in Paris, who spends his time working between Ireland, the UK and France. A partisan of ‘green’ living and conservation, he is particularly interested in information managementand how enterprise content management, analytics, big data and cloud computing impact on it.