Equity Research Analyst
With the start of the Premier League season, chances are you will have seen BT Sport’s latest advert for the beautiful game. Using machine learning, a type of artificial intelligence, BT has attempted to predict the course of the 2019/20 Premier League campaign. While it’s the results that will capture most people’s attention, it’s the use of AI that caught my eye.
Machine learning is simply pattern recognition done by computers, so what BT, Google and Opta have done is not technically difficult. But their collaboration does highlight the three things you need for good machine learning – a clean dataset, access to experienced – and scarce – data scientists, and a proper idea of what you want the machine learning to achieve, in this case, a clever bit of marketing/PR.
While I am by no means a football expert, the results don’t appear to be that different to what happened last year. Manchester City finish top, Liverpool a reasonably close second, with Tottenham and Chelsea in third and fourth place respectively. Of the three teams that were promoted in 2019, two get relegated, with Aston Villa the only one staying up. That should keep some of my colleagues in the Birmingham office happy at least.
The results ultimately show how far machine learning has come. Twenty years ago, machine learning would have been used sparingly, requiring vast supercomputers that only a few governments or corporates had access to. But over the past few decades, we have seen a vast improvement in computing power and a significant improvement in the availability of computer power. You can now rent a supercomputer through services like Amazon Web Services or Microsoft Azure, and it’s not even that expensive to do so.
Machine learning is now advancing on two fronts, becoming more widely used and able to deal with more complex problems. The next frontier is something called reinforcement learning, where a machine carries out a series of tasks to maximise its reward in a particular situation. This might be useful when you want to find the best behaviour or path to take in some scenario, including when the only way to collect information about an environment is to interact with it.
Alphabet, Google’s parent company, is widely seen as one of the machine learning experts, with the most data scientists of any company globally. There are also some interesting UK businesses, such as RELX, which provides information to insurance, legal and scientific and medical businesses and institutions.
Whereas it is the technology names that have driven the development and adoption of machine learning over the past decade, the potential beneficiaries are now spreading beyond the technology sector. It will be an interesting area to watch over the next five years, but be prepared for plenty more press releases like BT Sport’s.