Like millions of other people, Alice Simpson loves playing fantasy football.
Every season she picks a team of real-life players from England’s Premier League, and she gains – and loses – points according to how each of her players performs in their games.
Each week she can make a substitution – bring in a new player, and get rid of another. It is a very competitive hobby, and the fantasy managers take it very seriously.
To do well you need to be very knowledgeable about football, and follow the Premier League extremely closely. At least that used to be the case.
In recent years, managers have been able to turn to a number of providers of fantasy football artificial intelligence (AI) software programs. These programs do the studying for you, and suggest the best footballers for you to bring in.
It is very much the same with the American football version of the game. Yet, is it unsporting to use such systems?
Ms Simpson has been playing fantasy football for six years, and started using AI in 2018 to gain an edge on her friends in their mini-league.
“I see the AI as a source of information, maybe like talking to a well-informed friend for advice,” says the 31-year-old teacher, from Wiltshire. “But I think the best thing about AI, is that it helps you remove any biases.”
“Often, we may have a grudge against a player as they did not perform well for you the last time you owned them, or maybe they play for a rival team to one you support,” says Ms Simpson.
Currently in second place in her league as the 2021-22 football season draws to a close, she gets her AI assistance from one of the UK’s most popular providers – Fantasy Football Fix. Offering both a free and subscription-based premium service, it launched back in 2018, and says it now has 500,000 users.
Its self-learning software trawls through all the mass of data surrounding each and every Premier League player and team, to try to predict their future performance. And from this it suggests that fantasy football managers pick and change certain players.
Fantasy Football Fix’s Tom Brown claims that the tech works so well that when we “run bots using our AI tech they finish in at least the top 1% of all the nine million [Premier League fantasy football] managers”.
Yet, it’s not infallible, as Mr Brown’s colleague Adam Moss admits, the AI algorithms can get tripped-up, if a real life Premier League football manager rotates their players unpredictably, such as Manchester City’s Pep Guardiola.
“Despite all our efforts, there’s basically no rhyme or pattern for how he does things, and that makes it hard when you try to implement an algorithm,” says Mr Moss.
Mr Brown adds that AI can however, make very good predictions – “if it knows what players are going to be on the pitch”.
“But often, someone like Pep will decide randomly to drop one of his players for someone else, and there’s basically no way of predicting that sort of thing.”
If fantasy football is popular in the UK, with more than nine million players of the official Premier League game, it has nothing on the popularity of similar games in North America.
In the US and Canada, more than 59.3 million people play fantasy sports. Of that number, 78% play the American football or NFL (National Football League) version, while 39% play baseball and 19% basketball.
One of the most popular fantasy American football leagues is provided for free by broadcaster, ESPN.
Since 2020, users have been able to access AI assistance on which players to trade. This is thanks to a collaboration between ESPN and computer firm IBM, which asked its AI computer, Watson, to start studying the NFL.
The ‘Trade Assistant with IBM Watson’ function uses AI to not just trawl though all the available NFL stats, but it also listens to TV and radio shows, and podcasts to gauge the opinion of experts and other commentators.
It then uses all this information to suggest that a fantasy football manager makes a player change, and explains its reasoning.
Aaron Baughman, IBM’s AI and hybrid cloud lead engineer, is one of the architects behind the tie-up. He says that fantasy American football AI requires advanced algorithms “because this is a hard problem to solve”.
A keen player of fantasy American football himself, he says that he can enjoy the game more now that “the heavy lifting is done by AI”.
But is it fair to use AI to boost your performance in fantasy football or another sport? James Pritchard, a keen fantasy football player from North Wales, says that it certainly isn’t for him.
As the current 2021/22 season nears to an end, his team – Locomotive Llandudno – is top of a mini-league of 18 friends.
“I wouldn’t ever consider using AI,” says the 49-year-old public relations consultant. “It is ungentlemanly, and it feels suspiciously close to cheating.
“And anyway, I trust my judgement regarding football against any computer. I watch quite a lot of football, and keep a very close eye on the Premier League.
“It is all about bragging rights over my friends, and if I used AI I wouldn’t have them.”