Tuesday 19 August 2025 1:00 pm
 |  Updated: 

Tuesday 19 August 2025 10:20 am

Share

Last week City AM revealed that two thirds of Fantasy Premier League players would use AI to help pick their points-based teams throughout the season.

Last week City AM revealed that two thirds of Fantasy Premier League players would use AI to help pick their points-based teams throughout the season.

And it got us thinking; how accurate could AI be in predicting the outcome of Premier League football matches?

So in that spirit, hard earned cash was siphoned off this reporter’s monthly outgoings to have a flutter.

Ten £1 bets, one for each of the 10 Premier League games across the opening weekend, and one five-fold accumulator would amount to more than enough trust placed in AI to predict my winnings.

Using a paid-for AI app, Gemini’s 2.5 Flash, and bearing in mind this was the opening weekend and one that can often be hard to predict, we had a pop.

My prompt? “Give me a bet, to the best of your ability, which is likely to be successful for the xxx vs xxx game on xxx August 2025.”

Friday night fright

That nudge offered me a Liverpool win and over 2.5 goals for the Anfield club’s opening night fixture against Bournemouth. And goals courtesy of Hugo Ekitike, Cody Gakpo. Federico Chiesa and Mohamed Salah – plus a brace from Cherry Antoine Semenyo – secured that bet. Liverpool’s winning goals came in the 88th and 94th minutes – phew.

Strong AI Saturday?

Saturday’s prompts offered a straight up Aston Villa win against Newcastle United. Fail.

Then it suggested over 2.5 goals in both Sunderland’s match against West Ham United, and Brighton’s home tie versus Fulham. Eliezer Mayenda, Daniel Ballard and Wilson Isidor came home to secure that bet at the Stadium of Light, albeit the final goal came in the 92nd minute.

At the Amex, however, Fulham’s 97th minute equaliser was enough only for a 1-1 draw, and a failed bet.

At the Tottenham Hotspur Stadium Gemini reckoned the home team would win in a game that had at least three goals. It may not have predicted the Richarlison worldie, but it was bang on where it mattered.

Read more

Exclusive: Two thirds of fantasy football fans to use AI for FPL

In a smart run of wins, AI also backed a City win against Wolves on Saturday night.

Back to Earth

But Sunday was an abject failure; can AI run out of steam?

At Stamford Bridge both Chelsea and Crystal Palace were backed to score. Fail. At the City Ground, Nottingham Forest and Brentford were backed to score less than three goals between them. Fail. And both Manchester United and Arsenal were backed to score at Old Trafford. Fail.

AI resurrection

But Gemini had a resurrection as Monday Night Football hit our screens for the first time this season. Backing fewer than three goals, Gemini was correct in predicting Leeds United’s 1-0 victory over Everton.

As for the five-fold acca? The offering was as follows: Manchester City win; Aston Villa win; Tottenham Hotspur win; Liverpool win; Chelsea and Crystal Palace both teams to score. Three out of five; close but no cigar.

The verdict

In all, then, my £11 returned £8.72, which included various boosts and early pay-outs offered by the betting site. Make of that what you will, keeping in mind the aforementioned caveats.

For context, the odds offered out by Gemini – 4/7, 5/4, 8/11 (twice), 10/11, 21/20, 9/20, 7/10 (twice), 4/5 – ranged for the singles bets while the accumulator was offered at a healthy 9/1.

It was hardly 100/1 throwaways but those kinds of bets would have been rebellious ones for AI to make given the inserted prompt.

But much of the fun of responsible betting, especially on the opening weekend of the Premier League, is the feeling that you yourself have been the genius – not AI.

Take one William Hill punter this weekend, for example, who flung a fiver on a 10-fold accumulator across the Premier League, LaLiga and Ligue 1 and took home over £18,000.

Certainly one to revisit later in the season.

Read more

Exclusive: Premier League data partner Genius expands portfolio

Similarly tagged content:

Sections

Categories

People & Organisations