When I first started monitoring this service three months ago, I was raving about it. After all, what usually happens when I review a service, all the hitherto top results take an about turn and head for losses. After 11 days of AI Football Accas, I was 35 points up with a 50% strike rate which, for accumulators, is ridiculous. I had started to believe – but of course, I was wrong; it couldn’t be sustained. The following 47 bets found just two winners. After three months, the service has just broken even – less the cost of the subscription.

First, a word about AI (artificial intelligence). It’s been the buzzword for a couple of years now, how it’s going to transform the lives of all of us, and I have no doubt that’s true. But it’s not one size fits all, it very much depends on how the writers and programmers develop it for use in any particular service. It’s certainly possible to use it in sports prediction, and the more variables which can be included will make the programme more efficient. I remember in 1996 when Alex Ferguson changed the Manchester United team’s shirts at half time against Southampton because the players complained that they couldn’t see each other properly in grey (maybe an excuse as they were 3-0 down!). They still lost the game, but they won the second half. Whilst most sensible commentators will argue that the shirt colours are irrelevant to results, do we have data to show this? It will be very hard to show one way or the other because there are an enormous number of variables to consider. That’s why we need a computer to crunch the data. And even if you can show by simple stats that Arsenal get better results when they wear red shirts, it doesn’t show that the results are better because of the shirts. AI should be able to tell us the likelihood that the factor of shirt design has any significance in the performance. And further, for the team as a whole, and for individual players. And so it goes on.

I’m rambling. This is because a lot of betting service sellers are using AI to suggest that the service they’re promoting is somehow smarter than a human assessor. But there are many services out there providing data on football, horse racing and other sports which are accessible to us all. We’re not told about the nature of the AI programmes, if the seller has developed them for themselves, if they’re using services which are available to all, or if they’re really using any AI systems.

I think services which claim to use AI for betting suggestions would do well to engage with their subscribers in order that suggestions can be made to improve them. A developer simply can’t think of everything (the obvious problem here is that data for many variables, no doubt including shirt details, just doesn’t exist in one place anywhere. But a lot can be collected and developed over time).

Anyway, to be fair, break even is probably a better result than most of us would have guessed after a three month trial. Here are the numbers:

Total number of bets: 84

Total points staked: 171

Average number of legs per acca: 4-5

Average stake per acca: 2 points

Average odds per acca: 10.25

Average odds per winning acca: 7.23

Wins: 12

Losses: 72

Strike rate: 14.2%

Profit/loss: -0.78 points

ROI: 0%

So a shrug. I would say that there were a few errors on the emails which remained uncorrected, such as 12th January 19:30 – Germany – Bundesliga – Burnley vs Luton – Bet Both teams to score (y) @ 1.21: but it was usually possible to work out what the intended bet was.

Also, I’m not really enamoured about paid services which use their subscriber database to promote other services for affiliate fees, and this is one of those. 

Would I give this service the thumbs up? Well, I won’t be continuing with it myself, but there’s always the thrill factor when an acca comes up, so I’ll rate this as a neutral. What I take from this mostly, though, is the run of 47 bets with just two winners.

You can try AI Football Acca Here