Welcome to ALEX - the ALgorithmic EXpert.
ALEX applies sophisticated AI tools to a database of over 250,000 football match results to identify recurring underlying patterns of football 'behaviour'. She then looks at future matches, identifying the historic behaviour patterns that apply to these future matches. This gives her a unique insight into the likely outcomes of these future matches, allowing the identification of likely VALUE bets.
Betting exclusively on VALUE bets is the scientific way to make a profit. Only outrageously unlucky people (the sort of people you wouldn't want to get on a plane with) will lose money if they bet only on VALUE bets.
ALEX identifies these bets, and she makes a recommended stake size, with the objective of maximising profitability.
New features are constantly under development, but forecasts currently available are...
Home Win Recommendations, including suggested stakes, for forthcoming fixtures.
End-season winners and losers - who's going up, and who's going down?
Special bets on match outcome and both teams to score.
Don't bet unless - advice on the minimum you should accept for a given home win.
Weekend Recommendations
No recommendations to show during the close season. Instead there's a review of 2024-25' performance.
ALEX has had a hugely successful 2024-25 season. Over 25 weeks of betting, we saw 20 highly priced Home Wins delivering a return of £896 from stakes of £407 – a super profit of £489, just in excess of 120%. As ever, particular thanks are due to a few underdogs, and this season I am especially grateful to St Johnstone (v Celtic) and Heerenveen (v PSV Eindhoven) for rising superbly to the challenge and winning at 16/1 and 14/1 against.
Could it have been just luck? That’s always a very good question so I did a simulation of 1 million gamblers betting at random on the season’s 8,000 or so in-scope matches, with each allowed the same £407 total stake. The distribution of the profits that they achieved is shown below. The 1 million random gamblers’ results are all in the red peak to the left while ALEX’ performance is marked by the black line way to the right, almost off the scale.
Most of these gamblers lost money. 20% did make a profit, the highest (out of 1 million attempts) being £115 – that’s pretty good but it’s nowhere near the £489 made by ALEX.
Even if you have more than 1 million gamblers – and let’s say you have 1 gambler for every atom in the observable universe – then I’ve done the calculation and it’s still highly, highly unlikely that any of these atomic gamblers would do better than ALEX. So her performance is probably a bit more than just luck.
This is the 3rd successive year of profitability. In 2022-23 the profit was 80% (£748 return on stakes of £415) and in 2023-24 it was 52% (£708 return on stakes of £465). What does this mean for next season? Is success guaranteed in 2025-26? Sadly not, but it does mean that a poor season would be very much against the run of play.
And as for ALEX. You may think she is looking forward to a well-deserved rest during the close season. Well, that’s not going to happen. Like all good AIs she never complains about being worked hard so she will be spending the summer digging through the data looking for ever more profitable form patterns that can be exploited in future seasons.
Looking forward to next season with first forecasts sometime in September.
End-season Winners and Losers
update dated 4th April, 2025
With only 8 games to go for most premiership teams, it looks like the top 4 (per my ups and downs models) are going to be Liverpool, Arsenal, Forest and probably Man City 4th, though there is some support for Chelsea to be 4th, but none at all for Newcastle.
The relegation teams are currently likely to be Southampton, Leicester and Ipswich, but with still an outside chance of Wolves deposing one of those 3 - less than 5% probability - bookmakers are offering 14/1 so not worth it.
These forecasts are in line with William Hill’s cash out offers on my early season bets. They will give me £19 for my Forest top 4 bet (£29 full payout if I hang on) but only £1 for my Wolves relegation bet (full payout £25).
1st March, 2025
I have also been looking at end season outcomes and can confirm that the chance of Liverpool NOT being champions is pretty slim. To quantify this, I used a look-back model that simply evaluates where teams ended up historically based on their end-February position, and on this basis Liverpool’s winning chances are 66%.
A more sophisticated model that ‘plays’ out all remaining matches, taking current form into account, sees their chances at 96%.
At the other end of the table, most likely to go down are Leicester, Ipswich and Southampton, while Wolves have about a 40% chance of replacing one of these teams in that select group.
And at the other end of the English Pyramid, where I have a particular affection (some might say affliction) for Aldershot, it looks like we are heading for a 17th or 18th place finish, with relegation an outside chance at about 10% likelihood.
But remember, ALL models are wrong, but some are useful.
Match Result and Both Teams to Score
This section gives 3 selections for a Home win and both teams to score, and 3 selections for an Away win and both teams to score.
Note that I have not been able to find sufficient historic price data to test the models by hindcasting.
This approach is at an experimental stage. I have not been able to find sufficient historic price data to test the models by hindcasting.
But I can report that the results from the prior 6 weeks this season, after a very promising start, are currently showing a small loss - but hey, it's early days.
This section is suspended while more historic data is collected to enable full model hindcasting testing.
Don't bet unless...
This (sometimes rather large) table estimates the minimum odds that you should take to get a HOME WIN value bet.
Last updated for week 28 - weekend of Saturday 12th April, 2025
Latest blog posts...
Is there a role for ChatGPT? - posted on 14th March, 2025
I decided to get some help from ChatGPT this week. The numbers I use each week come from various sources, with a fair amount of manual processing so I thought I can get ChatGPT to do this for me, and save myself a little bit of time and a lot of tedium.
So as an easy start (and this one really is easy), I said to ChatGPT - please give me the home, draw and away odds for the weekend fixtures, as given on the William Hill site. Sure enough it gave me back a table of fixtures and numbers. But were they correct? Some of the odds appeared to beunusual. So I looked up one of the fixtures to check. Surprise, surprise - the fixture wasn’t even in the weekend list. When I investigated further I found that none of them was. They looked ok but they weren’t real.
So I fed this back to ChatGPT. Its response was that, for several reasons including web scraping challenges and the requirement to have interactions with the website (the things that make it tedious for me), it was difficult to get the William Hill numbers, so instead it had provided me with “placeholders”.
So I gave it another chance to come clean…
To be fair, ChatGPT does advise users to double check its advice, and it did appreciate my scrutiny
You'll notice that I now interact with ChatGPT as if he/she were a person. Would it pass the famous Turing test? Well, who better to ask than ChatGPT.
I think it shows great self-awareness in recognising and articulating some of its deficiences, and its potential.
What's happened to Home wins? - posted on 27th February, 2025
There was a story on the BBC this week reporting that Home Wins are at a low ebb in the Premiership. This matters for ALEX's recommendations as she only recommends home wins, and that could be a problem if there are less of them.
The BBC found that there has been a smaller proportion of home wins in the Premiership this season than in any prior season excluding the Covid season. The proportion so far is 39%, while typically you would expect to see about 45% - so that's a big reduction. There was a plausible hypothesis that the Covid season proportion was down because the lack of fans meant that referees were less likely to suffer home crowd intimidation, so fewer decisions were 'given' to the home team. That's plainly not the case now, so what else could be going on?
It's always good to check the data for reports like this, so I did just that. Indeed, the proportion in the Premiership is down as reported (to 39% from last season's 46%). But is this a widespread issue? I looked at this season's figures for the more than 20 of the top European divisions in ALEX's scope, and found that in only 2 other divisions has a similar trend been observed - the German Bundesliga 2 is down to 38% (46% last season) while the Greek Super League is at 39% (45% last season). At the other end of the spectrum, the Scottish Championship is at 52% (against a very low 37% last season) and the Scottish Division 2 is 52% (versus 41% last season). If these outliers are ignored the European average is 44%, comfortably near the long-term figure.
So while the BBC presents several pretty interesting theories to explain the 'change', which may indeed be having some impact, maybe it's no more than a statistical wobble. I will take another look at the end of the season and let you know.
Introductory blog - posted on 19th February, 2025
Hi there. I will add posts from time to time, looking at things that are football-interesting, gambling-interesting or anything else that takes my fancy-interesting.
