Supercomputer predicts 2023/4 league table, and gets it wrong!



By Tony Attwood

As we have been noting for some time, the media was utterly and totally hopeless when it came to predicting the 2022/23 season, and doubly so where the media in question claimed that a supercomputer was involved.

And again as we have pointed out, this use of a supercomputer for football prediction is incredibly unlikley.  As Factmr website says supercomputers have a “High cost of ownership and enormous power consumption” with the “cost of designing, building, and deploying a supercomputer” ranging from US$ 100 million to US$ 300 million.”

So two things.  One: is it likely that $200m machine with enormous running costs would be used to predict the outcome of the Premier League?  Two, is it likely that if such a machine (which is normally used for such minor matters as the defence of the realm), were to be given a bit of time off, it would be used to predict football results?  (Can you imagine it – Russia has just invaded but we missed it because the Sun was using the machine to predict football results…) 

Probably not, but the Sun thinks so and indeed despite its failure last season to get the forecast anything like right, the publishers think we are so utterly, pathetically stupid that we will forget last season and believe it this time.  (Or indeed believe that a supercomputer was involved).

This is what it predicted a year ago (with the actual result in the following column just so you can see how accurate (or perhaps otherwise) they were.


Predicted Position Actual position Club Predicted GD Actual GD Predicted points Actual points
1 1 Manchester City +55 +61 85 89
2 5 Liverpool +46 +28 80 67
3 12 Chelsea +30 -9 72 44
4 8 Tottenham Hots +21 +7 66 60
5 2 Arsenal +12 +45 61 84
6 3 Manchester United +6 +15 57 75


So how did they do?  They got one of the top four right – which is pretty pathetic – for a supercomputer. 

On goal difference they were 39 out with Chelsea.  With Arsenal they underestimated the number of points by a whopping 23 and the GD by 33.  With Chelsea the points were 28 wrong and the GD 53 out.    And this is allegedly a machine that is working on the nuclear defence of the realm!  On this basis, I wouldn’t be surprised if it had declared war on Greenland by the time you read this.

So, given that this is a load of mindless gibberish and utter tripe, what does the Sun have in line for us this time?

Well, maybe the first thing to notice is how seriously they take themselves despite the catastrophic set of errors in last year’s prediction (which curiously they don’t mention).  And they have the nerve to say that the “news” (so it is news now, not a wild and whacky prediction by a non-existent machine)  “will be bitterly disappointing for Arsenal fans who are hoping Mikel Arteta can go one better and win the league after a second-place finish last year.”

And this tells you what the Sun thinks of Arsenal fans.   A newspaper that invented a supercomputer, which then got its predictions utterly wrong last year (something that would be worrying in itself because these are the machines that tell our military when we are under nuclear attack) is now telling us that we are going to be “bitterly disappointed” because they have made the prediction that Arsenal “are in fact not set to even finish second.”  My goodness they have an inflated vision of their importance.

Now maybe because of the utter disaster of last season’s predictions, the Sun has chickened out of giving us anything other than the positions for the coming season.  Instead it prints the badges of each club, just in case we have forgotten, and because they think we won’t notice the lack of any other detail.  Ahhhh….   But we do get the order of the teams so here we go…

  1. Manchester City (last season 1st)
  2. Liverpool (last season 5th)
  3. Arsenal (last season 2nd)
  4. Manchester United (last season 3rd)
  5. Newcastle United (last season 4th)
  6. Chelsea (last season 12th)
  7. Tottenham Hotspur (last season 8th)

Now if you recall from our last post we showed that totals really jump around a bit from one season to the next.  They have got the same top five this coming season, as in the season just finished, but in a slightly different order.   Yes it can happen, but it is very unusual – and oh that was the mistake they made with their supercomputer last year.  Funny that.

What the people who write this stuff haven’t understood (presumably because they don’t bother to look at the maths, and of course aren’t using a real supercomputer because those machines are rather busy trying to ensure that no country launches a surprise nuclear attack on the kingdom) is that one season being like the previous season in terms of where clubs end up in the top half dozen or so places is very unusual.

But year after year they keep on predicting it will happen.

As John W Henry once tweeted,: “What do you think they’re smoking over there at Emirates?”  That comment was a bluff as he later admitted as Arsene Wenger had his facts absolutely right.  But we might be slightly more polite than Mr Henry and not allege any illegal activity but simply ask, “How much are they drinking at London Bridge Place?”



2 Replies to “Supercomputer predicts 2023/4 league table, and gets it wrong!”

  1. The thing is, even if it is a super computer, and that is a massive if, as we all know even a computer is only as good as the information, or data, you put in.

    The only ‘factual’ data you can put in is regarding stats such as how many points they achieved and goals for and against. That kind of thing. And that’s it. So if you put all those ‘facts’ in and nothing else, all a computer can do is predict the same again, because as far as the computer is concerned nothing has changed.

    But of course things have changed, in some cases lots of things, but what effect those changes will have on a teams performance are entirely subjective, and subjectivity is something a computer doesn’t like. A computer deals in facts, or as mentioned early, data.

    So if we don’t want the computer to just say, well based on that statistical data and nothing else, I predict exactly the same thing will happen as happened last season, which is more or less what the computer does, we need to input some changes, and therein lies the problem, because pretty much every single change and it’s effect on a team is ‘subjective’.

    Things such as:

    How have current players improved or regressed?

    How good are any new players?

    Will players that have left be missed?

    New manager?

    New coach?

    In reality, 10’s if not scores of changes, and the effect of every single one is subjective, and as I said computers hate subjectivity. They deal in facts.

    So how do we turn subjectivity in to a fact? Well, its actually WHO turns subjectivity in to a fact? Simple. The programmer.

    And this is how:

    On stats alone there is no way a computer could predict such a massive improvement for Liverpool? On statistics alone a computer would, could, never do that. It’s impossible. As I said earlier, all a computer can predict on statistics alone is the same again, because that’s how they work.

    So that improvement must be based on something other than raw data. So what is it?

    It must be based on all that ‘subjective’ information being inputted by the programmer as fact.

    So how does that work?

    Lets say Liverpool have made 3 signings. The computer doesn’t know that. Somebody has to input those 3 signings. So you put 3 names in. But what good is that to a computer? Just names.

    How good are they?

    Are they what Liverpool needed?

    Will they improve Liverpool?

    All subjective opinions, none of which a computer, even a super computer, could even begin to analyse. What we need is for the programmer to turn all that subjectivity in to facts.

    For example if he inputs that all 3 of Liverpool’s buys are fantastic 10 out of 10 players. That all 3 will settle in immediately. That Liverpool will improve considerably. They wont miss any of the players they have lost. Last year was a freak. But even doing that is difficult.

    But assuming he manages to programme in all that positivity of course the computer will make a positive analysis. But in reality it’s an analysis based almost entirely on the programmers opinion.

    Similarly if he programs in that Arsenals buys are 7 out of 10.They will miss the players that have left. The youngsters will be ‘traumatised’ by last seasons ‘collapse’. Last year was a freak. Everyone will now of ‘sussed’ Arsenal out. Of course, on the basis of inputs such as that it will not be a good prediction. But again it will be made, yes by a computer (Ah Hem), but again based almost entirely on the programmers opinion.

    To even suggest a computer is capable of accurately predicting something that is so subjective is mad. In other words as Tony says, it’s complete and utter bollocks.

Leave a Reply

Your email address will not be published. Required fields are marked *