Can we really believe Opta predictions any more?



By Tony Attwood

Opta is currently sending out email news items relating to its supercomputer’s predictions for the forthcoming season.   As ever, I have written back asking them the simple question, “which supercomputer they are using?”

And I ask this because last summer of the 500 or so supercomputers that there were in the world, only 11 were in the UK, so it would be good for us to do an interview with the owners of the supercomputer that Opta used and ask them why they allowed a machine like this to be used on football.

Indeed as Britannica says the term supercomputer, has “been used primarily for scientific and engineering work requiring exceedingly high-speed computations. Common applications for supercomputers include testing mathematical models for complex physical phenomena or designs, such as climate and weather, evolution of the cosmos, nuclear weapons and reactors, new chemical compounds  and cryptology.”

Now I think this is important, because if, as I suspect there is no such thing as the Opta supercomputer, but instead it is just a regular computer of the ordinary type, then we have a case of the leading supplier of statistics on football in England making up a story.

Even so, we thought it would be fun to see how well their alleged “supercomputer” did last year in its actual predictions compared with its results at the end of the season.  If poorly, then it might be that they were actually not using a supercomputer at all.  

Opta Analyst published an article in August 2022 which predicted the league positions of the 20 clubs by the end of the season

Predicted Premier League Positions in 2022-23 compared with reality

  1. Liverpool actually finished 5th, 22 points behind the final winners
  2. Manchester City actually finished 1st, 5 points ahead of the second-placed club
  3. Tottenham Hotspur actually finished 8th, 15 points behind the third-placed club
  4. Chelsea actually finished 12th, 27 points behind the fourth-placed club
  5. Manchester United actually finished 3rd, 8 points above the fifth-placed club
  6. Arsenal actually finished 2nd, 22 points above the sixth placed club
  7. West Ham United actually finished 14th, 21 points behind the seventh placed club
  8. Newcastle United actually finished 4th, 11 points above the eighth placed club
  9. Leicester City actually finished 18th, 25 points behind the ninth placed club
  10. Aston Villa actually finished 7th, 9 points above the 10th placed club.

Now these figures are shocking in several ways as we have oft noted before (although not with quite so much detail of the errors).  Shocking because

  • a) None of the ten predictions was right
  • b) Only one of the predictions was within one place of being right
  • c) The average error was almost five places
  • d) Not only was there never an apology for the gross level of errors achieved, the same people are undertaking predictions again this season without even acknowledging that they got every placement wrong last season, with some being eight or nine places out.

What is really needed at this time is not only a fulsome apology for such a gigantic cock up but also an explanation of how a supercomputer, the most powerful computer in the world could get it so wrong.

As we have noted before supercomputers are in charge of our defence systems, which of course rely on predictions.  So on the basis of this range of errors the UK’s weaponry could quite happily open fire on France, Germany, Sweden or indeed Lichtenstein.   This is indeed a worrying situation, and one that needs to be resolved quickly.  Unless of course this work wasn’t actually done by a supercomputer at all.

We should also ask, “Given the gross errors that arose last season what reprogramming has gone on, and how many of last season’s programmers have been sacked?”

And most of all we might ask, “Where is the apology?”

Anyway, the Mirror picks up the story and tells us that a “supercomputer prediction from King Casino Bonus has attempted to project this season’s table, based on betting markets and how those predict how the next 12 months of action will unfold – with no shortage of interesting results.”

In their fantasy league Liverpool are due to rise to second, and Arsenal drop to third, with Manchester United fourth.  

Opta-Simulated Premier League Table

Meanwhile Opta say that they have averaged the total number of points of each PL club across to give us this season’s final outcome.   Here’s the top ten; sadly Tottenham, with their urgent need for money to pay for the stadium, will not make Europe again, unless the places are extended.

1st: Manchester City – 88.81 average points
2nd: Arsenal – 72.23 average points
3rd: Liverpool – 71.47 average points
4th: Manchester United – 68.49 average points

5th: Newcastle United – 61.23 average points
6th: Chelsea – 58.90 average points
7th: Brighton & Hove Albion – 57.51 average points

8th: Tottenham Hotspur – 56.20 average points
9th: Aston Villa – 55.38 average points
10th: Brentford – 53.20 average points

And the relegated clubs

18th: Bournemouth – 36.52
19th: Sheffield United – 36.26
20th: Luton Town – 34.19

I suppose, since we now know how the league will finish, there’s not much point watching anymore.


7 Replies to “Can we really believe Opta predictions any more?”

  1. I’m not so sure the predictions used a computer of any description. If my suspicions and observations are right I think most of the calculations were determined and written on a beer mat in the Toppled Bollard perchance ?

  2. I did a piece on this a few months ago, and the basis of my argument was, a computer is only as good as the data that is imputed, and as most of the data that is imputed regarding football is ‘subjective’, the results aren’t going to be very good. Reason being, it is impossible for a computer to do calculations based on subjectivity with any accuracy because they basically work on mathematical certainties, more or less anyway, and none of these inputs are mathematical certainties.

    How Arsenal will do compared to last season, and in relation to every other team, depends almost entirely on subjective factors, such as:

    -How players have improved – subjective

    -How we will miss players sold – subjective

    -How players bought will play – subjective

    -How many players will come through the academy- subjective (or a guess at best)

    -How players bought will suit the team – subjective

    -Will other teams ‘sus’ our style of play – subjective

    -How many injuries will we get – subjective

    -How will be refereed – subjective

    A computer, super or otherwise, is simply not capable of dealing with this type of data. Or rather it is, but it’s conclusions are entirely dependent on whether, for example, the programmer inputs that player ‘A’ sold, will not be missed, while player ‘B’ bought will make a huge difference.

    If, for example the programmer inputs that:

    We have sold Xhaka, who he rates as a 7/10 player, but bought Declan Rice, who he rates as a 9/10 player, then the computer will analyse this as improvement because it’s a mathematical certainty. If he inputs the opposite ratings then we will be worse off. Another mathematical certainty.

    The validity of this particular example is irrelevant, it’s the demonstration that the prediction of the computer is entirely dependent on 100’s of, if not 1000’s of these entirely subjective inputs from the programmer.

    In other words it’s a load of bollocks.

  3. The media are hoping that they will be able to use the aforementioned computers as scapegoats when their predictions invariably turn out to be wrong. Unfortunately, they forgot about their own “get out of jail free” card.

  4. @ Nitram

    I agree with your fundamental hypothesis but to suggest the concept that ‘how we will be refereed’ is subjective is highly questionable. That’s one of the most predictable factors in any season 🙂

Leave a Reply

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