The Opta supercomputer has kept us company throughout the 2023-24 Premier League season. Now, we compare it’s pre-season projections with the end results.
“It’s over. It’s done.” – Frodo to Sam after destroying the One Ring in Lord of the Rings. Us to you after the conclusion of the 2023-24 Premier League season.
Sunday’s 10-game bonanza brought the curtain down on the campaign, settling everything that had gone to the wire. The final relegation spot was mathematically confirmed; the make-up of the European places was decided (well, sort of…); Crystal Palace finished above Brighton; some other stuff happened, too.
The 2023-24 campaign was momentous in so many ways. From setting a new record for goals scored over a season, to Manchester City becoming the first club in English football history to win the men’s top flight four years in a row; it also brought the end of Jürgen Klopp’s reign at Liverpool and saw Aston Villa secure their return to Europe’s top table for the first time since 1982-83 by qualifying for the Champions League.
The Opta supercomputer has been with us every step of the way, right from just before the opening weekend of the campaign when it chucked out its season predictions.
It’s been churning away, crunching those numbers as it estimated the probability of each match outcome (win, draw or loss) by using betting market odds and the Opta Power Rankings. The model considered the strength all teams by using these match outcome probabilities and simulated the remaining (or all) fixtures 10,000 times.
It was then able to create a Premier League table that showed in what percentage of the latest 10,000 simulations each team finished in each place.
Every week we revisited the supercomputer for match predictions and broader projections for the whole season, all the way up to the build-up of the final day.
Throughout 2023-24, the Opta supercomputer has entered mainstream discourse, with its projections increasingly referred to by pundits and online commentators alike. That’s right, people, we’re part of the zeitgeist.
Of course, it hasn’t been up everybody’s street.
“If the supercomputer changes its mind when a team wins then it’s not very ‘super’, is it?”
Opta supercomputer sceptic, April 2024
Yes, this is one we’ve seen a lot over the course of 2023-24.
And honestly, scepticism’s fine, but whether it’s rooted in light-hearted jest or sneering cynicism, we have to remember that the Opta supercomputer’s predictions work like any other projection model: feed it new information (i.e. a new round of Premier League results) and it’ll spit out new results that are almost certainly different to what came before.
Pointing out these changes isn’t a ‘gotcha’ moment, it’s simply highlighting the model works as intended.
Nevertheless, as with any projections – made by supercomputer, human or octopus (if you know, you know…) – it’s always interesting to revisit them with the clarity of finality.
Just how accurate were the supercomputer’s pre-first-ball-kicked predictions from way back in August? Has it brought shame to the rest of its little robot family and been banished into exile by Opta bosses? Or has it got everything spot-on and essentially rendered the future of the entire sport pointless due to its prescience robbing football of unpredictability?
Well, thankfully for everyone, it’s somewhere in the middle.
Put it this way, the supercomputer has certainly justified its existence and secured a contract extension with Opta through 2024-25, but it won’t be getting hired by Skynet anytime soon.
So, let’s start with the correct predictions.
The title race, then… Back in August, the supercomputer reckoned Manchester City, Arsenal and Liverpool would finish first, second and third respectively.
On Sunday 19 May, Manchester City, Arsenal and Liverpool finished first, second and third respectively.
Yes, yes, we know, even you predicted Man City would be champions, and picking them is like backing peak Usain Bolt at the Olympics.
But in the 10,000 pre-season predictions, they came out on top in 90.2%. Heading into the final day of the campaign, the supercomputer gave Pep Guardiola’s men an 84.3% probability of winning the title, thus highlighting a well-placed confidence that was still incredibly high even 37 weeks into the campaign. As good as Arsenal were, the big favourites were City and they ultimately delivered as the supercomputer expected.
Either way, the top trio go down as three correct projections for the supercomputer.
A bit further down in the August predictions, the supercomputer had Chelsea in sixth. Now, at the time this one even surprised us a little; sure, their 2022-23 campaign was rocky, but after hiring Mauricio Pochettino and spending a lot of money on some big talents in the transfer market, it was easy to see why many expected Chelsea to be higher and challenging for the top four. After all, that’s where they belong really.
But the supercomputer said no. It looked beyond the latest batch of flashy new signings and their 12-year contracts, and after whirring and screeching like an old dial-up modem (kids, you had to be there), it settled on sixth. Bingo. And we’ve still a few more to go.
Fulham and Wolves enjoyed relative mid-table mediocrity (not derogatory). We say “enjoyed” because for the former it meant they avoided the dreaded Second Season Syndrome following a very respectable 10th after promotion in 2022-23, while the latter managed to keep relegation flirtation to a minimum despite many dooming them to the drop in pre-season when Julen Lopetegui left on the eve of 2023-24 starting.
The supercomputer fancied the chances of both to be safe, however. Fulham finished 13th and Wolves one place lower, right where they appeared in the pre-season predictions.
The seventh correct prediction the supercomputer made impacted the relegation battle. Admittedly, it didn’t foresee Nottingham Forest’s deduction of four points for breaches of the Profit and Sustainability Rules (PSR), but considering they’d have still finished 17th even without that punishment, it doesn’t really matter. The supercomputer expected them to struggle, and they did, with their predicted finish of 17th coming to fruition.
So, what about the rest?
The supercomputer’s pre-season predictions may not have got the others bang-on, but there were another six instances of teams finishing two or fewer places away from where the model projected on 10 August 2023.
Among them were two of the three relegated teams. The predicted positions may have been slightly wrong, but the expectation of Sheffield United and Luton Town going down proved accurate. The Blades ended up entering the conversation of the worst teams in Premier League history, while both they and the Hatters conceded goals at remarkable rates, with Chris Wilder’s men breaking records.
Again, you didn’t have to be prescient to see those two going straight back to the Championship, but on 10 August, Sheffield United were given a 49.1% probability of being relegated and Luton a 61.5% likelihood. While Luton’s fate was only officially sealed on Sunday, they’d looked doomed for a few weeks; Sheffield United had for months.
The biggest overachievers in relation to the pre-season projections were Bournemouth; backed to go down with the Blades and Luton, the Cherries went on to finish 12th after a shaky start under new coach Andoni Iraola. That was six places higher than predicted and underlines what an impressive job their Basque boss did in 2023-24, especially after failing to win any of their first nine league games.
Just behind them in the overachievement stakes were Aston Villa. Put down to finish just inside the top half in ninth place, Unai Emery brilliantly led the club to fourth – their highest final position since also ending 1995-96 fourth – and into the Champions League. That means they’ll be dining at Europe’s top table next term for the first time since playing in the 1982-83 European Cup.
Brentford were the greatest underachievers, ending 2023-24 16th and six places lower than predicted, while Manchester United (eighth), Brighton (11th) and Burnley (19th) all fell short of their projected final positions by four places.
With the Clarets, the supercomputer clearly took into consideration their form during an impressive Championship promotion campaign that saw them accumulate 101 points; only four teams in the history of the second tier have ever managed more, and they finished 21 ahead of third place.
They weren’t able to translate that to the Premier League as well as the model suspected, though the supercomputer won’t have known Burnley had the second-youngest average starting XI age (24 years, 248 days) in the Premier League, and many would link that lack of experience to their struggles.
So, no, the Opta supercomputer hasn’t rid the world of sporting unpredictability (yet), which is probably a good thing for all of us. Then again, even the supercomputer doesn’t want the responsibility of being all-knowing and all-seeing – or at least that’s what it told us at the Christmas party.
Nevertheless, with well over half the division correctly placed to within two positions, the top three accurately reflecting pre-season picks and two of the bottom three correct, we think it’s fair to say the supercomputer’s given a good account of itself.
After a busy season, the machine – housed in a top-secret location and guarded 24/7 by football stats nerds rabid rottweilers – will now go into energy-saving mode to conserve bandwidth before doing it all over again for Euro 2024.
Premier League 2023-24 might be over, but the Opta supercomputer’s job is never done.
Enjoy this? Subscribe to our football newsletter to receive exclusive weekly content. You should also follow our social accounts over on X, Instagram, TikTok and Facebook.