Welcome to The Athletic’s Premier League predictions challenge, which I’m horrified to have to say has been won by the subscribers.
Yes, Arsenal fan Neel from New Delhi held his nerve on the final day to shake off the persistent and heroic challenge of six-year-old Wilfred — and the meandering, distinctly unheroic challenge of a 51-year-old Athletic writer.
And the final result — with the data-based algorithm ending up bottom of the table — gives us fairly irrefutable evidence that The Athletic subscribers are the smartest around. Too smart for this particular writer, that’s for sure.
Each week since the season began in August, I have been joined by a guest subscriber, six-year-old Wilfred and an algorithm in predicting the Premier League results with varying degrees of success.
We have awarded three points for a correct scoreline and one point for a correct result. We have also awarded a bonus point for any “unique” correct prediction, so for example Neel picked up four points for the subscribers as the only player to back Tottenham Hotspur to beat Everton 1-0 on Sunday.
Big for Spurs, big for subscribers (Richard Heathcote/Getty Images)
The subscribers started brilliantly with Manchester United supporter Vangeeh from India in week one. There were a few ups and downs along the way — for all of us — but there were some stellar performances from our subscribers all over the world, not least Sunderland fan Derek from Rhode Island, West Ham fan Steve from London, Bournemouth fan Chris from Kentucky, Manchester City fan Weston from New Jersey and Arsenal fan Jack from London and Newcastle United fan Adam from the Republic of Ireland.
Newcastle fans played a big part. In addition to Adam’s contribution, there were spectacular scores from Hannah from Bristol in gameweek 22 (14 points) and Elizabeth from San Francisco in gameweek 31 (13 points). Some of the readers in the comments section took issue with Elizabeth’s predictions in particular (a Newcastle fan predicting Sunderland to win the Tyne-Wear derby?!) but she and Hannah both played a blinder in weeks when the rest of us fell short.
Wilfred has been amazing, recovering from a tough start to force his way back into contention and to lead the table for much of the second half of the season.
“I feel proud of myself for coming second, but I would have liked to come first,” he says. “Congratulations to the subscribers. Next time, I’m going to beat you.”
Do I feel embarrassed at finishing behind a six-year-old? Yes, I do. Is my embarrassment tempered by the fact that this is a six-year-old so smart that he even backed his own team, Crystal Palace, to lose 3-0 at Manchester City the other week? Yes, I do. Do I wish we had chosen almost any other six-year-old on the planet to go up against me in this game? Yes, I very much do.
Going into the final day, Wilfred and the subscribers were locked on 253 points, with Wilfred in first place on the basis of having predicted more correct scorelines. I was a little further back on 246, with the algorithm bringing up the rear on 237.
Under the circumstances, you might have expected Wilfred and Neel to play it safe with lots of similar predictions.
But they went in different directions. They made two identical predictions — a 1-1 draw between Burnley and Wolves (spot-on) and a 2-0 win for Arsenal at Crystal Palace (one point apiece for a correct result). Both also went for victories for Liverpool, Manchester City and Bournemouth (all wrong) but predicted different outcomes for Brighton vs Manchester United, Fulham vs Newcastle United, Sunderland vs Chelsea, Tottenham Hotspur vs Everton and West Ham United vs Leeds United.
In the end, it was those games that made the difference. Wilfred got a point for West Ham’s victory over Leeds, but Neel got two points (including a bonus point) for Manchester United’s win, one point for Sunderland’s win and — the big one — four points for being the only one to predict that Tottenham would beat Everton 1-0. That made 11 points to Wilfred’s five.
I also picked up 11 points — including spot-on predictions for Nottingham Forest 1-1 Bournemouth and Sunderland 2-1 Chelsea — and I did briefly find myself wondering whether a late twist might suddenly swing things in my direction. All it would have taken was a Burnley winner against Wolves, allowing me to claim four precious points (including a bonus point) and deny the subscribers three.
Alas not. I ended up third, seven points adrift of the subscribers and one behind Wilfred.
And if you honestly think I’m petty and small-minded enough to point out that overall I made far more correct predictions (172) than everyone else — Wilfred (160), subscribers (158) and the algorithm (145) — then you’re quite right.
That, as David Brent said after watching Finchy throw a shoe over a building in episode three of The Office (the original and superior UK version), is the real quiz.
But seriously, my lack of correct scorelines was a bit of a head scratcher. It’s not as if, with a few exceptions, I was going for wild scores; I thought I was playing it safe in a low-scoring, predicting a lot of 1-1s and 2-1s, but I kept falling short. That is one of various trends I’ll investigate in an end-of-season debrief, which will be coming soon.
At least all three human players — even this one — managed to outperform the algorithm. There were times when its obsession with 1-1 and 2-1 scorelines bordered on the disturbing. On the final weekend of the season, hilariously, it backed nine of the fixtures to finish 1-1.
But three of those 1-1 predictions ended up correct, meaning the algorithm finished off with a very solid 10 points.
The algorithm was something of an enigma: started off strongly, losing its way in comical fashion and finished strongly (40 points in the final three rounds and ended up only 17 points off the top, having been 37 points adrift with just three weeks to play.
A source close to the algorithm, speaking on the condition of anonymity in order to protect relationships, told The Athletic: “The algorithm can be satisfied with its first season operating at this level. Its commitment to analysing games in the same precise, considered fashion for 38 weeks is admirable in a sport that is often led by narrative and hunches. The machine will now enjoy some valuable ctrl-alt-delete time and a leisurely boot.”
Whatever, Algo. I’ll try to get some proper answers for that end-of-season debrief, which will demonstrate just how wildly I overestimated some teams and underestimated others.
Anyway, that is to follow. For now, I just want to thank three of my editors, Alex Brodie, Duncan Alexander and Andrew Fifield, for their help in making this feature a success and also thank and congratulate Wilfred for being a great sport and utterly brilliant — often far too brilliant for my liking.
And finally, I want to thank all the 38 subscribers who have played this year (and the thousands of others who applied and who have read and engaged with the predictions and articles on a weekly basis) and to congratulate you on making me look stupid. I’m already plotting my revenge.
