I did this exercise last year, and we’re running it back for 2020-21. Let’s take a look at all 20 of my Premier League predictions for this season, and see if there’s anything we can learn from just how wrong -- and occasionally right -- I was.

1. Manchester City (Prediction: 2)

I just couldn’t help myself. I knew Liverpool weren’t as good as Manchester City. It was there in the numbers. It was there in all the close wins. It was there in the goal differential. It was there when the two teams played after the pandemic-pause and City absolutely annihilated the just-crowned champs. Hell, I even said it when I made my predictions:

I’ll probably be wrong — gun to my head, well, I can’t really imagine being in that kind of situation, but were you to press a revolver against my cranium, this prediction may well be different than it appears here. It’s just that City’s defense really is vulnerable against any team they play. They can pummel teams into oblivion that Liverpool cannot, but they can also lose to teams that Liverpool just don’t seem vulnerable against.

Did I feel pretty good about myself after three games, when City’s defense looked as vulnerable as it ever had under Pep, after a loss against Leicester and a draw against Leeds? And Liverpool steam-rolled Leeds, Chelsea, and Arsenal by a combined 9-4 scoreline? You bet I did -- and LFC were the betting market favorite at that point, too. Then, of course, something something seven goals against Villa, and Pep completely flipped the script on his team. They slowed things down to a crawl and became invulnerable against the kinds of counter-attacking mid-table teams they lost to the year before. After the defeat to Spurs in November and before clinching the title, City only lost to Manchester United, Chelsea, and Leeds United -- and the latter two were fluke results in games City totally dominated.

This was predictable -- and the betting markets did predict it -- but not in the way it happened. City went from a high-wire defense that gave up a tiny number of high-quality chances ... to a defense that gave up a tiny number of terrible chances. They conceded the joint-lowest xG per shot and the fewest shots per game in the league. (All stats in the piece are via FBref.) That’s it. That’s the whole ballgame. And on the other end, after three years of City players totally dominating all of the attacking-stats leaderboards over at FBref, the team’s leading goalscorer was Ilkay Gundogan with 13, and the only other guy in double digits was Raheem Sterling, who became a bench player in the second half of the season. But that’s always a potential outcome when you pick a fantastic squad with an all-time manager: They find a totally new way to be great.

2. Manchester United (Prediction: 5)

Got this one wrong, but I don’t think it was that far off. I was low on United because I didn’t think they could win 14 penalties again, and because I thought they were too reliant on Bruno Fernandes and Paul Pogba in what seemed like it would be a grueling season where everyone was more likely to get hurt. Pogba only started 21 matches, but Bruno seemingly played more than the maximum 3,420 minutes that any single player can accrue in one season. Also, they won ... 11 penalties this season. That’s 12, 14, and 11 in the past three seasons in a statistic that really helps you win -- penalties get converted about 75 percent of the time -- and has also not really shown itself to be predictive or repeatable. But even with the penalties baked in, United finished with just the fourth-best xG differential in the league. And it’s not like they were particularly good at doing the things that come before the shots, either:

Mark @markrstats

#PremierLeague Expected Threat created/allowed Best attack - #Liverpool, best defence - #ManCity Good to see all UCL qualified teams in top-right corner (Tier A) #Brighton, #Arsenal, #Leeds, #AstonVilla and #Leicester can be considered as Tier B teams

May 24th 2021

15 Retweets86 Likes

The main reason United landed in a clear-ish second rather than in a final-day scrum is that they were lights out in front of goal, scoring 70 goals on 60.2 xG. That being said, I picked them fifth, and they were one of the four-best teams in the league, variance be damned.

3. Liverpool (Prediction: 1)

One of the weirdest seasons I can remember: Start off flying, seem like you’re gonna win the league, get absolutely destroyed by a team that just barely avoided relegation, lose your superstar center back, rally to seem like you’re gonna win the league again, lose all of your other center backs, forget how to shoot, plummet down the table to the point where it seems like Europa League might be a stretch, get rescued by an injury-time header from your freakin’ goalkeeper, and somehow ultimately finish third by multiple points. If Alisson didn’t score against West Brom, this team would’ve still finished third. What the hell?

I picked Liverpool to win it all on the basis of this theory: “I trust them to navigate chaos more than anyone else. After all, that’s the whole point of hiring Klopp”. And frankly, they navigated the chaos extremely well! It was just way, way, way more chaos than anyone could have reasonably expected. Still, the xG gap between them and City was pretty big last season; I shouldn’t have ignored that. 

4. Chelsea (Prediction: 3)

One spot off, still top four, I’ll take a W on this one. Congrats to me for picking the only team that spent hundreds of millions of dollars last summer to qualify for the Champions League.

OK, there’s another reason why Liverpool and Chelsea weren’t comfortably top four and United were: these teams couldn’t shoot for shit. Liverpool scored 65 goals on 72.6 xG, and Chelsea were slightly worse for 56 goals on 64 xG. Timo Werner, meanwhile, ends the season as the Premier League’s Underperformance Champion, scoring six goals on 11.9 xG. Get your memes in now, people. For RB Leipzig in the Bundesliga last year, he exceeded his xG by nearly the same amount he underperformed this season: 5.6 goals. Finishing is incredibly volatile -- on both a player and team level -- and relying on an apparent ability to convert chances at a significantly higher rate than everyone else is a reckless way to go about things.

I’m updating the numbers Statsbomb’s James Yorke used here, but despite all the volatility in the table this year and the injuries and the firing of managers, the only thing that really changed at the top was how well certain teams kicked the ball into the net 15 or 20 separate times:

-City’s per-game non-penalty xG differential in 19-20 and 20-21: 1.36 to 1.12
-Liverpool: 0.75 to 0.76
-Chelsea: 0.66 to 0.70
-United: 0.35 to 0.33

This sport: pretty dumb.

5. Leicester (Prediction: 4)

Let’s go back to the same number to explain why I picked the Foxes to finally finish fourth. Last season, their npxG differential per game was a really healthy plus-0.51. Despite their second-half collapse, that’s a significantly better marker than what United produced, and given that much of United’s top performance came after the pandemic pause, “Leicester above United” seemed like a pretty easy data-informed choice. Plus, they also do employ Jamie Vardy, who actually does seem to be one of the few players who’s consistently able to draw penalties.

Well, Vardy led the league in penalties won this season -- but everything else was ... average. They were 10th in both npxG and npxG conceded per 90 minutes, which added up to a differential of plus-0.01 per game. Leicester were basically playing their opponents even from week to week, finishing their chances, and winning a ton of close games. Last year, they were a really good team who couldn’t capitalize on anything over the last 10 weeks of the season. This year, they were an average-ish team that pretty much capitalized on everything until the last five or so weeks of the season, where their lack of quality finally caught up to them.

I could take a W here for being one spot off and, well, also this:

OptaJoe @OptaJoe

242 - Leicester ended more days in the Premier League’s top four than any other side this season and yet failed to finish inside the top four places (242 days). Unlucky.

May 23rd 2021

543 Retweets2,115 Likes

But despite narrowly missing out on the Champions League and winning the FA Cup, this team was significantly worse than I expected them to be. The lack of Wilifried Ndidi and especially Ricardo Pereira, who was probably the best right back in the league after Trent Alexander-Arnold over the previous two seasons, seems like it was just a total killer.

6. West Ham (Prediction: 14)

I wrote my predictions when I thought that I had -- and was told by a medical professional that I probably did have -- COVID. Turns out: I didn’t! But anyway, the second half of the predictions was just a list of teams, in the order I predicted them to finish. No commentary. So, West Ham didn’t even get a write-up, and why they hell would they? Everyone got West Ham wrong:

Omar Chaudhuri @OmarChaudhuri

City are champions, with 3 fewer points than what was expected of them before a ball was kicked. West Ham finish with 22 more points than was originally expected - a hugely impressive overachievement. Aston Villa too, at +18. Easy to forget how unfancied they both were in Sept.

May 23rd 2021

43 Retweets251 Likes

It wasn’t a total fluke either. They finished sixth, with the eighth-best npxG differential in the league. If you told me that Michail Antonio wouldn’t even play 2,000 minutes and that Thomas Soucek would be their joint-leading goal scorer, I would’ve told you they were gonna get relegated. Instead: David Moyes would be my manager of the year.

7. Tottenham (Prediction: 8)
8. Arsenal (Prediction: 7)

Got the spots slightly wrong, but I think I nailed both of these in spirit. 

Despite being betting-market favs to finish fifth, Arsenal were mostly terrible under Mikel Arteta last season. They just didn’t take any shots and were relying on an aging striker to continue finishing his chances at a super-high rate despite a career-long sample that suggested he would not, in fact, continue to do that. Pierre-Emerick Aubameyang both got worse and stopped finishing his chances. 

Tottenham, meanwhile, really just hadn’t been good for a sustained period of time since long before Mauricio Pochettino left the club. They had a nice start to the season, but Jose Mourinho eventually Mourinho’d the hell out of the thing. Harry Kane was the best player in the league, Son Heung-min probably had his best season in a Spurs shirt, Gareth Bale led all players in goals+assists per 90 minutes, and they still really weren’t even close to finishing top four.

The lesson for these two predictions: trust the numbers, not the names on the jerseys.

9. Leeds United (Prediction: 12)

The “If You Want to Know Which Promoted Team Is Going to Be Good, Just Look at Whichever One Conceded the Fewest Goals” Theory continues to be undefeated. Still, I didn’t think Leeds would be this good, in this fashion:

-Fourth in npxG per game
-Fourth in shots per game
-Sixth in passes into the penalty area
-First in our hearts
-And first in anything related to RUNNING YOUR FACE OFF:

Two points back of Arsenal in their first season back up in the Premier League. We must protect Marcelo Bielsa at all costs.

10. Everton (Prediction: 9)

Last year, I learned to not put too much stock into offseason additions when making preseason predictions: a single player can only have so much of an effect in a game that involves 22 guys, plus sometimes it takes a lot of time for players to settle in and truly have a positive impact on a new team’s performance. Everton signed James Rodriguez and Allan -- two good-to-great veterans -- and I avoided the temptation of bumping them up because of it.

One spot off, but I’m gonna give myself this one, too. Everton paid a lot of money to hire Carlo Ancelotti -- and he just hasn’t had an effect on the team’s performance in any way. If anything, the team is more talented than before he arrived but actually performing at a worse level. Their per-game npxG differential was slightly positive last year; this season it dropped down to minus-0.17. They went big on some win-now moves, and now they’re stuck in no man’s land once again.

11. Aston Villa (Prediction: 17)

If you stare into the sun for five straight hours and then come back to look at these predictions, you might think that I got this one right. Villa were a relegation team, according to the betting markets, and after all: I did not predict them to get relegated! But this team finished with a positive xG differential and a plus-9 goal differential a year after sort of magically and improbably avoiding the bottom three right at the end of last season. Jack Grealish was the best player in the league for the first half of the season, but barely played in the last third of the year. I just didn’t see Grealish making that kind of leap, and I didn’t see Villa, who conceded the most xG in the league last season, not conceding a shit ton of goals.

12. Newcastle (Prediction: 18)

This is not the 12th-best team in the league, but this was also not quite a bottom-three-quality team, either. Callum Wilson had a really nice season -- averaged more than a goal or an assist every other game -- and Joe Willock briefly turned into Messi, scoring eight goals on 3.6 xG in an amazing late-season flurry that will almost definitely never happen again! They’re the only team in the league that didn’t average at least 17 touches in the penalty area per match, and only West Brom conceded more shots per game. Still seems like this is gonna fall apart at some point, but I was wrong in thinking it would be this season.

13. Wolves (Prediction: 6)

Probably my worst prediction of the season! To be fair to myself -- and shouldn’t we all strive to do that? -- I made the prediction before Diogo Jota was sold to Liverpool. Also, then Raul Jimenez fractured his skull and missed most of the season. Wolves’s entire structure was built around their ability to defend with numbers and then create enough attack through one- or two- or three-man counters with Jota, Jimenez, and Adama Traore. Without two of those three, the attack declined, and the defense also took a hit, probably in part because Joao Moutinho has been playing professional soccer since the first term of George W Bush’s presidency and because the attack just wasn’t able to give the defense any rest. Now, their manager, Nuno Espirito Santo, is gone, too. Before last season, Wolves seemed like they’d become maybe the most consistent and predictable team in the league. Not anymore, though.

14. Crystal Palace (Prediction: 16)

Sure? I’ll take it? I guess? Why not? This team is really bad. Only Sheffield United and West Brom had worse npxG differentials, and only all three relegated teams had worse goal differentials. On top of that, when weighted by playing time, they were the oldest team in the Premier League this year, with an average age of 29.1. Also, their one consistent young player, Eberechi Eze, suffered a serious Achilles injury just a week ago. With Roy Hodgson leaving the club, I’m having a hard time seeing why anyone would take this manager job -- beyond the pay. Unless something drastic changes, this team is going to be fighting relegation next season.

15. Southampton (Prediction: 10)

Here’s what I said about them before the season:

Great coach, great goalscorer, great pressing system, and a not-great rest of the roster. Right on the edge of the top-half sounds about right.

Danny Ings both got worse and got injured, and this season was not kind to any pressing team not coached by Marcelo Bielsa. But I think there’s another lesson in here that connects with Wolves: don’t be too confident in mid-table-ish teams who are overly reliant on one or two top attackers to keep them above water. 

16. Brighton (Prediction: 15)

On the one hand: I was only one spot off.

On the other:

I’m not gonna ... actually, you know what? Screw it. Brighton basically had the 2015-16 Leicester season, just without finishing and against much better competition. I wouldn’t be able to sleep at night if I gave myself credit for accurately predicting that.

17. Burnley (Prediction: 13)

Last year, Burnley finished with a minus-7 goal differential. This year, they dropped down to minus-22. I think 13th was a bullish take for Burnley, but they ended last season with better xG numbers than both Arsenal and Spurs. What happened this year? Nothing, and that’s the problem. They squeezed some production out of Ashley Barnes and Jay Rodriguez last season, and neither one played much this year ... because they’re both journeymen now in their 30s. On top of that, Dwight McNeil didn’t improve, which is normal since he’s still only 20. And the rest of the team just got old. They’re the second-oldest squad other than Palace, and they’re pretty simply just worse at creating shots and preventing them than they used to be. That took them from “comfortably midtable” to “fourth-worst team in the league”. This is their new baseline; let’s see if I remember that when I do the predictions again in August.

18. Fulham (Prediction: 20)

They probably did play well enough to stay up. This was a much better go ‘round than their last promotion from the Championship, but ultimately Fulham were doomed by one thing: they didn't have anyone who could score! Their leading scorer was Bobby Reid, who netted just five times. Scott Parker opted to move away from their targetman Aleksandar Mitrovic, who scored 26 in the Championship last season and 11 in the Premier League the year before that. Perhaps Parker thought that gave the team more balance, but per FBref, Fulham were about 0.4 expected goals better when Mitrovic was on the field, compared to when he was off of it. We’ll see Fulham again in two years.

19. West Brom (Prediction: 19)

Unimpressive Championship team without a good defense struggles in the Premier League: it’s a tale as old as time.

20. Sheffield United (Prediction: 11)

I, uh, had these guys finishing a spot ahead of Leeds. Last season, they finished ninth, with an even goal differential and xG numbers about equal to Tottenham’s. It was a super-impressive start to life in the Premier League for Chris Wilder and Co., but there was one massive warning season: they, too, could not score. They won 54 points despite only scoring 39 goals; that’s basically the absolute max in terms of points-per-goal. To fix that, they signed 20-year-old Rhian Brewster from Liverpool for £24 million ... and he proceeded to not score or assist a single goal the entire season. It was a bad move, from basically every angle, but also no one could’ve expected it to go that badly. 

The Blades ended the season with just 20 goals, and it was even worse on the other end, as a team that did it with their defense in year one conceded 63 goals in year two. Pretty much everything went wrong for them; they weren’t even clearly the worst team in the league over the first half of the season, but still made it through that stretch with only two points. Eventually it seemed like their performances started to match their results -- rather than the other way around -- and they ended the season as quite clearly the worst side in the competition.

The lesson I take from this one? If a team completely shocks the league with a collection of late-career unknowns and unique tactics, don’t expect them to do it again unless they continue to evolve.