Predicting how all 20 Premier League teams close the season

Even through 20 of 38 games, the Premier League table doesn't predict the future all that well.

At this point last year, Arsenal were five points clear of second-place Manchester City. Liverpool were tied with Brentford in seventh. Aston Villa were in 11th, a point back of Chelsea. Leeds United were five points clear of the relegation places, which were occupied by Southampton, West Ham and Everton -- all tied on 15 points.

Over the remaining 18 matches, Manchester City won nine more points than the second-best pair of teams, one of which was Liverpool. Arsenal (fourth most) won only one more point than Aston Villa (fifth most). Chelsea won more points than only two other teams -- total -- one of which was Leeds United. Meanwhile, Everton finished with the 14th-most points over the home stretch, and West Ham gathered the 10th-best haul from the last few months of matches.

On average, a team's rank after 20 games shifted by 3.5 spots when compared to its rank over the final 18 games last season. In fact, Brentford were the only team to rank in the exact same spot over the first 20 games and the final 18.

Of course, each team's place in the final table was significantly stickier -- the points accrued in those first 20 matches account for more than half of the final point total. But when you look at this season's current table, there's still a good deal of unpredictability between how each team has played so far, and how they're going to play from here on out.

Today, we're going to try a slightly different prediction exercise. We're going to rank all 20 Premier League teams based on how many points we think they're going to win from now until the end of the season. So, if the table were erased and we played only the games left on the schedule, how would things look come the end of May?

Our Premier League prediction methodology

When it comes to making better predictions, the wisdom of the crowds is maybe the most powerful tool. Everyone has their own biases, and everyone also probably has their own useful bits of insight, too. So when you aggregate all of those opinions or predictions together, you're much more likely to be closer to the truth than any one individual prediction is.

As Phillip Tetlock, perhaps the world's preeminent expert in the mechanics of making better predictions, writes in the book Superforecasting: "Aggregating the judgment of many consistently beats the accuracy of the average member of the group." So, to abide by this idea, we're going to turn to the betting markets, which are essentially one big wisdom-of-the-crowd model, aided by the added incentives of both sides of the bets (the sportsbook and the bettor) either losing or winning money based on their predictions.

The basis of our prediction, then, comes from Sporting Index, which has a constantly updating point-total projection for each Premier League team. Bettors can place bets if they think the ultimate total will be higher or lower. From there, I've made a couple minor, personal adjustments based on what I think will happen. For the best possible overall prediction, perhaps I should just keep it exactly as the betting odds suggest, but that's no fun.

Below, we're listing each team's current point total, along with the Sporting Index projection. And we're also including each team's per-game non-penalty expected-goal (xG) differential, which is the single-best statistical predictor of how the team will perform going forward. Unless otherwise noted, all stats come from Stats Perform.

With that out of the way, let's get to the predictions.

  1. Manchester City

• Current point total: 40 (tied for third)
• Projected rest-of-season total: 44 (first)
• Non-penalty xG differential: minus-0.9 (second)

This seems like the easiest prediction of them all. The defining question for the rest of the season isn't, "Will Manchester City win the most points from here on out?" Rather, it's, "How many more points will Manchester City win than everyone else?"

We're also cheating slightly here. City have 19 games remaining, while all but three other teams have only 18 games left. With a game in hand, plus Kevin De Bruyne, Erling Haaland and Jérémy Doku all returning from injury, it's really hard to see a world where City don't top the table from now until the end of the season.

 2. Liverpool

• Current point total: 45 (first)
• Projected rest-of-season total: 37 (second)
• Non-penalty xG differential: plus-0.91 (first)

On the one hand, Trent Alexander-Arnold, Mohamed Salah, Wataru Endo, Dominik Szoboszlai, Thiago, Andy Robertson, Joël Matip and Kostas Tsimikas are all either out injured or away on international duty. On the other hand, their underlying numbers are as good as City's, and they don't have the Champions League to worry about.

I give them the edge over Arsenal because they're more likely to be in the title race, so they're more likely to need to focus on every league match. The Gunners, meanwhile...

 3. Arsenal

• Current point total: 40 (tied for third)
• Projected rest-of-season total: 36 (third)
• Non-penalty xG differential: plus-0.87 (third)

Although I'd put their recent slump down to little more than finishing variance, the dropped points have significantly hampered their chances at winning the league.

Throw in the fact that they're third favorites, per ESPN BET, to win the Champions League, and it seems like there's a very real possibility that the Gunners have a top-four spot locked up later in the season and then spend a couple weekends resting players amid a deeper run in Europe.

 4. Chelsea

• Current point total: 28 (tied for 10th)
• Projected rest-of-season total: 30 (fifth)
• Non-penalty xG differential: plus-0.37 (fifth)

Here's where it starts to get really tricky. I have a high degree of confidence that City wins the most points from here, and then a not-quite-as-high degree of confidence that Liverpool and Arsenal -- in some order -- rank second and third. But now we enter a group of, I don't know, seven teams with the possibility of winning the fourth-most points.

Chelsea have been much better than their point-total suggests this season, and I think their massive squad gives them a higher potential of in-season improvement than any of the other teams below them. At least, there's a much bigger chance that there's some unused configuration of Chelsea players -- if manager Mauricio Pochettino finds it -- that will significantly raise the team's performance level. At the same time, there's also a chance that Pochettino gets fired, it all goes to hell and they finish 10th.

 5. Aston Villa

• Current point total: 32 (second)
• Projected rest-of-season total: 28 (tied for eighth)
• Non-penalty xG differential: +0.38 (fourth)

The markets aren't buying Villa's current point total or their current performance level. I understand the former, but don't feel as strongly about the latter. While I didn't expect this group of players to be this good, Villa have been a top-four team, if not better, for more than a full year under Unai Emery at this point. I trust them to perform at a competent level more than I trust anyone below them -- or even the team I just picked above them.

 6. Everton

• Current point total: 16 (17th)
• Projected rest-of-season total: 32 (fourth)
• Non-penalty xG differential: +0.33 (seventh)

Uh ... what?

Two things are happening here: The first thing is that Everton have been impressive this season. That xG differential is nearly just as good as Villa's. And the second thing is that there's a decent chance their point deduction gets overturned, and they suddenly receive an extra 10 points.

That projected number seems to suggest that it's likely Everton will get the 10 points back. But I'm gonna take a slightly more conservative approach here and bump them down a few slots from where the markets have them. Also, though: read back that sentence! I just said that it's "conservative" to say Everton will win more points than 14 other teams in the Premier League before the end of the season. Second-season manager Sean Dyche might be special.

 7. Newcastle

• Current point total: 29 (ninth)
• Projected rest-of-season total: 29 (tied for sixth)
• Non-penalty xG differential: +0.25 (ninth)

This is what Newcastle's non-penalty xG differential looks like for every match this season, in order:

Up through November, they were out-creating their opponents consistently ... and then it all fell apart. Among this glut of seven or so teams, no one has come close to showing the same ceiling that we saw with Newcastle over the first three months of the season. Unfortunately, no one has shown a lower floor, either.

 8. Brighton

• Current point total: 31 (tied for seventh)
• Projected rest-of-season total: 28 (tied for eighth)
• Non-penalty xG differential: +0.32 (eighth)

They have the eighth-best underlying numbers, and the betting markets project them to win the eighth-most points. No need to overcomplicate this one.

 9. Tottenham

• Current point total: 39 (fifth)
• Projected rest-of-season total: 29 (tied for sixth)
• Non-penalty xG differential: +0.15 (10th)

Ange Postecoglou & Co. are outperforming their non-penalty xG differential by a larger margin than any team in the league this season: +17 from an xG differential of +3.01.

At this stage of last season, Fulham, too, were about 13 goals ahead of their underlying numbers. Over the first 20 games, they averaged 1.55 points per match. Over the final 18, it dropped to 1.17. Tottenham might be headed for a similar kind of slowdown.

 10. Brentford

• Current point total: 19 (16th)
• Projected rest-of-season total: 21 (tied for 13th)
• Non-penalty xG differential: +0.35 (sixth)

Brentford might be without Bryan Mbeumo for the rest of the season, but Ivan Toney is returning from his gambling suspension later this month. They have an extra game in hand, and their level of play has been quite good despite their results.

However, the market remains quite low on Brentford's prospects for the rest of the season. Perhaps that's because of the chance that Toney leaves during the transfer window? I'm buying their process, though. It wouldn't shock me if they won, say, the seventh-most points from here on out.

 11. Manchester United

• Current point total: 31 (tied for seventh)
• Projected rest-of-season total: 27 (10th)
• Non-penalty xG differential: -0.22 (14th)

They're going to get healthier as the season goes on, and some of the performances this season have been almost impossibly bad, given the amount of money invested into this roster. It should improve.

But such is the state of things at United that the 10th-best rest-of-season points projection actually seems slightly bullish when compared to their performance level over the first 20 games of the season.

 12. Bournemouth

• Current point total: 25 (12th)
• Projected rest-of-season total: 24 (11th)
• Non-penalty xG differential: -0.14

Take that image from the Newcastle section, flip it upside down, and you've basically got Bournemouth's season:

Through October, they produced a -0.89 xG differential -- fourth worst in the league. From November on? It's +0.70 -- fourth best in the league, behind only the top three teams on this list. In his first season in England, it sure seems like manager Andoni Iraola and his players have started to figure it out.

 13. West Ham

• Current point total: 34 (sixth)
• Projected rest-of-season total: 23 (12th)
• Non-penalty xG differential: -0.31 (16th)

Yes, they're five points out of fifth. But they're also a below-average attacking and defending team:

Per the site FBref, West Ham's average age (weighted by minutes played ) is 28.8, second oldest in the league. They've been running hot, and there aren't many reasons to expect any kind of organic, internal improvement either.

 14. Wolverhampton

• Current point total: 28 (tied for 10th)
• Projected rest-of-season total: 21 (tied for 13th)
• Non-penalty xG differential: -0.15 (12th)

Let's loop these guys in with the next team...

 15. Crystal Palace

• Current point total: 21 (14th)
• Projected rest-of-season total: 21 (tied for 13th)
• Non-penalty xG differential: -0.21 (13th)

I used to think that Crystal Palace were the platonic ideal of a mid-table team: no real upside, nothing particularly interesting about how they try to win games, and no real risk of being relegated. This season, though, Wolves are staking their own claim as the masters of mid-table mediocrity. These two clubs play on May 11th, in what might be the lowest-stakes game in the history of the Premier League. I wouldn't have it any other way.

 16. Nottingham Forest

• Current point total: 20 (15th)
• Projected rest-of-season total: 18 (17th)
• Non-penalty xG differential: -0.27 (15th)

Since bringing in Nuno Espirito Santo to replace Steve Cooper as manager, Nottingham Forest have beaten Newcastle and Manchester United, and they lost to Bournemouth on a 94th-minute goal despite going down a man in the 23rd minute. The bullish case for Forest has always been that somewhere, within that randomly assembled 800-man roster, there was hidden an effective combination of 11 professional soccer players. Has Nuno found it?

 17. Fulham

• Current point total: 24 (14th)
• Projected rest-of-season total: 21 (tied for 13th)
• Non-penalty xG differential: -0.40 (17th)

They're the oldest team in the league (an average age of 29.3), and they have the worst underlying numbers among any of the non-promoted clubs. Unless something changes, Fulham will be serious relegation candidates next season.

 18. Luton Town

• Current point total: 15 (18th)
• Projected rest-of-season total: 15 (19th)
• Non-penalty xG differential: -1.11 (19th)

The big advantage Luton Town have on the other two teams in the relegation zone doesn't have anything to do with coaching or talent or tactics or schedule strength. No, their big trump card is that they have an extra game left to play.

 19. Burnley

• Current point total: 11 (19th)
• Projected rest-of-season total: 17 (18th)
• Non-penalty xG differential: -0.82

After a brief period of improvement at the beginning of December, Burnley have been terrible since. Vincent Kompany's measured approach as manager -- Burnley move the ball more slowly upfield than any team in the league other than Manchester City -- just doesn't work at this level, with these players. I think they're probably better than Luton, in a vacuum, but not better enough to make up for the fact that they have one fewer game remaining.

 20. Sheffield United

• Current point total: 9 (20th)
• Projected rest-of-season total: 13 (20th)
• Non-penalty xG differential: -1.19 (20th)

Look at the numbers inside of the parentheses. It's bad. It's real bad.