ESPN put out deep dive statistically based end of season Premier League predictions

In partnership with

the Chelsea News

·04 de novembro de 2025

ESPN put out deep dive statistically based end of season Premier League predictions

ESPN writers have put out a a deep dive statistically based end of season Premier League predictions list.

Obviously when I saw this come out I was keen to see where they were putting Chelsea at, and I think it will come as little surprise to see where they believe we will finish.

Vídeos OneFootball

EXCLUSIVE! When Cole Palmer is expected to return to the Chelsea squad – plans laid out!

But we’ll come to that in a second, first and foremost, how did ESPN forecast this and come to their conclusions here?

The author, Ryan O’Hanlon, writes:

‘Rather than doing a much more complex and perhaps much more mathematically sound projection, this approach will be something simpler: (1) because it’s easier to understand, and (2) because that makes it easier to glean insights from what the projection tells us.

‘I’ve written about this so many times that I fear I will soon develop a very specific kind of hand arthritis from the repeated motion of my fingers toward these letters, but my favorite single-number to project future team performance is what I call “adjusted goal differential.” This comes from a study by former AC Milan data researcher Ben Torvaney, which discovered that a blend of 70% expected goals and 30% actual goals better predicted future point totals than either goals or xG alone or any other blend of the two.

‘Knowing that, I then looked at the relationship between a team’s adjusted goal differential after 10 games and their points won over the final 28 matches. Based on data from the past 10 Premier League seasons, a team with a neutral adjusted goal differential after 10 matches would be expected to win 1.39 points per game for the rest of the season. And then every goal increase of adjusted goal differential increases the point per game expectation by 0.47 points.

‘Of course, 10-game performance can’t tell us everything; otherwise, we could just shut up shop right now and give these guys seven months off until the men’s World Cup.

‘And, well, there is one number that’s even more powerful in predicting future performance than this blend of goals and expected goals: how much teams pay their players. Even more so than the 10-game performance horizon, the correlation between player wages and team performance has been heavily studied. Most recently, analyst Paul Johnson has looked at how the estimated player values on Transfermarkt can predict how many points a team wins in a season. He found, among other things, that a 10% increase in a team’s value increases a team’s point total in a season by about a point and a half.

‘Now, neither of these numbers can tell us everything, but I’m using them because there’s an intuitive logic to each one. Adjusted goal difference acknowledges the importance of creating better chances than your opponent as the main driver in team success, but it also accounts for the fact that sometimes teams play differently after a goal is scored and it also picks up some of the stuff for which xG models can’t fully account.

‘And then market values are just a proxy for talent. Even the best managers will tell you that player quality is the main driver of team performance. But sometimes a bunch of great players play poorly at the same time, or a bunch of good players get injured at the same time or a talented group of players needs more time to figure out how to play together. Adjusted goal differential wouldn’t account for any of this, but we can all agree that these things happen. The market value value input at least attempts to account for this.

‘Finally, to figure out how to weigh each input, I looked at betting projections for season point totals. Given the financial incentives at play and the continued, insanely profitable existence of sports books, these are going to be the most accurate publicly available “projection systems.” I then chose whatever weight brought these projections closest to being in line with the markets: 64% transfer values and 36% adjusted goal differential.’