Why did Brighton start so well in the Premier League and where will they finish?

Brighton are fourth in the Premier League – a Champions League place. Last season they finished 16th, two places above the relegation zone. On expected goals, though, they ranked fifth. According to expected goals they should have qualified for the Europa League.

Were Brighton’s expected goals stats from last season a clue as to how well they would start this season? And will they carry on getting such good results?

Brighton are below Chelsea, Liverpool and Manchester United. They are above Manchester City. Last season on expected goal difference they were below only those four teams: City, Chelsea, Liverpool and United.

What are expected goals and what do they tell us?

They are an attempt to answer this question: with the same number of shots from the same positions, how many goals would an average team score?

Expected goals remove two things that can be found in real goals: skill and luck. Expected goals ignore differences in skill and differences in luck. The last bit is helpful but the first bit is not.

Skill is enduring. Not all teams are average. Some players have higher than average skills, others have lower than average skills. If one player was more skilful than another last season they will be more skilful again this season.

Luck tends not to be enduring. Some teams have been luckier than others. But if one team were more fortunate than another last season they probably will not be more fortunate again this season.

How well do expected goal difference and actual goal difference in one season predict actual goal difference in the next?

I studied six pairs of seasons – 2014-15 and 2015-16 to 2019-20 and 2020-21 – in five top divisions: the English Premier League, Spanish La Liga, German Bundesliga, Italian Serie A and French Ligue 1. I used Understat’s expected goals.

Expected goal difference from last season was better than real goal difference from last season as a predictor of real goal difference in the next season. This suggests that disparities between expected and real goal difference in one season are due more to disparities in luck, which are likely to be temporary, than to disparities in skill, which are not. But how much more?

I found that I could have obtained even more accurate predictions of real goal difference in the next season by using a mixture of the previous season’s expected and real goal differences. The optimum mixture of the previous season’s data was one part real goal difference to two parts expected goal difference.

This suggests that discrepancies between expected and real goal difference in a season may be explicable more fully as one-third discrepancies in skill, two-thirds discrepancies in luck.

I applied that formula to real and expected goal differences from last season to predict real goal differences for this season. The prediction for Brighton this season put them ninth – higher than they finished in the Premier League last season, but lower than they had ranked for expected goals.

There are other ways of evaluating performances and potential. Many would have projected a finishing position for Brighton this season lower than ninth. Probably last season Brighton were better than their results suggested, but not as good as their expected goals stats suggested.

Where might they finish this season?

So far it has been the opposite of last season. Brighton’s results have been better than their expected goals. They have played five games, winning four and losing one. They have scored seven goals and conceded four. Undertstat have expected goals totals for Brighton of six for and against. So do other compilers I checked.

I compute ratings from different stats, including shots. They anticipate Brighton gathering points less quickly over the remaining 33 games. They put Brighton’s most likely end-of-season total as 53 points. This is inside Spreadex’s quote, so it does not seem unreasonable. Teams with 53 points typically finish ninth.

How second thoughts can help us to make better bets

Talking about real and expected goals I said that one gave better predictions than the other but the best predictions came from combining them. Over and again I have found that I can get more accurate predictions by combining two or more bits of information.

If I am using a formula I need make the prediction only once. Otherwise, I now know, I should do it twice.

Researchers Stefan Herzog and Ralph Hertwig explained why in a paper published in the Journal of Experimental Psychology. I read about it in a book called Noise that I mentioned a few weeks ago – a superb book.

Herzog and Hertwig asked people to estimate something. Then they gave them these instructions:

“First, assume that your estimate is off the mark. Second, think about a few reasons why that could be. Which assumptions and considerations could have been wrong? Third, what do these new considerations imply? Was the first estimate rather too high or too low? Fourth, based on this new perspective, make a second, alternative estimate.”

Herzog and Hertwig found that the most accurate predictions were not the first, or the second, but an average of the two.

This is what I can add, based on what I have learned from all the times I got things wrong. If your first estimate was that someone would do really well, it is more likely to have been too high than too low. If your first estimate was that someone would do really badly, it is more likely to have been too low than too high.

Not got a bet365 account? Sign up today and get up to £100 in bet credits

Up to £100 in Bet Credits for new customers at bet365. Min deposit £5. Bet Credits available for use upon settlement of bets to value of qualifying deposit. Min odds, bet and payment method exclusions apply. Returns exclude Bet Credits stake. Time limits and T&Cs apply.

CLAIM OFFER HERE

MORE FREE BETS

Today's top sports betting stories

Follow us on Twitter @racingpostsport