Trying to predict the exact order that 20 soccer teams will finish in a league table is a fool’s errand. While the opinions of pundits are often wide of the mark, the use of predictive models will help us make more accurate predictions. Can expected goals help us predict the Premier League table? Read on to find out.
In theory you have far more chance of correctly predicting the winning numbers for the lottery than the Premier League table after a 38-game season. Or at least you would if all 20 teams were identical; in reality we know that this is not the case.
Anyone with even a passing interest in the Premier League could easily make a sensible estimate of where most clubs will finish in May - the same teams regularly feature at the top of the table and newly promoted teams will often be favourites in the Premier League relegation betting.
- Expert insight: Outright Premier League betting.
The benefit of using expected goals
Because having a very specific idea of where in the league a team will finish can make a profit in terms of betting, using a system as opposed to a vague hunch will benefit bettors.
In a previous article we looked at how to calculate expected goals (or ‘xG’ for short). That article focused on how to use that information to forecast the outcome of individual matches. However, we can also use this process to look at the previous season’s total data to predict the following season’s Premier League table.
Stoke finished 9th in 2015/16, whilst only ranking 13th for expected goals. Where did they finish in the Premier League the following season? Thirteenth.
The 2015/16 season was a remarkable one in the English top flight. Who would’ve predicted that Leicester would win the title, or that the reigning champions Chelsea would stumble home in tenth place? But despite the unusual look to the final table, we could’ve used the total expected goals figures from that season to predict the final 2016/17 table with a reasonable degree of accuracy.
By using the expected goals both for and against for the 2015/16 campaign as a whole, we can get each team’s expected goals difference and then make an alternate ‘expected’ league table.
If we’d have then used that for our prediction for 2016/17, we’d have found in May that teams finished on average 2.9 places away from their expected goals difference rank from the previous season. That’s allowing for the fact that Leicester sank back to mid-table whilst Chelsea surged up to take the title back, too.
When using data from a previous season, the model is not aware of heavy transfer spending or a new star player who has a major impact on team performance.
So why not just use the previous season’s league table and assume it will repeat again? After all, most teams finish in roughly the same place year after year. By using xG we can eradicate the random nature of finishing at both ends of the pitch and get a truer reading of each team’s overall quality.
Additionally, when using the 2015/16 table to predict the final standings in 2016/17, we find that teams moved by 3.9 places on average. This illustrates that the expected goals method was more accurate.
Premier League predictions: How accurate is expected goals?
There were some interesting examples of where xG difference from 2015/16 gave a greater insight into a team’s finishing position in 2016/17 than the league table did. For instance, Stoke finished 9th in 2015/16, whilst only ranking 13th for expected goals. Where did they finish in the Premier League the following season? Thirteenth.
Despite the unusual look to the final table in 2015/16, we could’ve used the total expected goals figures from that season to predict the final 2016/17 table with a reasonable degree of accuracy.
West Ham surprised a lot of people by finishing 7th in 2015/16, but their expected goals figures suggested they were in fact the 10th best team, and they duly came 11th a year later. Meanwhile Liverpool trundled home in 8th as they focussed on their Europa League commitments in Jurgen Klopp’s first campaign, but their underlying numbers ranked them as 5th; and they finished 4th last season.
There are also some other interesting examples from further back. Newcastle surprised a lot of people (not least those of us with an interest in expected goals) in 2011/12 by finishing 5th in the Premier League. They actually ranked a lowly 16th on the xG chart that year, which is exactly where they came in the 2012/13 league table.
That season was Brendan Rodgers’ first at Liverpool and it was underwhelming on the face of it as the Reds finished 7th. Their expected goals difference was the second best in the division though and it all clicked into place the following year as they came incredibly close to winning the league.
The limitations of expected goals
As with every system, expected goals is certainly not infallible. When using data from a previous season, the model is not aware of heavy transfer spending or a new star player who has a major impact on team performance. Clubs often change managers during the close season, and new tactics may render the form from the previous season a distant memory.
Teams changing stadia - which is often seen as a disadvantage - is also becoming a more regular occurrence in modern soccer; will Spurs be able to repeat their second place finish in both the xG and actual tables from 2016/17 now that they will be playing their home games at Wembley?
Using last season’s expected goals data suggests that Manchester City will pip Tottenham to the title next May, with Chelsea and Manchester United making up the rest of the top four. It will be interesting to see if that’s correct after all 380 games have been played in the 2017/18 Premier League.
Ready to use expected goals in betting? Get the best Premier League odds online at Pinnacle.