I think this is about as close to plagiarising my own work that I can possibly get. After looking at whether it’s possible to predict how relegated teams will perform in the championship I’m going to go ahead and look at teams going the other way and get promoted to the premiership. First the ‘average’ season for the group as a whole, with points and ± on a per game basis
|Championship (‘Year 1’)||1.88||+0.67||0.564||1062|
|Premiership (‘Year 2’)||1.00||-0.56||0.426||972|
So the promoted teams don’t get on so well in the premiership (as a group their SOTR decreases and their PDO regresses to the mean). Again this is expected, the quality of competition is substantially higher in the premiership than the championship. Obviously some of the teams that are promoted are better than others (for example we’d expect the champions to typically be better than a sixth placed team that goes up via the playoffs) so what I really want to know is whether it’s possible to predict how an individual team will perform rather than the population as a whole.
To do this I’ve taken the performance of each team and made a plot of the values in ‘year 1’ against the values in ‘year 2’. The correlation (R2 value) has been determined and is shown in the table below (the higher the R2 value the stronger the correlation).
|Variable||Correlation (R2 value)|
As for relegation the values are all still very low but there is one obvious outlier. Shots on target ratio gives an R2 value that is approaching statistical significance and seems to be a good indicator of future performance and this lines up well with what I observed when I was examining future performance in the premiership.
There’s also another chance here to highlight how important shots are compared to points and goals when it comes to predicting future performance. This table follows the same methodology as above to compare the correlation (R2 value) of performance in successive seasons for goals, shots on target and total shots.
|Shots on target||0.503||0.522|
Bear in mind that points in successive years return an R2 value of 0.026 and the relative predictive power of shots becomes apparent.
So can we predict how a promoted team will perform in the premiership? Well we certainly have a better idea than teams relegated to the championship but the correlation is still relatively weak. Overall I’d be surprised if it was possible to consistently predict within 5 points how the teams will finish.