With the Premier League neatly poised at its 190 match mid-point it’s a good time to take stock of where things are.
I use my Adjusted Goals rating measure to assess teams’ performance. This isn’t anything too complicated – it’s simply uses a combination of goals, shots on target and shots to rate each team’s attack and defence. The formula is:
Adjusted goals = 45.0%*Goals + 8.4%*”Shots on target” + 2.8%*Shots, I’ve found this measure predicts future performance better than goals scored or conceded (or points attained) as explained here.
1st half Performance
Manchester City are rated the top team, but it’s notable that Spurs are second – performing well in both attack and defence (2nd place for both – behind Man C for attack and Man U for defence). Arsenal are only 3rd – which I know is contrary to many expected-goal models that rate them higher. At the bottom end, surprisingly Sunderland and Newcastle are lower ranked than Villa.
Improvement from last season
To assess improvement, I’ve compared teams’ adjusted goals rating from last season. Spurs again take the honours, even managing to better Leicester’s remarkable half season. Chelsea have suffered the greatest decline. Obviously.
Over/ Underperformance (or luck)
This one’s a bit tenuous, but I use the adjusted goals measure to work out what points a team would normally get for a given rating. It points to Leicester over-performance of nearly 8 points and Villa underperformace of the same amount. Again there’s good news for Spurs here- they could have expected 3.5 points more from their 1st half performance.
This season appears more unpredictable than recent seasons, but it needs measuring. I’ve done this in a couple of ways. Firstly correlation between 1st half season adjusted goals and previous full season. Adjusted goals normally correlates well from season to season. Correlation coefficient of 68% is much lower than any of the prior 6 seasons.
I’ve also looked at the market return for a unit bet on every match – a highly positive return indicates that the market is poorly assessing true match probability. A current yield of 10% is far higher than any recent full season (Chelsea and West Ham have been responsible for much of this) .
This season has been unpredictable so far, which is great – because it shows that even in this world of big data the market can still get it wrong.
Average goals scored per match (2.61) have been low compared to recent seasons, but similar to last year’s 2.57. More notable is the diminishing home field advantage –it’s only 0.25 goals so far this season. It will be interesting to see if this continues.
The second half.
The Premier League is beautifully poised for a competitive finish. In my projections I have Manchester City and Arsenal neck and neck for the title.
For future forecasts I use a combination of this season’s and last season’s adjusted goals. Currently it’s a 50/50 weighting, as we’re half way through the season. Because of this I still don’t rate Spurs as highly as Arsenal (or Man City).But if Spurs can maintain their current performance level they’re in with a great chance too. Add the Leicester, Chelsea and Man U stories into the mix and we’re in for a fascinating second half.