Regular readers will know that I spent last season compiling chance quality information for the Premier League, to see which teams created scoring opportunities in the best and worse areas. Simon Gleave is compiling pre-season predictions on how the league table will look at the end of 2014/15 season, so I’ve decided to give it a go using this system.
It may turn out that chance quality is not consistent from season-to-season (and indeed I do expect this to be the case), but there’s only one way to find out, so I will attempt to forecast the final 2014/15 Premier League table using this data.
A quick reminder of how the chance quality data is calculated. Take a look at this pitch map, which explains the six different chance types I record.
Using the average conversion rate for chances in each of the six zones, and knowing how many chances in each area the teams created and allowed, I was able to calculate what a team’s ‘expected assists’ figure would be for both ends of the pitch.
Before deciding how to use this data for my 2014/15 forecast though, I had to see how various uses of the data compared to last season’s actual league table. In the same vein as standard goal difference (attacking score minus defensive score), I used four types of difference for a comparison:
- Chances Created – a simple look at the raw figures
- Central Box Chances – as these are converted at the highest rate, the more a team creates and fewer they allow should determine their quality.
- Average Chance Quality – what was the percentage chance that a team’s created opportunities would be scored?
- Expected Assists – as mentioned above, by using league average conversion rates, how many assists should a team have seen at both ends of the pitch?
I then ranked the teams by above the measures, to see how different the orders were from the actual 2013/14 Premier League table. Here’s what I found:
There was a clear winner; Expected Assists Difference. On average, teams were only 2.3 places away from their actual league finish using this data, and sixteen of the twenty teams (80% of them) were within three places of their true league placing. Here is a table of the teams sorted by their Expected Assists Difference, which also shows how that compared to the actual top flight table.
We can see that there were four outliers of five places or more: Southampton and particularly West Bromwich Albion underachieved against this metric, whilst Crystal Palace and West Ham displayed the merits of playing direct football whilst being hard to beat by comfortably outperforming their expected results. More importantly though almost half of the teams were within one place of their genuine league standing, which seems a decent verification of the system.
To be able to predict next season’s table though, I also needed to factor in the three promoted teams (Leicester, Burnley, and QPR) and I don’t have chance quality data for the Championship.
I have however been able to locate the number of shots on target these teams had themselves and allowed their opponents, and I found with the Premier League that on average a team’s shots on target ratio (the proportion of the shots on target in their matches that they had) was within 2.2% of their Expected Assists ratio, so I used this for the promoted teams.
It would be foolish to think that they can match their Championship output in the top flight though, so I amended the shots on target figures by twenty-percent in each direction; fewer themselves with more for their opponents. I seem to recall Ted Knutson claiming a Championship player’s output dropped by twenty percent when moving up to the Premier League, hence why I used that figure. Blame Ted if it proves to be inaccurate!
The only other amendment I made to the above findings was to move Southampton below the superior seven teams, as it would’ve been crazy to forecast them to finish third, even before they lost most of their key players in the summer. However, they could perform nineteen percent worse at both ends of the pitch than they did last season and still be eighth in this table on merit, so it seems a reasonable place to put them.
I also wanted to include a forecast of how many points the teams will get. As this is the first season of compiling this data though, I had no idea how the expected assists translates to points; I couldn’t simply use last years figures, as the teams are in a different order.
I recently wrote how the number of points for the top teams seems to be lower following World Cup summers, so I have used that as the basis for my points predictions. The figures I have used are the average for that league position in the last four seasons that have been directly preceded by a World Cup.
Putting all of that together gave me this; my Premier League table forecast for 2014/15:
Whilst I would be amazed if this turns out to be entirely accurate, at the same time it certainly doesn’t look wholly unreasonable either. The points totals do perhaps look a little low, but as I said at the beginning of the piece, I don’t particularly expect this to be entirely accurate anyway. Perhaps this should be thought of more as an experiment rather than an outright prediction, but it will be interesting to see how it accurate is at the end of the season.