23/24 - The Context #1
Early observations from Gameweeks 1 - 4.
Hello - welcome to The Context #1 of the 23/24 season.
Football data is always better with context. This series aims to provide additional insights into my Premier League team strength model by delving into the underlying data. Future editions will incorporate footballing perspectives from fans in the future, potentially as a blog post or a podcast. I have tried to respect your time and keep this balanced and short - your feedback and suggestions for the future are highly appreciated.
In this issue, we'll focus on four teams I found interesting: Brighton, Brentford, Wolves, and Everton. Let's begin with the latest Model Snapshot:
Brighton
Attack Strength: 1.71 (+0.02)
Defence Strength: 1.38 (-0.06)
Overall Strength: 1.33 (-0.04)
The Context: Brighton’s attack continues to improve and defy expectations - pick an attacking metric and they are topping the league table in it. But that isn’t the real story here. Brighton's Overall Strength has suffered a notable drop of -0.04 since the start of the season and this decline is predominantly due to a substantial decrease in their Defence Strength (-0.06 which is the largest defence decrease in the model). For a long time, the miracle of RDZ was turning the attack to 11 whilst maintaining most of the defence strength. The evidence so far suggests the miracle is finally fading on the defence side.
They currently sit at approximately 17th for xGA having played Luton, Wolves, West Ham and Newcastle. Despite winning three of those games (and winning on xG), an opening 4 fixtures like that should not see Brighton giving up ~8 xGA. The model would roughly expect Brighton to concede ~3.3 xG from these opening fixtures.
The much-talked-about West Ham game is an interesting example:
xThreat: 2.73 - 0.81
Possession: 78.0% - 22.0%
Field Tilt: 84.5% - 15.5%
Def Action Height: 52.9 - 29.4
Just looking at the above data, you would probably not guess that the xG and goal outcomes were 1.5 - 3.0 and 1 - 3. Brighton may dominate games across virtually all the attacking metrics but they are giving up too many chances through their playstyle.
Stat I enjoyed: this sums up a lot I think - Brighton ranks 20th in the league for xG conceded from counters, 0.53 p90 compared to the Premier League average of 0.14 p90.
Brentford
Attack Strength: 1.40 (+0.05)
Defence Strength: 1.39 (-0.01)
Overall Strength: 1.02 (+0.04)
The Context: There were doubters - but even believers did not expect a start like this. They currently sit 11th in the table but on xPoints would sit ~3rd. Brentford's Overall Strength has seen a significant improvement of +0.05 since the beginning of the season. This remarkable progress is primarily driven by a substantial increase in their Attack Strength, highlighting their improved offensive capabilities even without the presence of Ivan Toney.
Pick your xG model and you will find Brentford sitting right near the top of the xG and npxG tables. The fixture schedule has been relatively kind but ~2.1 npxG p90 has seen Brentford surpass model expectations every week so far. For context, my model has been higher on Brentford than both other models and the bookies for ~2 seasons now - which makes the 0.05 attack improvement even more impressive.
As expected, Brentford’s attack continues to be leaning on set-piece play - having generated the most xG from set-pieces this season (0.68 p90 to the league average of 0.38 p90).
Stat I enjoyed: early days but Brentford continue the trend of forcing low shot quality against them = 0.08 npxG conceded per shot compared to the league average of 0.12.
Wolves
Attack Strength: 0.98 (+0.05)
Defence Strength: 1.61 (-0.01)
Overall Strength: 0.36 (+0.03)
The Context: I think this is a good time to give a reminder that when I talk about strength it is from a model perspective. When a team’s strength is changing in the model, it is not saying that the team’s actual strength is ‘objectively’ increasing/decreasing (although that is very likely happening… I hope) - what it is saying specifically is that the team’s strength is changing according to the model. In other words, the model is changing to be a more accurate reflection of that team’s strength. Please feel free to read the FAQ if you want more information on how the model changes.
That was a long-winded way of saying I think Wolves are very bad. I think they have been bad for quite a while now - plummeting in strength pretty much since the end of the Nuno era. O’Neil seems to be getting a tune out of their attack but it is hard to know how much of the +0.05 attacking strength improvement is due to a genuine improvement or whether I was too low on Wolves to begin with. Any input from Wolves fans would be appreciated!
Stat I enjoyed: Wolves rank first in the league for Successful Dribbles (60) and Tackles Won (53).
Everton
Attack Strength: 1.19 (+0.05)
Defence Strength: 1.64 (-0.03)
Overall Strength: 0.55 (+0.02)
The Context: I can’t quite remember the last time a team got themselves into such a finishing deficit this early in a season. Only 4 games in, Everton sits bottom of the Goals - xG table, with just 2 goals from ~8 xG. Almost all of this xG was put up against Fulham, Wolves and Sheffield United. On a more ‘neutral’ look at the fixture schedule it might seem unimpressive, but that amount of xG is up there with the big teams who have had an equally nice opening set of fixtures.
In the context of the model, Everton was expected to put up ~4.2 xG across these fixtures. Although the goals are not there right now, this is a significant amount of overperformance and the performances seem to suggest the output will continue. Everton is rivalling Brentford with xG from set pieces, 0.66 p90 compared to the Premier League average of 0.38 p90.
Stat I enjoyed: Early days but Everton’s average Field Tilt % has improved from 43% last year to 55% this year.
Thanks for reading and making it until the end! I am just doing what I enjoy and find interesting but I am happy people want to come along for the journey. If you enjoyed this or have any feedback/input on what you would like to see in the future as I mentioned in the opening paragraph - then please do let me know directly on Twitter!
elevenify