Open play xA+xG mapped onto position
All stats from Football Whispers
With analytics it is easy to get sucked into thinking you are producing useful and meaningful content. Most of the time it is just interesting.
I think this is probably the case for this, but I'm going to do it anyway because interesting is better than nothing.
Today I'm looking at Open play expected goals + open play expected assists broken down by playing positions.
My hypothesis is that Everton are relatively poor on goal/assist contribution from midfielders. For this to be true I would expect that other teams generate a greater proportion of their open play goals and open play assists from midfield than Everton.
So for this I am going to take the Football Whispers data for each team, then total the Open play expected goals + expected assists for each player under the categories defender, midfielder and forward. If I wanted to I could break it down into full back, centre back, but I don't want to spend the time.
What follows is the team name followed by the total open play xG+xA for all players who are listed as playing in those positions for the season so far.
Arsenal
Defenders 13.21
Midfielders 28.06
Attackers 24.84
Bournemouth
Defenders 6.07
Midfielders 22.25
Attackers 18.57
Brighton
Defenders 6.38
Midfielders 13.05
Attackers 11.63
Burnley
Defenders 6.75
Midfielders 13.93
Attackers 17.22
Cardiff
Defenders 15.38 (Callum Paterson classed as defender)
Midfielders 14.14
Attackers 4.51
Chelsea
Defenders 11.91
Midfielders 16.59
Attackers 37.39
Crystal Palace
Defenders 11.2
Midfielders 19.53
Attackers 10.35
Everton
Defenders 9.45
Midfielders 12.76
Attackers 21.46
Fulham
Defenders 6.44
Midfielders 16.13
Attackers 17.71
Huddersfield
Defenders 9.32
Midfielders 10.91
Attackers 8.52
Leicester
Defenders 8
Midfielders 19.61
Attackers 11.19
Liverpool
Defenders 11.47
Midfielders 11.99
Attackers 46.76
Manchester City
Defenders 4.75
Midfielders 35.5
Attackers 45.09
Manchester United
Defenders 6.65
Midfielders 22.69
Attackers 25.8
Newcastle
Defenders 8.25
Midfielders 13.63
Attackers 15.51
Southampton
Defenders 8.79
Midfielders 16.72
Attackers 14.4
Tottenham
Defenders 8.65
Midfielders 23.12
Attackers 18.62
Watford
Defenders 6.04
Midfielders 19.82
Attackers 23.26
West Ham
Defenders 6.69
Midfielders 24.04
Attackers 18.37
Wolves
Defenders 13.2
Midfielders 12.1
Attackers 23.94
So in terms of the percentage of the teams open play expected goals and assists coming from defenders the table looks like this
The issue with Cardiff is that Callum Paterson is classed as a defender, whilst he has been playing as a centre forward. However I'm not going to correct that for two reasons; one it highlights a problem with just using stats without doing analysis, and two, I can't be bothered.
Midfield next
My theory that Everton don't create from the midfield seems to be supported by this, even with Bernard and Sigurdsson classed as midfielders. Liverpool rock bottom of this table.....
Attack
And Liverpool top this table by a long way. Their midfield exists to provide quick balls to the front 4 who do the rest. It seems to work but they are at the extreme end of this dataset.
So what does all this tell us?
Firstly that how we define what a defender/midfielder/attacker is in a database has a big impact. Players move around, Callum Paterson ruined my spreadsheet. But equally it is very tricky to see Sigurdsson as a midfielder when he is most often the furthest forward Everton player on an average position map.
Formation plays an issue too, the 4-5-1 will likely have far more midfield involvement than a 4-3-3 and thus far more xG and xA involvement.
Also perhaps the fact that the lower half of the league creates a lower percentage through their attackers, and a higher percentage through their defenders is because they tend to flood the box for set pieces (and I'm not sure how an open play xG + xA model classifies headers from set pieces or second balls after set pieces).
Tottenham seem further down the attacking list then I would assume. Dele Alli is classed as a midfielder however, but he is only on about 4.5 open play goals+assists in my dataset. Kane and Son are on lower figures than I would expect. Maybe they are just overperforming or scoring more from deadballs?
I was surprised at how low the involvement of the defenders in open expected goals and assists was at Manchester City though. But it probably just is the case that the more functioning your midfield and attack is the less you require attacking contributions from defenders.
All in all probably quite interesting. You could probably turn it into a lovely graphic breaking down each team how they attack. Feel free to do so if you wish!
Comments
Post a Comment