While this graphic is awesome, it obviously carries a bias related to what teams a player has played on, and what league they have played in. A forward playing alongside good playmakers will score much more frequently than one without them, at the same time a playmaker with a good forward will get much more assists. Also, the play-style in different leagues can often be different, with more goals scored on average in some leagues than others.
I wonder how one could create a statistic that takes those things into account, by somehow “normalizing” the number of goals/assists a player has based on who they played with and what league they played in.
Also, I think it’s really interesting how that graph “curves up”. It looks like there’s a kind of trend where players getting a few more assists correlates to them scoring a lot more goals, and while almost noone has more than .35 assists per 90, there are quire a few that have more than .4 goals per 90.
We’re over a decade into the era of advanced statistics in football. While we haven’t cracked it completely yet the way baseball has, great strides have been made in data analytics. You’re absolutely right: simple measurements of goals/assists are mostly meaningless and should not be seen as indicative of performances or abilities.
I wouldn’t go as far as calling it “mostly meaningless”, but it definitely carries an enhancement bias, in that the better players will tend to be placed on the same team, where they mutually enhance each others goal/assist stats.
From a recruiting perspective for instance, I would assume that it would be interesting to devise a statistic to indicate how much a player improves the team they’re on, while somehow factoring out the effects of the other players.
At the same time of course, a lot (most?) of what makes a good team is not just the skills of the individual players, but how the different players utilise each other’s strengths and cover each other’s weaknesses.
I wouldn’t go as far as calling it “mostly meaningless”
Well, to start off goals and assists only even begin to measure one type of player - attackers. They do not represent the impact someone like Rodri or Toni Kroos has.
Even then, goals scored and assists created have proved a poor statistic in terms of predictive power - hence why so called “advanced statistics” began being developed. Expected Goals (or xG) was an early development attempting to measure the quality of chances created, for example. The field is much more advanced at this point however, with xG expanded into concepts like Expected Threat as early as 2018[1] and Possession Value models [2]
From a recruiting perspective for instance, I would assume that it would be interesting to devise a statistic to indicate how much a player improves the team they’re on, while somehow factoring out the effects of the other players.
Attempts to translate statistics such as “X Above Replacement” has been made, see this article by Dan Altman for instance [3]
While this graphic is awesome, it obviously carries a bias related to what teams a player has played on, and what league they have played in. A forward playing alongside good playmakers will score much more frequently than one without them, at the same time a playmaker with a good forward will get much more assists. Also, the play-style in different leagues can often be different, with more goals scored on average in some leagues than others.
I wonder how one could create a statistic that takes those things into account, by somehow “normalizing” the number of goals/assists a player has based on who they played with and what league they played in.
Also, I think it’s really interesting how that graph “curves up”. It looks like there’s a kind of trend where players getting a few more assists correlates to them scoring a lot more goals, and while almost noone has more than .35 assists per 90, there are quire a few that have more than .4 goals per 90.
We’re over a decade into the era of advanced statistics in football. While we haven’t cracked it completely yet the way baseball has, great strides have been made in data analytics. You’re absolutely right: simple measurements of goals/assists are mostly meaningless and should not be seen as indicative of performances or abilities.
I wouldn’t go as far as calling it “mostly meaningless”, but it definitely carries an enhancement bias, in that the better players will tend to be placed on the same team, where they mutually enhance each others goal/assist stats.
From a recruiting perspective for instance, I would assume that it would be interesting to devise a statistic to indicate how much a player improves the team they’re on, while somehow factoring out the effects of the other players.
At the same time of course, a lot (most?) of what makes a good team is not just the skills of the individual players, but how the different players utilise each other’s strengths and cover each other’s weaknesses.
Well, to start off goals and assists only even begin to measure one type of player - attackers. They do not represent the impact someone like Rodri or Toni Kroos has.
Even then, goals scored and assists created have proved a poor statistic in terms of predictive power - hence why so called “advanced statistics” began being developed. Expected Goals (or xG) was an early development attempting to measure the quality of chances created, for example. The field is much more advanced at this point however, with xG expanded into concepts like Expected Threat as early as 2018[1] and Possession Value models [2]
Attempts to translate statistics such as “X Above Replacement” has been made, see this article by Dan Altman for instance [3]
1 ↩︎
2 ↩︎
3 ↩︎