Sorry, I should clarify: I read your excellent post on the components of transfer value and found it very interesting and insightful. In the example (Julian Alvarez) and in the above post on Rogers, the only on-field outputs I noticed was goals/90 and minutes played. For an AM, I would think that there would be more to that which would determine their value.
But perhaps that's the interesting tidbit: the model is probably quite good at predicting what someone is likely to pay in transfer fees, but to assess whether someone has "overpaid", you would need to figure out how likely that player is to contribute to wins/improved play, and teams are still overall not very good at figuring that out before shelling out a lot of money for someone.
Example: I would bet that the transfer fee which Man United paid for Bryan Mbeumo and/or Cunha were reasonably within expectation, but that is a very different discussion to "how much on-field value is this player likely to generate?" which is how people tend to THINK about the question of over-paying.
if we use a mix of expected goals, expected assists and chance creation (all per 90, with goals being weighted higher and chance creation watered down a bit) and filter for players who logged more than 2000 PL mins last season, Morgan Rogers comes out as around the 30th-best attacker, right on par with Brendan Aaronson and John McGinn. Obviously this may leave out other valuable traits such as defensive actions and/or prog. carries/passes, but it's a better predictor of "how much will this player help the team win next season" than just goals/90.
Anyway, to me it's a fun to try to get at the distinction of "did this club overpay based on expected transfer fees" vs. "did this club overpay based on expected output." The two aren't (always) in agreement!
What data are the basis for both the baseline AM output and the productivity above average?
Sorry, I should clarify: I read your excellent post on the components of transfer value and found it very interesting and insightful. In the example (Julian Alvarez) and in the above post on Rogers, the only on-field outputs I noticed was goals/90 and minutes played. For an AM, I would think that there would be more to that which would determine their value.
But perhaps that's the interesting tidbit: the model is probably quite good at predicting what someone is likely to pay in transfer fees, but to assess whether someone has "overpaid", you would need to figure out how likely that player is to contribute to wins/improved play, and teams are still overall not very good at figuring that out before shelling out a lot of money for someone.
Example: I would bet that the transfer fee which Man United paid for Bryan Mbeumo and/or Cunha were reasonably within expectation, but that is a very different discussion to "how much on-field value is this player likely to generate?" which is how people tend to THINK about the question of over-paying.
if we use a mix of expected goals, expected assists and chance creation (all per 90, with goals being weighted higher and chance creation watered down a bit) and filter for players who logged more than 2000 PL mins last season, Morgan Rogers comes out as around the 30th-best attacker, right on par with Brendan Aaronson and John McGinn. Obviously this may leave out other valuable traits such as defensive actions and/or prog. carries/passes, but it's a better predictor of "how much will this player help the team win next season" than just goals/90.
Anyway, to me it's a fun to try to get at the distinction of "did this club overpay based on expected transfer fees" vs. "did this club overpay based on expected output." The two aren't (always) in agreement!