RSS Player of the Season 2008-09
milivella
these position:
- Goalkeeper
- Defender
- Midfielder
- Forward
- Overall
Rules are pretty simple:
- You can vote more than a player (e.g. three best goalkeepers, or two
best player overall), and each of them will in any case have one
point. [1]
- Deadline: July 1, 23:59 GMT.
[1] This can be useful to vote "against" a player that you don't like
without having to discard your real preference. E.g. you think that
Mascara is the best player overall, but you know that he's not in the
set of the likely winners, that is instead composed by Messi and C.
Ronaldo. Among these, you think that Messi is totally overrated. In
this poll you can vote for *both* Mascara (showing your preference for
him) and C. Ronaldo (minimizing the chances that Messi will win). See
http://en.wikipedia.org/wiki/Approval_voting
---
In the past editions, we have awarded Kaka and C. Ronaldo, months
before France Football, FIFA, World Soccer, FIFPro, etc. did:
2006-07
Goalkeeper: Cech
Defenders: Daniel Alves, Vidic
Midfileder: Kaka
Forward: Van Nistelrooy
Overall: Kaka
Poll: http://groups.google.com/group/rec.sport.soccer/browse_thread/thread/8722127baa73f545/
Results: http://groups.google.com/group/rec.sport.soccer/browse_thread/thread/3cfbc5463863c72c/
2007-08
Goalkeeper: Casillas
Defenders: Vidic
Midfileder: Ribery
Forward: Cristiano Ronaldo
Overall: Cristiano Ronaldo
Poll: http://groups.google.com/group/rec.sport.soccer/browse_thread/thread/676a1cc8c67282eb/
Results: http://groups.google.com/group/rec.sport.soccer/browse_thread/thread/f59f53a027eef11a/
--
Cheers
milivella
milivella
> Vote who is, in your opinion, the best player of the season in each of
> these position:
Here I go.
> - Goalkeeper
Julio Cesar
> - Defender
Daniel Alves, Ferdinand, Maicon, Puyol
> - Midfielder
Iniesta
> - Forward
Rooney
> - Overall
Iniesta
--
Cheers
milivella
MH
milivella wrote:
> Vote who is, in your opinion, the best player of the season in each of
> these position:
> - Goalkeeper
Robert Enke
> - Defender
hard... equal votes to Jagielka, Lahm, and Rio Ferdinand
> - Midfielder
Xavi
> - Forward
Messi
> - Overall
>
Xavi
Google Beta User
- Defender: FERDINAND
- Midfielder: XAVI, GERRARD
- Forward: <no vote>
- Overall: GERRARD
Alkamista
> milivella wrote:
> > Vote who is, in your opinion, the best player of the season in each of
> > these position:
> > - Goalkeeper
>
> Robert Enke
>
> > - Defender
>
> hard... equal votes to Jagielka, Lahm, and Rio Ferdinand
You think Jagielka was one of the top 3 defenders in Europe last year?
Enzo
> - Goalkeeper
Victor Valdez
> - Defender
Alves
> - Midfielder
Xavi
> - Forward
Forlán
> - Overall
Messi
Jussi Uosukainen
> Vote who is, in your opinion, the best player of the season in each of
> these position:
> - Goalkeeper
Casillas
> - Defender
Ferdinand, Puyol
> - Midfielder
Xavi, Iniesta, Gerrard
> - Forward
Villa
> - Overall
>
Gerrard
--
/jussi
It has always been the prerogative of children and half-wits to point
out that the emperor has no clothes. But the half-wit remains a
half-wit, and the emperor remains an emperor.
* The Sandman
chris m
<8d96500e-738b-410f...@j32g2000yqh.googlegroups.com>,
milivella <mili...@gmail.com> wrote:
> Vote who is, in your opinion, the best player of the season in each of
> these position:
> - Goalkeeper-Tim Howard (no kidding
> - Defender
> - Midfielder
> - Forward
> - Overall
Goal keeper-Tim Howard (no kidding, Everton in Europe, USA, he's been
getting shelled all year and still managed to win. I'd like to see how
Casillas and the rest would do with the defenses he's had to work with)
Defender-Vidic (he scores goals too, a force in the opponent's box)
Midfielder-Xavi he almost makes it look to easy
Forward- Cristiano Ronaldo
Overall-Cristiano Ronaldo- even with the injury and attitude and very
poor taste in used up blondes he is still the one player in world I
would want to start a team with.
chrism
Jesus Petry
> Vote who is, in your opinion, the best player of the season in each of
> these position:
> - Goalkeeper
Júlio César
> - Defender
Daniel Alves, Rio Ferdinand
> - Midfielder
JS Verón, Andrés Iniesta, Xavi, Kaká
> - Forward
Nilmar, Luís Fabiano, Messi, Rooney
> - Overall
Luís Fabiano
Tchau!
Jesus Petry
Mark V.
> Vote who is, in your opinion, the best player of the season in each of
> these position:
> - Goalkeeper
> - Defender
> - Midfielder
> - Forward
> - Overall
>
> Rules are pretty simple:
> - You can vote more than a player (e.g. three best goalkeepers, or two
> best player overall), and each of them will in any case have one
> point. [1]
> - Deadline: July 1, 23:59 GMT.
>
> [1] This can be useful to vote "against" a player that you don't like
> without having to discard your real preference. E.g. you think that
> Mascara is the best player overall, but you know that he's not in the
> set of the likely winners, that is instead composed by Messi and C.
> Ronaldo. Among these, you think that Messi is totally overrated. In
> this poll you can vote for *both* Mascara (showing your preference for
>
> ---
>
> In the past editions, we have awarded Kaka and C. Ronaldo, months
> before France Football, FIFA, World Soccer, FIFPro, etc. did:
>
> 2006-07
>
> Goalkeeper: Cech
> Defenders: Daniel Alves, Vidic
> Midfileder: Kaka
> Forward: Van Nistelrooy
> Overall: Kaka
>
> Results:http://groups.google.com/group/rec.sport.soccer/browse_thread/thread/...
>
> 2007-08
>
> Goalkeeper: Casillas
> Defenders: Vidic
> Midfileder: Ribery
> Forward: Cristiano Ronaldo
> Overall: Cristiano Ronaldo
>
GK: Buffon
Def: Evra, Ferdinand,
MF: Xavi, Gareth Barry
AM: Kaka
F: C. Ronaldo, Eto'o
All-around: Messi
MH
I don't know. He was very good. Vidic is normally better but produced
some big match goofs. Nobody in Italy impressed me that much, ditto
Spain where Puyol is losing it, and Pique has a way to go yet.
milivella
> GK: Buffon
> Def: Evra, Ferdinand,
> MF: Xavi, Gareth Barry
> AM: Kaka
> F: C. Ronaldo, Eto'o
> All-around: Messi
While voting for a player just for the "overall" award (and not for a
"positional" award) is a legitimate move, I just want to be clear: you
(but the same applies to Enzo) are not giving any point to Messi as a
FW (or MF).
(Of course you can change your votes in any moment.)
--
Cheers
milivella
kolexy
Puyol
Xavi
Kaka
Messi
Gabbage
> Mark V.:
>
> > GK: Buffon
> > Def: Evra, Ferdinand,
> > MF: Xavi, Gareth Barry
> > AM: Kaka
> > F: C. Ronaldo, Eto'o
> > All-around: Messi
>
> While voting for a player just for the "overall" award (and not for a
> "positional" award) is a legitimate move, I just want to be clear: you
> (but the same applies to Enzo) are not giving any point to Messi as a
> FW (or MF).
What do you do when a rebel like Mark nominates a player in the AM
category?
Jordi
Obviously biased towards the Liga and Barcelona.
GK: Casillas
Defender: Alves, Maicon
Mid: Iniesta, Xavi
Forward: Messi
All-around: Iniesta.
milivella
:)
I've considered it as a vote for midfielder. I'm open to suggestions,
though (1 vote both as MF and as FW?).
--
Cheers
milivella
Mark V.
"The line between rebelliousness and carelessness is an obfuscated one
indeed." -- Jack Ruby
Here is my revised list:
GK: Buffon
Def: Evra, Ferdinand,
MF: Xavi, Gareth Barry, Kaka
F: C. Ronaldo, Eto'o, Torres, Messi
All-around: Messi
milivella
> "The line between rebelliousness and carelessness is an obfuscated one
> indeed." -- Jack Ruby
:)
> Here is my revised list:
OK, casted.
--
milivella
Google Beta User
> Vidic is normally better but produced some big match goofs.
Correct.
Plus he was never the same after his encounter with you know who.
I don't have a problem with Jaglieka (or Lescott). Both played well.
Futbolmetrix
news:8d96500e-738b-410f...@j32g2000yqh.googlegroups.com...
> Vote who is, in your opinion, the best player of the season in each of
> these position:
> - Goalkeeper
Julio Cesar
> - Defender
Maicon, Dani Alves. Defenders that are good at defending are overrated.
> - Midfielder
Xavi, Iniesta. can't distinguish between the two generic midgets anyway.
> - Forward
Messi, Cristiano Ronaldo, Ibrahimovic
> - Overall
Messi
milivella
> > - Defender
>
> Maicon, Dani Alves. Defenders that are good at defending are overrated.
I knew that reading Futbolmetrix = smiling!
Anyway, you're not really Italian, mate. I voted Rooney because he's a
forward that is good at defending!
--
Cheers
milivella
milivella
> > - Defender
>
> Maicon, Dani Alves. Defenders that are good at defending are overrated.
Anyway, here are the top 10 "universal forwards", as defined by Elko's
formula (geometric mean of goals and assists), in some 2008-09
competitions. All data from
http://www.transfermarkt.co.uk/en/
(hence the English-Deutsch mix)
UEFA Champions League
Lionel Messi FC Barcelona Forward 6.71
Franck Ribéry FC Bayern München Midfield 5.29
Miroslav Klose FC Bayern München Forward 5.29
Frank Lampard FC Chelsea Midfield 4.58
Juninho Olympique Lyon Midfield 3.87
Thierry Henry FC Barcelona Forward 3.87
Robin van Persie FC Arsenal Forward 3.87
Xavi FC Barcelona Midfield 3.74
Lisandro López FC Porto Forward 3.46
Cristiano Ronaldo Manchester United Midfield 3.46
UEFA Cup
Mladen Petric Hamburger SV Forward 4.47
Ivica Olic Hamburger SV Forward 4.24
Fabio Quagliarella Udinese Calcio Forward 4.00
Dimitri Payet AS Saint-Etienne Forward 3.87
Luis Suárez Ajax Amsterdam Forward 3.87
Stephen Ireland Manchester City Midfield 3.46
Jadson Schachtjor Donezk Midfield 3.46
Piotr Trochowski Hamburger SV Midfield 3.46
Vágner Love ZSKA Moskau Forward 3.32
Diogo Olympiakos Piräus Forward 3.16
English Premier League
Steven Gerrard Liverpool FC Midfield 13.86
Nicolás Anelka Chelsea FC Forward 11.53
Frank Lampard Chelsea FC Midfield 10.95
Gabriel Agbonlahor Aston Villa Forward 10.49
Robin van Persie Arsenal FC Forward 10.49
Cristiano Ronaldo Manchester United Midfield 10.39
Dirk Kuyt Liverpool FC Forward 10.39
Dimitar Berbatov Manchester United Forward 9.49
Wayne Rooney Manchester United Forward 9.17
Stephen Ireland Manchester City Midfield 9.00
Liga
Diego Forlán Atletico Madrid Forward 17.89
Lionel Messi FC Barcelona Forward 15.91
Gonzalo Higuaín Real Madrid CF Forward 15.56
Kun Agüero Atletico Madrid Forward 14.28
Juan Manuel Mata FC Valencia Forward 13.67
David Villa FC Valencia Forward 12.96
Samuel Eto'o FC Barcelona Forward 12.25
Raúl Real Madrid CF Forward 12.00
Thierry Henry FC Barcelona Forward 11.53
Xavi FC Barcelona Midfield 11.49
Serie A
Zlatan Ibrahimovic Inter Mailand Forward 14.14
Diego Milito FC Genua 1893 Forward 13.86
Antonio Cassano UC Sampdoria Forward 13.42
Kaká AC Milan Midfield 13.27
Alessandro Del Piero Juventus Turin Forward 12.49
Fabrizio Miccoli US Palermo Forward 11.83
Alberto Gilardino AC Florenz Forward 10.68
Gaetano D'Agostino Udinese Calcio Midfield 9.95
Marco Di Vaio FC Bologna Forward 9.80
Mauro Matías Zárate Lazio Rom Forward 9.54
Bundesliga
Grafite VfL Wolfsburg Forward 17.55
Edin Dzeko VfL Wolfsburg Forward 16.12
Mario Gomez VfB Stuttgart Forward 12.96
Zvjezdan Misimovic VfL Wolfsburg Midfield 11.83
Vedad Ibisevic TSG 1899 Hoffenheim Forward 11.22
Stefan Kießling Bayer 04 Leverkusen Forward 10.39
Alexander Frei Borussia Dortmund Forward 10.39
Patrick Helmes Bayer 04 Leverkusen Forward 10.25
Franck Ribéry Bayern Munich Midfield 9.95
Luca Toni Bayern Munich Forward 9.90
Ligue 1
Michel Bastos OSC Lille Midfield 13.49
Yoann Gourcuff Girondins Bordeaux Midfield 11.49
Mamadou Niang Olympique Marseille Forward 10.58
Kevin Gameiro FC Lorient Forward 10.49
Guillaume Hoarau FC Paris St. Germain Forward 10.10
André-Pierre Gignac FC Toulouse Forward 9.80
Marouane Chamakh Girondins Bordeaux. Forward 8.83
Mevlüt Erdinç FC Sochaux-Montbeliard Forward 8.77
Karim Benzema Olympique Lyon Forward 8.25
Steve Savidan SM Caen Forward 8.06
Copa Libertadores (up to the quarter-finals)
Jorge Daniel Núñez Nacional FC Asunción Midfield 5.29
Cleiton Xavier Sociedade Esportiva Palmeiras Midfield 4.24
Rodrigo Texeira Deportivo Cuenca Forward 3.74
Keirrison Sociedade Esportiva Palmeiras Forward 3.46
Souza Gremio Foot-Ball Porto Alegre Midfield 3.16
Rodrigo Sebastián Palacio Club Atlético Boca Juniors Forward 3.16
Martín Palermo Club Atlético Boca Juniors Forward 3.16
Darío Damián Figueroa Caracas FC Midfield 3.16
Ismael Óscar Villalba Deportivo Cuenca Forward 3.00
Lucas Barrios CSD Colo Colo Forward 3.00
Copa Sudamericana (2008)
Nilmar Sport Club Internacional Forward 3.16
Omar Arellano Deportivo Guadalajara Forward 3.00
Marco Jhonfai Fabián Deportivo Guadalajara Midfield 2.83
Carlos Alberto Botafogo de Futebol e Regatas Midfield 2.00
Sebastián Abreu Club Atlético River Plate Forward 1.73
Julio Gutiérrez CD Universidad Católica Forward 1.73
Marcos Alberto Angeleri Estudiantes de la Plata Defense 1.73
Milován Mirosevic CD Universidad Católica Midfield 1.41
Malher Tressor Moreno Club San Luis Midfield 1.41
Carlos Flores Club Sport Ancash Midfield 1.41
Campeonato Brasileiro Série A (2008)
Guilherme Cruzeiro Esporte Clube Forward 11.22
Alex Mineiro Sociedade Esportiva Palmeiras Forward 9.75
Leandro Amaral Club de Regatas Vasco da Gama Forward 9.38
Nilmar Sport Club Internacional Forward 9.17
Lúcio Flávio Botafogo de Futebol e Regatas Midfield 9.00
Alex Sport Club Internacional Midfield 8.94
Borges São Paulo Futebol Clube Forward 8.66
Iarley Goias Esporte Clube Forward 8.49
Edmundo Club de Regatas Vasco da Gama Forward 8.06
Washington Fluminense Football Club Forward 7.94
Argentina (Apertura)
José Gustavo Sand Club Atletico Lanus Forward
7.75
Alejandro Darío Gómez Club Atlético San Lorenzo de Almagro Midfield
6.32
Maxi Moralez Club Atletico Velez Sarsfield Midfield
5.48
Rubén Ramírez Racing Club de Avellaneda Forward
5.48
Lucas Ezequiel Viatri Club Atlético Boca Juniors Forward
5.29
Martín Gerardo Morel Club Atletico Tigre Midfield
5.10
Sebastián Blanco Club Atletico Lanus Midfield
4.90
Gonzalo Bergessio Club Atlético San Lorenzo de Almagro Forward
4.90
Pablo Barrientos Club Atlético San Lorenzo de Almagro Midfield
4.58
Diego Eduardo Lagos Club Atletico Lanus Forward
4.47
Argentina (Clausura. 1 round left)
José Gustavo Sand Club Atletico Lanus Forward 7.42
Matías de Federico Club Atlético Huracán Forward 7.07
Daniel Gastón Montenegro CA Independiente de Avellaneda Forward 6.00
Esteban Óscar Fuertes Club Atletico Colon (Santa Fe) Forward 5.74
Victor Alberto Figueroa Godoy Cruz Midfield 5.48
Leandro Rubén Caruso Godoy Cruz Forward 4.90
Luciano Leguizamón Arsenal de Sarandi FC Forward 4.90
Sebastián Fernández Club Atletico Banfield Forward 4.90
Rodrigo López Club Atletico Velez Sarsfield Forward 4.69
Javier Pastore Club Atlético Huracán Midfield 4.58
Confederations Cup
Kaká Brasilien Midfield 2.00
David Villa Spanien Forward 1.73
Fernando Torres Spanien Forward 1.73
Maicón Brasilien Defense 1.73
Cesc Fàbregas Spanien Midfield 1.41
Robinho Brasilien Forward 1.41
Jozy Altidore Vereinigte Staaten Forward 1.41
Landon Donovan Vereinigte Staaten Forward 1.41
Dani Alves Brasilien Defense 1.00
Charlie Davies Vereinigte Staaten Forward 1.00
--
Cheers
milivella
milivella
> here are the top 10 "universal forwards", as defined by Elko's
> formula (geometric mean of goals and assists), in some 2008-09
> competitions.
And here are the top 18 WC performances 1966-2006 by the same metrics
(I haven't computed the square root, so it's just goals*assists):
30
Mueller 70
25
Maradona 86
20
Pele 70
16
Zico 82
Belanov 86
15
Rensenbrink 78
14
Lato 74
12
Boniek 82
Brolin 94
Hagi 94
Ronaldo 98
Ballack 02
10
Szarmach 74
Littbarski 82
Careca 86
Andersson 94
Romario 94
Rivaldo 02
---
Compare them with the best performances computed taking the minimum
number between goals and assists:
5
Maradona 86
4
Pele 70
Zico 82
Belanov 86
3
Torres 66
Mueller 70
Rivelino 70
Cruijff 74
Rensenbrink 78
Boniek 82
Giresse 82
Brehme 90
Brolin 94
Hagi 94
Bergkamp 98
Ronaldo 98
Ballack 02
--
Cheers
milivella
El Kot
across different competitions. The best would be to do it per minutes
played, or at least per number of games. That would allow comparison
across leagues. Maybe normalized by the total scoring of the league too?
Just giving ideas here. :)
milivella wrote:
>
> Anyway, here are the top 10 "universal forwards", as defined by Elko's
> formula (geometric mean of goals and assists), in some 2008-09
> competitions. All data from
> http://www.transfermarkt.co.uk/en/
> (hence the English-Deutsch mix)
>
> UEFA Champions League
>
> Lionel Messi FC Barcelona Forward 6.71
> Franck Rib�ry FC Bayern M�nchen Midfield 5.29
> Miroslav Klose FC Bayern M�nchen Forward 5.29
> Frank Lampard FC Chelsea Midfield 4.58
> Juninho Olympique Lyon Midfield 3.87
> Thierry Henry FC Barcelona Forward 3.87
> Robin van Persie FC Arsenal Forward 3.87
> Xavi FC Barcelona Midfield 3.74
> Lisandro L�pez FC Porto Forward 3.46
> Cristiano Ronaldo Manchester United Midfield 3.46
>
> UEFA Cup
>
> Mladen Petric Hamburger SV Forward 4.47
> Ivica Olic Hamburger SV Forward 4.24
> Fabio Quagliarella Udinese Calcio Forward 4.00
> Dimitri Payet AS Saint-Etienne Forward 3.87
> Luis Su�rez Ajax Amsterdam Forward 3.87
> Stephen Ireland Manchester City Midfield 3.46
> Jadson Schachtjor Donezk Midfield 3.46
> Piotr Trochowski Hamburger SV Midfield 3.46
> V�gner Love ZSKA Moskau Forward 3.32
> Diogo Olympiakos Pir�us Forward 3.16
>
> English Premier League
>
> Steven Gerrard Liverpool FC Midfield 13.86
> Nicol�s Anelka Chelsea FC Forward 11.53
> Frank Lampard Chelsea FC Midfield 10.95
> Gabriel Agbonlahor Aston Villa Forward 10.49
> Robin van Persie Arsenal FC Forward 10.49
> Cristiano Ronaldo Manchester United Midfield 10.39
> Dirk Kuyt Liverpool FC Forward 10.39
> Dimitar Berbatov Manchester United Forward 9.49
> Wayne Rooney Manchester United Forward 9.17
> Stephen Ireland Manchester City Midfield 9.00
>
> Liga
>
> Diego Forl�n Atletico Madrid Forward 17.89
> Lionel Messi FC Barcelona Forward 15.91
> Gonzalo Higua�n Real Madrid CF Forward 15.56
> Kun Ag�ero Atletico Madrid Forward 14.28
> Juan Manuel Mata FC Valencia Forward 13.67
> David Villa FC Valencia Forward 12.96
> Samuel Eto'o FC Barcelona Forward 12.25
> Ra�l Real Madrid CF Forward 12.00
> Thierry Henry FC Barcelona Forward 11.53
> Xavi FC Barcelona Midfield 11.49
>
> Serie A
>
> Zlatan Ibrahimovic Inter Mailand Forward 14.14
> Diego Milito FC Genua 1893 Forward 13.86
> Antonio Cassano UC Sampdoria Forward 13.42
> Kak� AC Milan Midfield 13.27
> Alessandro Del Piero Juventus Turin Forward 12.49
> Fabrizio Miccoli US Palermo Forward 11.83
> Alberto Gilardino AC Florenz Forward 10.68
> Gaetano D'Agostino Udinese Calcio Midfield 9.95
> Marco Di Vaio FC Bologna Forward 9.80
> Mauro Mat�as Z�rate Lazio Rom Forward 9.54
>
> Bundesliga
>
> Grafite VfL Wolfsburg Forward 17.55
> Edin Dzeko VfL Wolfsburg Forward 16.12
> Mario Gomez VfB Stuttgart Forward 12.96
> Zvjezdan Misimovic VfL Wolfsburg Midfield 11.83
> Vedad Ibisevic TSG 1899 Hoffenheim Forward 11.22
> Stefan Kie�ling Bayer 04 Leverkusen Forward 10.39
> Alexander Frei Borussia Dortmund Forward 10.39
> Patrick Helmes Bayer 04 Leverkusen Forward 10.25
> Franck Rib�ry Bayern Munich Midfield 9.95
> Luca Toni Bayern Munich Forward 9.90
>
> Ligue 1
>
> Michel Bastos OSC Lille Midfield 13.49
> Yoann Gourcuff Girondins Bordeaux Midfield 11.49
> Mamadou Niang Olympique Marseille Forward 10.58
> Kevin Gameiro FC Lorient Forward 10.49
> Guillaume Hoarau FC Paris St. Germain Forward 10.10
> Andr�-Pierre Gignac FC Toulouse Forward 9.80
> Marouane Chamakh Girondins Bordeaux. Forward 8.83
> Mevl�t Erdin� FC Sochaux-Montbeliard Forward 8.77
> Karim Benzema Olympique Lyon Forward 8.25
> Steve Savidan SM Caen Forward 8.06
>
> Copa Libertadores (up to the quarter-finals)
>
> Jorge Daniel N��ez Nacional FC Asunci�n Midfield 5.29
> Cleiton Xavier Sociedade Esportiva Palmeiras Midfield 4.24
> Rodrigo Texeira Deportivo Cuenca Forward 3.74
> Keirrison Sociedade Esportiva Palmeiras Forward 3.46
> Souza Gremio Foot-Ball Porto Alegre Midfield 3.16
> Rodrigo Sebasti�n Palacio Club Atl�tico Boca Juniors Forward 3.16
> Mart�n Palermo Club Atl�tico Boca Juniors Forward 3.16
> Dar�o Dami�n Figueroa Caracas FC Midfield 3.16
> Ismael �scar Villalba Deportivo Cuenca Forward 3.00
> Lucas Barrios CSD Colo Colo Forward 3.00
>
> Copa Sudamericana (2008)
>
> Nilmar Sport Club Internacional Forward 3.16
> Omar Arellano Deportivo Guadalajara Forward 3.00
> Marco Jhonfai Fabi�n Deportivo Guadalajara Midfield 2.83
> Carlos Alberto Botafogo de Futebol e Regatas Midfield 2.00
> Sebasti�n Abreu Club Atl�tico River Plate Forward 1.73
> Julio Guti�rrez CD Universidad Cat�lica Forward 1.73
> Marcos Alberto Angeleri Estudiantes de la Plata Defense 1.73
> Milov�n Mirosevic CD Universidad Cat�lica Midfield 1.41
> Malher Tressor Moreno Club San Luis Midfield 1.41
> Carlos Flores Club Sport Ancash Midfield 1.41
>
> Campeonato Brasileiro S�rie A (2008)
>
> Guilherme Cruzeiro Esporte Clube Forward 11.22
> Alex Mineiro Sociedade Esportiva Palmeiras Forward 9.75
> Leandro Amaral Club de Regatas Vasco da Gama Forward 9.38
> Nilmar Sport Club Internacional Forward 9.17
> L�cio Fl�vio Botafogo de Futebol e Regatas Midfield 9.00
> Alex Sport Club Internacional Midfield 8.94
> Borges S�o Paulo Futebol Clube Forward 8.66
> Iarley Goias Esporte Clube Forward 8.49
> Edmundo Club de Regatas Vasco da Gama Forward 8.06
> Washington Fluminense Football Club Forward 7.94
>
> Argentina (Apertura)
>
> Jos� Gustavo Sand Club Atletico Lanus Forward
> 7.75
> Alejandro Dar�o G�mez Club Atl�tico San Lorenzo de Almagro Midfield
> 6.32
> Maxi Moralez Club Atletico Velez Sarsfield Midfield
> 5.48
> Rub�n Ram�rez Racing Club de Avellaneda Forward
> 5.48
> Lucas Ezequiel Viatri Club Atl�tico Boca Juniors Forward
> 5.29
> Mart�n Gerardo Morel Club Atletico Tigre Midfield
> 5.10
> Sebasti�n Blanco Club Atletico Lanus Midfield
> 4.90
> Gonzalo Bergessio Club Atl�tico San Lorenzo de Almagro Forward
> 4.90
> Pablo Barrientos Club Atl�tico San Lorenzo de Almagro Midfield
> 4.58
> Diego Eduardo Lagos Club Atletico Lanus Forward
> 4.47
>
> Argentina (Clausura. 1 round left)
>
> Jos� Gustavo Sand Club Atletico Lanus Forward 7.42
> Mat�as de Federico Club Atl�tico Hurac�n Forward 7.07
> Daniel Gast�n Montenegro CA Independiente de Avellaneda Forward 6.00
> Esteban �scar Fuertes Club Atletico Colon (Santa Fe) Forward 5.74
> Victor Alberto Figueroa Godoy Cruz Midfield 5.48
> Leandro Rub�n Caruso Godoy Cruz Forward 4.90
> Luciano Leguizam�n Arsenal de Sarandi FC Forward 4.90
> Sebasti�n Fern�ndez Club Atletico Banfield Forward 4.90
> Rodrigo L�pez Club Atletico Velez Sarsfield Forward 4.69
> Javier Pastore Club Atl�tico Hurac�n Midfield 4.58
>
> Confederations Cup
>
> Kak� Brasilien Midfield 2.00
> David Villa Spanien Forward 1.73
> Fernando Torres Spanien Forward 1.73
> Maic�n Brasilien Defense 1.73
> Cesc F�bregas Spanien Midfield 1.41
> Robinho Brasilien Forward 1.41
> Jozy Altidore Vereinigte Staaten Forward 1.41
> Landon Donovan Vereinigte Staaten Forward 1.41
> Dani Alves Brasilien Defense 1.00
> Charlie Davies Vereinigte Staaten Forward 1.00
--
No, no, you can't e-mail me with the nono.
milivella
> it is difficult to compare them
> across different competitions. The best would be to do it per minutes
> played, or at least per number of games. That would allow comparison
> across leagues. Maybe normalized by the total scoring of the league too?
> Just giving ideas here. :)
Given that transfermarkt put minutes (and games) in the same table as
goals and assists, dividing by minutes would have needed no effort...
if I had thought to do it. :( And now I haven't time to copy-paste it
again. :( I am curious, though, 'cause different competitions have
different levels of difficulty, so I think that the ranking are only
good to compare players inside a given competition (Messi has been the
best player in UCL, Gerrard in EPL, Ibrahimovic in Serie A...).
--
Cheers
milivella
El Kot
> El Kot:
>
>> it is difficult to compare them
>> across different competitions. The best would be to do it per minutes
>> played, or at least per number of games. That would allow comparison
>> across leagues. Maybe normalized by the total scoring of the league too?
>> Just giving ideas here. :)
>
> Given that transfermarkt put minutes (and games) in the same table as
> goals and assists, dividing by minutes would have needed no effort...
> if I had thought to do it. :( And now I haven't time to copy-paste it
> again. :(
Too bad. :( I usually save all data for such things, assuming I
copied it to begin with. Like I did with the result grids for the energy
rankings of the leagues (do you have any opinion on the Italian ones, BTW?).
Another thing - I took a look at the transfermarkt site, but it was
not immediately intuitively clear how to get to that data. Which
function did you use - player search? something else?
> I am curious, though, 'cause different competitions have
> different levels of difficulty, so I think that the ranking are only
> good to compare players inside a given competition (Messi has been the
> best player in UCL, Gerrard in EPL, Ibrahimovic in Serie A...).
Well, that's exactly the reason I mentioned normalization by total
goals scored in the competition. That way, we can get a measure of the
proportion of the total goals that the player in question was involved
in. And then maybe further normalize by number of players? or by total
goals scored by his team? I'm getting too much into normalization, I
guess. :) Probably just by minutes played (to measure efficiency) and
by total goals scored in the league should be enough.
Benny
> From : mili...@gmail.com
> Vote who is, in your opinion, the best player of the season in each of
> these position:
> - Goalkeeper
Julio Cesar.
> - Defender
Rio Ferdinand.
> - Midfielder
Xavi.
> - Forward
Messi.
> - Overall
Messi because Xavi didn't have a good Confederations Cup.
--
http://soccer-europe.com
Rss feed : http://soccer-europe.com/RSS/News.xml
milivella
> I usually save all data for such things
You're right. I started to do it without even thinking, then I thought
that it could have been interesting to post the rankings, but I hadn't
saved the first leagues, so I went on without saving...
> Like I did with the result grids for the energy
> rankings of the leagues (do you have any opinion on the Italian ones, BTW?).
Genoa above Fiorentina in all the rankings: interesting. The squashing
ranking has the minimum distance Palermo - Catania (10th vs. 13th
instead than 8th vs. 15th), so it's the best one. ;) Really, I'm not
sure that I understand the "meaning" of the 3 rankings: could you
please give an intuitive explanation?
> Another thing - I took a look at the transfermarkt site, but it was
> not immediately intuitively clear how to get to that data. Which
> function did you use - player search? something else?
Go to a competition, and then to "Club statistics" (odd, I know). E.g.
http://www.transfermarkt.co.uk/en/wettbewerb/GB1/premier-league/datenfakten/torschuetzen.html
> > I am curious, though, 'cause different competitions have
> > different levels of difficulty, so I think that the ranking are only
> > good to compare players inside a given competition (Messi has been the
> > best player in UCL, Gerrard in EPL, Ibrahimovic in Serie A...).
>
> Well, that's exactly the reason I mentioned normalization by total
> goals scored in the competition. That way, we can get a measure of the
> proportion of the total goals that the player in question was involved
> in.
It's banal but let me say: of course we should take in account that
some competitions have more teams than others.
--
Cheers
milivella
Abubakr
> > - Overall
>
> Messi
So much for statistical validations.
Abubakr
> Vote who is, in your opinion, the best player of the season in each of
> these position:
> - Goalkeeper
> - Midfielder
> - Forward
> - Overall
Goalkeeper: Julio Cesar
Defender: Pique,
Midfielder: Iniesta, Xavi, Melo
Forward: Messi
Overall: Messi, Iniesta
FF
Defender: Messi
Midfielder: Messi
Forward: Messi
Overall: Messi
;-)
Seriously now, FWIW I nominate Messi for midfielder and over-all.
Elko Tchernev
> El Kot:
>
>> Like I did with the result grids for the energy
>> rankings of the leagues (do you have any opinion on the Italian ones, BTW?).
>
> Genoa above Fiorentina in all the rankings: interesting. The squashing
> ranking has the minimum distance Palermo - Catania (10th vs. 13th
> instead than 8th vs. 15th), so it's the best one. ;) Really, I'm not
> sure that I understand the "meaning" of the 3 rankings: could you
> please give an intuitive explanation?
OK, I'll try a simple subset of the Spanish liga, with results
slightly modified for illustrative purposes. Suppose the matches between
Barca, Real, Mallorca and Recreativo had ended with these scores:
Barca Mallo Real Racing
Barca 3-1 2-0 1-1
Mallo 2-1 0-2 1-0
Real 2-6 1-1 1-0
Racing 1-2 1-2 0-2
This would produce a force grid between them (positive value meaning the
team has to be higher than the other in the ranking; the magnitude is
the distance at which the force becomes 0):
Barca Mallo Real Racing
Barca 0.0 0.1 2.6 1.1
Mallo -0.1 0.0 -1.2 2.2
Real -2.6 1.2 0.0 2.3
Racing -1.1 -2.2 -2.3 0.0
In the linear energy case, these force vectors are just added to each
other; for each team, it is a linear superposition of forces. The sums
become:
Initial forces acting on each team point, when the coordinates of each
are at 0:
1 Barca 3.8
2 Mallo 0.9
Real 0.9
4 Racing -5.6
Not surprisingly, since everything is linear, the final positions of the
teams after energy minimization, preserve the ordering from the initial
forces:
Positions where the forces acting on each team are 0:
1 Barca 0.190
2 Mallo 0.045
Real 0.045
4 Racing -0.280
This corresponds to the traditional ranking performed in leagues (with 2
points per win), with total points and goal difference from all games.
Both Real and Mallorca would have 3 wins, 2 losses and 1 draw, with the
same goal difference of -1, and would have to be separated according to
some other criteria. It can be easily seen, that the influence of any
change (be it 1 point for a win, or 1/10 of a point for 1 goal
difference, is the same no matter what the actual magnitude of the value
is - 0.9, or 3.8).
With the quadratic function, we obtain the following energy minimum
positions:
1 Barca 0.53
2 Mallo 0.17
3 Real 0.06
4 Racing -0.56
The large force exerted by Barca on Real (that force vector is the
largest in the system) has pushed Real down past Mallorca. Real have
advantage over Mallorca, but that vector is more than 4 times weaker
than the one from Barca in the quadratic case, and doesn't help. Not
only that, but since the force between Barca and Mallorca becomes 0 only
at distance 0.1, and the distance here is greater than that, Barca is
actually pulling Mallorca up, while pushing Real down. :)
Similar was the case with the EPL quadratic ranking - Liverpool had
a combined force vector of 8.20 against the other 6 top teams, and
regardless of its overall smaller sum, forced its way to the top.
Finally, with the squashing function, the energy positions become:
1 Barca 0.105
2 Real 0.050
3 Mallo 0.009
4 Racing -0.165
With Barca far away from Real and Mallorca, the force it exerts tends to
saturate at some level, sufficient to keep them both away, but not
sufficient enough to override the advantage Real has over Mallorca
head-to-head.
El Kot
> El Kot:
>> milivella wrote:
>
>>> I am curious, though, 'cause different competitions have
>>> different levels of difficulty, so I think that the ranking are only
>>> good to compare players inside a given competition (Messi has been the
>>> best player in UCL, Gerrard in EPL, Ibrahimovic in Serie A...).
>> Well, that's exactly the reason I mentioned normalization by total
>> goals scored in the competition. That way, we can get a measure of the
>> proportion of the total goals that the player in question was involved
>> in.
>
> It's banal but let me say: of course we should take in account that
> some competitions have more teams than others.
I think this will be implicitly taken care of when we normalize by
playing time. More teams - more matches - more playing time.
Futbolmetrix
> Futbolmetrix:
>
> Anyway, you're not really Italian, mate. I voted Rooney because he's a
> forward that is good at defending!
That's overrated as well!
But seriously, suppose that we could calculate the +/- of each player
(as we've discussed in the past), as they do in basketball. That is,
the extra number of goals that the team scores when player X is on the
field (relative to the "average" player), minus the number of goals
that the team concedes when player X is on the field (relative to the
"average" player).
Well, my hunch is that the variation in goals conceded must be quite a
bit smaller than the variation in goals scored. In other words, the
best defender in the world is maybe worth 0.1 goals per game, while
the worst defenders is worth -0.1 goals per game. But the best
attacker is maybe worth 0.25 goals per game.
Now, in terms of attacking prowess, Maicon is probably in the top 25%
of players in Serie A (4 goals and 5 assists). So his "+" value is
probably around 0.1. Let's assume that he's a slightly below average
defender, so his "-" value is -0.02. His value is therefore 0.08. So,
the best defender in the world would have to be worth at least -0.02
in terms of attacking prowess to be equivalent to Maicon.
Too bad this is only a wild conjecture.
D
Enzo
It is due to combined performances in the Liga and ECL, not the liga
alone
Abubakr
Fair enough point. I have this theory that coaching at junior levels
should be geared towards producing as many forwards/attackers as
possible, in as many shapes and sizes as possible. Once the time comes
to make the grade and step into serious pro/semi stage, the lesser
talented big forwards can be converted into central defenders, the
less talented winger types can be converted into full/wing-backs, the
more hard working less talented ones can be converted into
midfielders, and the rest who made up the numbers and gave them
practice discarded altogether.
This type of development seems to have worked for Nesta, Lucio, Cafu
etc...
Enzo
> Campeonato Brasileiro Série A (2008)
>
> Guilherme Cruzeiro Esporte Clube Forward 11.22
Can my deer Brasilian friends here say when he is going
to reward me with a cap ?
Enzo
> On Jun 28, 7:30 pm, milivella <milive...@gmail.com> wrote:
>
> > Vote who is, in your opinion, the best player of the season in each of
> > these position:
> > - Goalkeeper
>
>
> > - Defender
>
> Daniel Alves, Rio Ferdinand
>
> > - Midfielder
>
> JS Verón, Andrés Iniesta, Xavi, Kaká
>
> > - Forward
>
> Nilmar, Luís Fabiano, Messi, Rooney
>
> > - Overall
>
> Luís Fabiano
Hmm, a very ordinary season for Sevilla.
milivella
> milivella wrote:
> > El Kot:
>
> >> Like I did with the result grids for the energy
> >> rankings of the leagues (do you have any opinion on the Italian ones, BTW?).
>
> > Genoa above Fiorentina in all the rankings: interesting. The squashing
> > ranking has the minimum distance Palermo - Catania (10th vs. 13th
> > instead than 8th vs. 15th), so it's the best one. ;) Really, I'm not
> > sure that I understand the "meaning" of the 3 rankings: could you
> > please give an intuitive explanation?
>
> OK, I'll try a simple subset of the Spanish liga, with results
> slightly modified for illustrative purposes.
Thanks a lot!
Let's see whether I've got at least the basics...
- We have teams as magnets in a up-down oriented line. Each result
(two-legged, points and GD wise) is a force that tries to put the two
teams (the winner up, the loser down) at a distance that is
proportional to that result.
- Linear = force equals result = all the results have the same weight.
- Quadratic = force equals result^2 = results involving largest
differences have more weight.
- Squashing = forces have less effect when the distance between the
two points is more.
What have I got wrong? :)
--
Cheers
milivella
milivella
> suppose that we could calculate the +/- of each player
> (as we've discussed in the past), as they do in basketball. That is,
> the extra number of goals that the team scores when player X is on the
> field (relative to the "average" player), minus the number of goals
> that the team concedes when player X is on the field (relative to the
> "average" player).
>
> Well, my hunch is that the variation in goals conceded must be quite a
> bit smaller than the variation in goals scored. In other words, the
> best defender in the world is maybe worth 0.1 goals per game, while
> the worst defenders is worth -0.1 goals per game. But the best
> attacker is maybe worth 0.25 goals per game.
>
> Now, in terms of attacking prowess, Maicon is probably in the top 25%
> of players in Serie A (4 goals and 5 assists). So his "+" value is
> probably around 0.1. Let's assume that he's a slightly below average
> defender, so his "-" value is -0.02. His value is therefore 0.08. So,
> the best defender in the world would have to be worth at least -0.02
> in terms of attacking prowess to be equivalent to Maicon.
>
> Too bad this is only a wild conjecture.
Yes, too bad. Because it's a very interesting conjecture. What are the
obstacles to prove it? the fact that we have too little data? the fact
that we should take in account the strength of the opponent team? the
number of matches involved?
The only computations (slightly related to this) that I've done in the
past are
http://groups.google.com/group/rec.sport.soccer/browse_thread/thread/892cadc83aa97258/
http://groups.google.com/group/rec.sport.soccer/msg/8b0aebe62248f459
and subsequent posts in the thread
They can't prove anything, but there is something in the direction of
"forwards are more important than defenders" (Garrincha, Zico, Pele,
Cruijff; Kaka, Seedorf...).
--
Cheers
milivella
El Kot
>
> Thanks a lot!
>
> Let's see whether I've got at least the basics...
> - We have teams as magnets in a up-down oriented line. Each result
> (two-legged, points and GD wise) is a force that tries to put the two
> teams (the winner up, the loser down) at a distance that is
> proportional to that result.
> - Linear = force equals result = all the results have the same weight.
> - Quadratic = force equals result^2 = results involving largest
> differences have more weight.
> - Squashing = forces have less effect when the distance between the
> two points is more.
>
> What have I got wrong? :)
I think you got it right. Let me just re-iterate one even simpler
example, to remove any doubt.
Suppose you have only two teams, and team A has 2 wins over team B with
1-0 each; the force vector will be (according to my convention) equal to
2.2.
Which means that:
- in all cases, the distance between them, where they will settle
when the force between them becomes 0, is equal to that same 2.2; A will
be at +1.1, and B will be at -1.1
- in the linear case, the initial force between the teams, when
they are both at 0, will be equal to that same 2.2, and will decrease
linearly with the distance between them, until the distance reaches the
desired 2.2. At half the distance. the force will be half the initial
one - that is, 1.1
- in the quadratic case, the initial force will be equal to the
squared 2.2, that is, to 4.84, and will change proportionally to the
squared distance change. At half the distance, the force will be 1.1
squared - that is, 1.21
- in the squashing case, where the function is the hyperbolic
tangent. the initial force tanh(2.2) will be 0.98, and will decrease
much more slowly, to 0.80 at half the distance of 1.1, and then in the
vicinity of the target distance of 2.2, tanh(~0) is almost linear.
Conversely, when the teams are close to their final positions (or
when the difference between them is small - let's say just one goal
difference), then the quadratic dependency is much smaller - 0.1^2 is
0.01. We might say that the quadratic case is less sensitive to goal
difference, the squashing case is less sensitive to overall wins, and
the linear case is equally sensitive to both.
Jussi Uosukainen
> Vote who is, in your opinion, the best player of the season in each of
> these position:
> - Goalkeeper
> - Midfielder
> - Forward
> - Overall
>
A thought crossed my mind: why don't we make this a bit more
sophisticated? First round of votes determines the 5 candidates for each
position, in the second round we use a rank based voting system (like
http://en.wikipedia.org/wiki/Condorcet_method) to determine the actual
winner.
Could someone bother to set this up? I know that a software is probably
needed, but with only 5 positions and 5 candidates, it probably isn't
too much of a hassle. I'll excuse myself with very limited internet
access in the coming weeks (summer holiday), but if anyone would be
willing to set this up, I would be very happy...
--
/jussi
The woman is a highly socially-practised
Master in body language, dab-handed actress
She's calculating all the mad facts and the figures
While you're pretending to listen staring at her tits
* The Streets
milivella
> As every Englishman knows, Tomas Brolin was one of the best to lace 'em up.
> On par with the likes of Cruyff, Bergkamp, Boniek, Ronaldo, Brehme, not far
> away from people like Pele.
:)
Let me play the devil's advocate: England has seen a broken Brolin,
because of his 1994 injury. And, as my uncle says (really), "put a
Brolin in Parma, and you have two European cups and one more final in
three years (Cup Winners' Cup winner 1992-93 + runner-up 1993-94, UEFA
Cup winner 1994-95, and a UEFA Super Cup to boot). Put Stoichkov
instead of him, and you have a 6th place in the league and an
elimination at the quarter finals of the CWC (1995-96)."
--
Cheers
milivella
milivella
> Well, my hunch is that the variation in goals conceded must be quite a
> bit smaller than the variation in goals scored. In other words, the
> best defender in the world is maybe worth 0.1 goals per game, while
> the worst defenders is worth -0.1 goals per game. But the best
> attacker is maybe worth 0.25 goals per game.
And the theoretical reason could be something like: if you have the
best attacker in the world, give him the ball 100% of the times; if
the opponent have the best defender in the world, attack *on the other
side* 100% of the times. A discussion from 2006 raised similar
thoughts:
http://groups.google.com/group/rec.sport.soccer/browse_thread/thread/ecdda5d4f7e7a2d1/
(OK, the opponent team could mark the best attacker in the world with
3 men, but this would leave 2 free attackers, so he's still worth a
lot.)
--
Cheers
milivella
milivella
> I think you got it right.
No, I hadn't. But now I hopefully understand it: thanks for the
simpler example! :)
Is there any physical phenomenon that I can think at to visualize the
energy relationships among teams?
--
Cheers
milivella
milivella
> A thought crossed my mind: why don't we make this a bit more
> sophisticated? First round of votes determines the 5 candidates for each
> position, in the second round we use a rank based voting system (likehttp://en.wikipedia.org/wiki/Condorcet_method) to determine the actual
> winner.
It would be great. Good good idea. Who wants to run it? (I call myself
out.)
--
Cheers
milivella
Futbolmetrix
news:95dfa61d-ab5d-48e3...@a36g2000yqc.googlegroups.com...
> Yes, too bad. Because it's a very interesting conjecture. What are the
> obstacles to prove it? the fact that we have too little data?
That's the main obstacle. Plus the fact that goals occur too infrequently in
soccer, and one bad day can skew the results quite a lot.
> the fact that we should take in account the strength of the opponent team?
If the data were there, taking into account opponent strength wouldn't be a
big issue.
Anyway, maybe my conjecture was wrong, or maybe I was just way off in terms
of magnitudes:
Juventus central defenders, 2008-2009, all official matches (Serie A + CL +
Coppa Italia):
Player gpg allowed gpg allowed
starts when started when didn't start Diff
------------------------------------------------------------
Legrottaglie 33 0.727 1.153 -0.567
Chiellini 33 0.788 1.176 -0.389
Mellberg 26 1.154 0.667 +0.487
Ariaudo 4 1.250 0.891 +0.359
Knezevic 4 1.750 0.848 +0.902
-------------------------------------------------------------
JUVENTUS 50 0.92
There was nearly a 0.5 differences in gpg allowed between Legrottaglie and
Mellberg.
On the other hand, the same type of calculation would show that Manninger
was a much better goalkeeper than Buffon this year. Which, of course, may
very well be true.
D
milivella
> "milivella" <milive...@gmail.com> wrote in message
>
> news:95dfa61d-ab5d-48e3...@a36g2000yqc.googlegroups.com...
>
> > Yes, too bad. Because it's a very interesting conjecture. What are the
> > obstacles to prove it? the fact that we have too little data?
>
> That's the main obstacle. Plus the fact that goals occur too infrequently in
> soccer, and one bad day can skew the results quite a lot.
Right!
> > the fact that we should take in account the strength of the opponent team?
>
> If the data were there, taking into account opponent strength wouldn't be a
> big issue.
So why haven't you done it in your Juve defenders example? ;)
> Juventus central defenders, 2008-2009, all official matches (Serie A + CL +
> Coppa Italia):
>
> Player gpg allowed gpg allowed
> starts when started when didn't start Diff
> ------------------------------------------------------------
> Legrottaglie 33 0.727 1.153 -0.567
> Chiellini 33 0.788 1.176 -0.389
> Mellberg 26 1.154 0.667 +0.487
> Ariaudo 4 1.250 0.891 +0.359
> Knezevic 4 1.750 0.848 +0.902
> -------------------------------------------------------------
> JUVENTUS 50 0.92
Just to check the noise: what's the gpg _scored_ by Juventus when
these players started?
(I'll hopefully follow with some stats.)
--
Cheers
milivella
milivella
Yes, but you must take in account that Jesus thinks something in the
line of "if starts in the NT instead of Pato he must be the best
player in the world!" ;)
--
Cheers
milivella
milivella
> (I'll hopefully follow with some stats.)
Very rough ones, indeed. It's the result of the formula
GD_per_game_with_that_player - GD_per_game_without_that_player
for Barcelona players, but:
- It's a "game with" even if he played 1 minute.
- It doesn't take in account neither the strength of the opponents nor
who played instead of that player.
So it's practically useless. Still, here it is:
G Valdes 2.72
G Pinto -2.47
G Olazabal -2.92
D Marquez 1.39
D Puyol 0.60
D Alves 0.38
D Pique -0.12
D Caceres -0.34
D Abidal -0.36
D Sylvinho -0.84
D Botia -1.89
D Muniesa -2.92
M Xavi 2.36
M Iniesta 0.99
M Toure 0.81
M Sanchez 0.19
M Keita 0.08
M Hleb -0.32
M Busquets -0.48
M Gonzalez -1.89
M Thiago -2.92
M Jeffren -3.00
M Torres -3.00
F Henry 1.39
F Messi 1.38
F Krkic 0.40
F Gudjohnsen -0.36
F Eto'o -1.22
F Pedro -1.59
Another issue with this stat is that it doesn't take in account that,
if a player played a lot of matches (e.g. 37) or very few matches
(e.g. 1), results could be very skewed. I.e. these fictional players
will all have a score of -1:
- Pele: plays 37 matches and his team always win 3-0, then misses just
one match in which his team wins 4-0
- Denilson: plays just 1 match and his team wins 3-0, then misses 37
matches in which his team always win 4-0
- Ronaldinho: plays 19 matches and his team always wins 3-0, then
misses 19 matches in which his team always wins 4-0
Of course we can be pretty sure about Ronaldinho's real value, but we
can't say much about Pele and Denilson, because our knowledge is in
both cases based on just one matches (the one missed or played by that
player).
But, starting with the assumption that the score of a player of which
we know nothing is 0, you can compute a "less rough" score using the
following Bayesian formula (I hope that I'm not completely mistaken!):
(19 - min (matches_with, matches_without)) * 0 + min (matches_with,
matches_without) * rough_score =
min (matches_with, matches_without) * rough_score
This way we would have Pele and Denilson -1, but Ronaldinho -19.
So here is the "less rough" Barcelona ranking:
G Valdes 8.17
G Olazabal -2.92
G Pinto -4.94
D Marquez 20.87
D Puyol 6.00
D Alves 1.53
D Pique -1.60
D Botia -1.89
D Muniesa -2.92
D Caceres -4.48
D Abidal -4.64
D Sylvinho -12.61
M Iniesta 11.85
M Toure 10.56
M Xavi 7.09
M Sanchez 1.35
M Keita 0.76
M Gonzalez -1.89
M Thiago -2.92
M Hleb -6.00
M Jeffren -6.00
M Torres -6.00
M Busquets -6.67
F Henry 12.55
F Messi 9.68
F Krkic 6.00
F Eto'o -2.44
F Gudjohnsen -5.08
F Pedro -9.56
--
Cheers
milivella
Futbolmetrix
> Futbolmetrix:
>> If the data were there, taking into account opponent strength wouldn't be
>> a
>> big issue.
>
> So why haven't you done it in your Juve defenders example? ;)
Wouldn't be a big issue in terms of it taking about an hour or so to get the
data, figure out some adjustment, and do the calculations. As opposed to
several days of work to compile a full set of data on the minutes played by
each player in Serie A, and how many goals were scored in that time period.
>> Juventus central defenders, 2008-2009, all official matches (Serie A + CL
>> +
>> Coppa Italia):
>>
>> Player gpg allowed gpg allowed
>> starts when started when didn't start Diff
>> ------------------------------------------------------------
>> Legrottaglie 33 0.727 1.153 -0.567
>> Chiellini 33 0.788 1.176 -0.389
>> Mellberg 26 1.154 0.667 +0.487
>> Ariaudo 4 1.250 0.891 +0.359
>> Knezevic 4 1.750 0.848 +0.902
>> -------------------------------------------------------------
>> JUVENTUS 50 0.92
>
> Just to check the noise: what's the gpg _scored_ by Juventus when
> these players started?
You know that now I will not resist the urge to go and check. You are an
evil man.
D
milivella
> "milivella" <milive...@gmail.com> wrote in message
>
> news:afa7bc12-40e4-42e2...@a7g2000yqk.googlegroups.com...
>
> > Futbolmetrix:
> >> If the data were there, taking into account opponent strength wouldn't be
> >> a
> >> big issue.
>
> > So why haven't you done it in your Juve defenders example? ;)
>
> Wouldn't be a big issue in terms of it taking about an hour or so to get the
> data, figure out some adjustment, and do the calculations. As opposed to
> several days of work to compile a full set of data on the minutes played by
> each player in Serie A, and how many goals were scored in that time period.
You didn't need to reply, I was kidding. But thanks for giving us an
idea of how much time is needed for such a stat.
> > Just to check the noise: what's the gpg _scored_ by Juventus when
> > these players started?
>
> You know that now I will not resist the urge to go and check. You are an
> evil man.
Yes, but you are too easy to tempt...
--
Cheers
milivella
milivella
> It's the result of the formula
> GD_per_game_with_that_player - GD_per_game_without_that_player
> for Barcelona players
...in national league, I should have added. BTW, maybe there is some
consistence in the stat, because results are not dissimilar for the
UCL. E.g. in Liga "rough" score leaders for each position are:
> G Valdes 2.72
> D Marquez 1.39
> M Xavi 2.36
> F Henry 1.39
and in UCL they are... Valdes (2.74), Marquez (2.20), Xavi (2.64) and
Henry (2.25).
Or, for "less rough" scores:
> G Valdes 8.17
Valdes (2.64)
> D Marquez 20.87
Marquez (11)
> M Iniesta 11.85
Iniesta is indeed last in UCL (-1.55)! But the next two in Liga...
> M Toure 10.56
> M Xavi 7.09
...are the top 2 in UCL (Toure 3, Xavi 2.64).
> F Henry 12.55
Henry 6.75.
But maybe it's just a coincidence, because the other positions (2nd,
3rd, etc.) show more variance.
--
Cheers
milivella
El Kot
Not exactly the same, but positions of atoms in a crystal lattice,
or in big molecules, follow similar energy behavior. When an atom moves
away from its position, it experiences forces that bring it back. The
most notable difference is that in the soccer case, each team has unique
forces acting on it, while the atoms (especially in the lattice case)
are many and similar.
Futbolmetrix
>
> Juventus central defenders, 2008-2009, all official matches (Serie A + CL +
> Coppa Italia):
Updated tables, including goals scored:
Unadjusted
>
> gpgF gpgF gpgA gpgA
> Player Pld with without Diff_F with without Diff_A DIFF_TOT ------ ---------------------------------------------------------------------------
> Legro 35 1.66 1.76 -0.11 0.74 1.29 -0.55 0.44
> Chiel 34 1.88 1.33 0.55 0.79 1.17 -0.37 0.92
> Mellb 27 1.52 1.88 -0.36 1.15 0.68 0.47 -0.83
---------------------------------------------------------------------------------
So Chiellini and Legrottaglie were not only better at defending than
Mellberg, they also contributed more offensively. Of course it's very
possible that the two things were linked. If Mellberg was letting in
goals at the back, maybe it was more difficult for Juve to score
because their opponents were parking the bus. Or vice versa.
Adjusted for opponent strength (Serie A only) [*]
> gpgF gpgF gpgA gpgA
> Player Pld with without Diff_F with without Diff_A DIFF_TOT ---------------------------------------------------------------------------------
> Legro 26 0.45 0.60 -0.15 -0.39 -0.10 -0.28 0.13
> Chiel 25 0.70 0.11 0.59 -0.55 0.18 -0.72 1.31
> Mellb 20 0.44 0.56 -0.12 -0.06 -0.56 0.50 -0.62
---------------------------------------------------------------------------------
[*] This is calculated by simply subtracting the average gpg of the
opponents in each match. So if Juve draws 1-1 with Fiorentina (53
goals for, 1.39 gfpg; 38 against, 1.00 gapg), the offensive
contribution to the player's score is 1-1.00 = 0.00, and the defensive
contribution is 1-1.39 = -0.39. But if Juve draws 1-1 with Catania,
the respective contributions are -0.34 and -0.08.
The ranking stays the same, not sure what we can say about the
magnitudes because things are not in the same scale.
I liked Mellberg and I actually thought that he had given a solid
contribution at the back. But, looking at the numbers, I have to say
that I'm happy he's leaving for Olympiakos. Not so sure that Cannavaro
will be better, though.
More analysis tomorrow (maybe)
D
milivella
> So Chiellini and Legrottaglie were not only better at defending than
> Mellberg, they also contributed more offensively. Of course it's very
> possible that the two things were linked. If Mellberg was letting in
> goals at the back, maybe it was more difficult for Juve to score
> because their opponents were parking the bus. Or vice versa.
Right.
> Adjusted for opponent strength (Serie A only) [*]
Thanks.
> The ranking stays the same, not sure what we can say about the
> magnitudes because things are not in the same scale.
Why? Isn't the sum of the scores supposed to be the same in both cases
(i.e. unadjusted and adjusted)?
E.g. simplest example possible, considering only goals against (and
not taking in account home/away):
Fiorentina scores in average 1 goal per game
Catania scores in average 0 goals per game
Juventus_with_Chiellini 1-1 Fiorentina
Juventus_with_Mellberg 1-1 Catania
Unadjusted score:
[player, gpgA with, gpgA wihout,Diff_A]
Chiellini,1,1,0
Mellberg,1,1,0
Chiellini+Mellberg=0
Adjusted score:
Chiellini,0,1,-1
Mellberg,1,0,1
Chiellini+Mellberg=0
Maybe I'm missing something very banal, but I've 3.5 hours of sleep
and 2 flights in the last 41 hours, so please forgive me. :)
> I liked Mellberg and I actually thought that he had given a solid
> contribution at the back. But, looking at the numbers, I have to say
> that I'm happy he's leaving for Olympiakos.
La Gazzetta dello Sport's average rating:
Legrettaglie 6.24
Chiellini 6.21
Mellberg 5.84
(i.e. out of 86 Serie A defenders with at least 19 appearances,
Legrottaglie is 2nd, Chiellini 4th, Mellberg 45th)
So it's Futbolmetrix vs. both stats and pundits... (I support
Futbolmetrix, of course!)
> Not so sure that Cannavaro
> will be better, though.
Eh...
> More analysis tomorrow (maybe)
So we hope!
--
Cheers
milivella
milivella
> Futbolmetrix:
>
> > The ranking stays the same, not sure what we can say about the
> > magnitudes because things are not in the same scale.
>
> Why? Isn't the sum of the scores supposed to be the same in both cases
> (i.e. unadjusted and adjusted)?
I.e. it looks to me that diffs are in the same scale. And if you want
to put everything in the same scale, just consider unadjusted scores
as based on matches against the average Serie A team (that has 1.30
goals for per game, and of course 1.30 against):
> gpgF gpgF gpgA gpgA
> Player Pld with without Diff_F with without Diff_A DIFF_TOT
> Legro 35 0.36 0.46 -0.11 -0.56 -0.01 -0.55 0.44
> Chiel 34 0.58 0.03 0.55 -0.51 -0.13 -0.37 0.92
> Mellb 27 0.22 0.58 -0.36 -0.15 -0.62 0.47 -0.83
Now everything should be in the same scale of the adjusted score, that
I quote here:
> gpgF gpgF gpgA gpgA
> Player Pld with without Diff_F with without Diff_A DIFF_TOT
> Legro 26 0.45 0.60 -0.15 -0.39 -0.10 -0.28 0.13
> Chiel 25 0.70 0.11 0.59 -0.55 0.18 -0.72 1.31
> Mellb 20 0.44 0.56 -0.12 -0.06 -0.56 0.50 -0.62
(Or add 1.3 to all the gpg's in the adjusted table, of course.)
But something doesn't work: I expect that, if Legro's gpgF_with goes
up in the adjusted score (because he played against above-the-average
team), gpgF_without should go down (because from the previous point
you derive that, when Legore wasn't there, Juventus played against
below-the-average team). But in some cases my expectations aren't
fulfilled. Why?
--
Cheers
milivella
Futbolmetrix
> milivella:
>
> I.e. it looks to me that diffs are in the same scale. And if you want
> to put everything in the same scale, just consider unadjusted scores
> as based on matches against the average Serie A team (that has 1.30
> goals for per game, and of course 1.30 against):
Yes, that's a convenient way of looking at it. Now formally, I think
that all the averages should be calculated excluding the games played
by the team against Juve, but that shouldn't change things much.
> > gpgF gpgF gpgA gpgA
> > Player Pld with without Diff_F with without Diff_A DIFF_TOT
> > Legro 35 0.36 0.46 -0.11 -0.56 -0.01 -0.55 0.44
> > Chiel 34 0.58 0.03 0.55 -0.51 -0.13 -0.37 0.92
> > Mellb 27 0.22 0.58 -0.36 -0.15 -0.62 0.47 -0.83
>
> Now everything should be in the same scale of the adjusted score, that
> I quote here:
>
> > gpgF gpgF gpgA gpgA
> > Player Pld with without Diff_F with without Diff_A DIFF_TOT
> > Legro 26 0.45 0.60 -0.15 -0.39 -0.10 -0.28 0.13
> > Chiel 25 0.70 0.11 0.59 -0.55 0.18 -0.72 1.31
> > Mellb 20 0.44 0.56 -0.12 -0.06 -0.56 0.50 -0.62
>
> (Or add 1.3 to all the gpg's in the adjusted table, of course.)
>
> But something doesn't work: I expect that, if Legro's gpgF_with goes
> up in the adjusted score (because he played against above-the-average
> team), gpgF_without should go down (because from the previous point
> you derive that, when Legore wasn't there, Juventus played against
> below-the-average team). But in some cases my expectations aren't
> fulfilled. Why?
The unadjusted table was based on all of Juve's official matches, not
just Serie A.
Futbolmetrix
> Legrettaglie 6.24
> Chiellini 6.21
> Mellberg 5.84
> (i.e. out of 86 Serie A defenders with at least 19 appearances,
> Legrottaglie is 2nd, Chiellini 4th, Mellberg 45th)
Well, it's reassuring to see that the +/- stat gives results that are
not too dissimilar from the player ratings. There's not really that
much separating Chiellini and Legrottaglie (especially in terms of
gpga_with), while Mellberg is clearly a notch below.
> So it's Futbolmetrix vs. both stats and pundits... (I support
> Futbolmetrix, of course!)
Stats obviously wins, because they're based on a full sample of
matches, not just the ones I watched.
D
milivella
> Well, it's reassuring to see that the +/- stat gives results that are
> not too dissimilar from the player ratings.
Yes it it! :)
--
Cheers
milivella
milivella
> It's the result of the formula
> GD_per_game_with_that_player - GD_per_game_without_that_player
> for Barcelona players, but:
> - It's a "game with" even if he played 1 minute.
> - It doesn't take in account neither the strength of the opponents nor
> who played instead of that player.
> So it's practically useless. Still, here it is:
...and then a not so bad ranking of Barcelona players followed. After
all, it had "Valdes - Marquez Puyol Alves Pique - Iniesta Toure Xavi -
Henry Messi Krkic" as the best XI. But the method falls apart when
applied to Manchester United (league games):
G Foster 0.72
G Kuszczak -0.71
G van der Sar -1.25
D Neville 10.96
D Evans 6.00
D O'Shea 5.79
D Vidic 5.18
D Rafael 4.27
D Eckersley 3.89
D De Laet -0.16
D Brown -0.33
D Evra -1.78
D Ferdinand -9.17
M Fletcher 6.16
M Gibson 6.00
M Welbeck 3.89
M Tosic 3.89
M Nani 3.08
M Sung 1.60
M Martin -0.16
M Carrick -0.37
M Possebom -1.60
M Hargreaves -3.50
M Giggs -10.22
M Scholes -13.24
M Anderson -17.52
F Ronaldo 3.21
F Berbatov 1.13
F Macheda 0.41
F Tevez -3.29
F Rooney -7.27
While everyone but me will be happy to see Anderson in the dead last
position, the ranking seems odd to me. OTOH, United had 0.29 goals
against per game in the 14 games *without* Ferdinand, while 0.92 with
him on the field; and United's goal difference per game doubles in the
8 matches with Rooney off (1.88 vs. 0.97). I guess that weighting
opponent's strength is necessary to avoid skewing.
Anyway, can someone who knows United's season better than me (i.e.
anyone in the ng) please have his say about this ranking? Do you see
any insight in it? Is it totally rubbish?
--
Cheers
milivella
milivella
> But the method falls apart when
> applied to Manchester United (league games):
It was the Bayesian (or "less rough") score, I forgot to write it.
To make judgments easier, here are the Barca and United players who
departed more from the average GD of their team (listed always from
the best to the worst in that segment):
Very good (more than +10)
Barca: Marquez, Henry, Iniesta, Toure
United: Neville
Good (from +5 to +10)
Barca: Messi, Valdes, Xavi, Puyol, Krkic
United: Fletcher, Evans, Gibson, O'Shea, Vidic
Bad (from -5 to -10)
Barca: Gudjohnsen, Hleb, Jeffren, Torres, Busquets, Pedro
United: Rooney, Ferdinand
Very bad (less than -10)
Barca: Sylvinho
United: Giggs, Scholes, Anderson
--
Cheers
milivella
milivella
> To make judgments easier, here are the Barca and United players who
> departed more from the average GD of their team (listed always from
> the best to the worst in that segment):
Let's see where Barca's and United's players of the seasons
http://www.goal.com/en/news/12/spain/2009/05/13/1260682/goalcom-special-who-is-barcelonas-player-of-the-season
http://www.goal.com/en/news/1717/editorial/2009/05/11/1258464/goalcom-special-who-is-manchester-uniteds-player-of-the-season
are placed:
> Very good (more than +10)
> Barca: Marquez, Henry, Iniesta, Toure
> United: Neville
Henry, Iniesta: in the top 5 according to Goal.com
> Good (from +5 to +10)
> Barca: Messi, Valdes, Xavi, Puyol, Krkic
> United: Fletcher, Evans, Gibson, O'Shea, Vidic
Messi, Vidic: best player
Xavi, Puyol: in the top 5
Here, between "good" and "bad" players, there are the "average"
players (from +5 to -5), and some of them have been called by
Goal.com:
Ronaldo, van der Sar: in the top 5
> Bad (from -5 to -10)
> Barca: Gudjohnsen, Hleb, Jeffren, Torres, Busquets, Pedro
> United: Rooney, Ferdinand
Rooney: in the top 5
> Very bad (less than -10)
> Barca: Sylvinho
> United: Giggs, Scholes, Anderson
Giggs: in the top 5
Again, it seemes that things worked well enough for Barca, but not for
United.
--
Cheers
milivella
Futbolmetrix
> milivella:
>
>> But the method falls apart when
>> applied to Manchester United (league games):
>
> It was the Bayesian (or "less rough") score, I forgot to write it.
Your formula is interesting, but it has little to do with Bayes' rule. What
do the unadjusted scores look like?
Or try this:
m0 = prior mean = 0
s0 = prior standard deviation = 1 (This is the tricky part...)
m_i = player i's mean goal difference (with - without)
s_i = player i's standard deviation of the goal difference
n_i = number of matches played by i.
(1/s0^2)*m0 + (n_i/s_i^2)*m_i
Player i's score = -----------------------------
(1/s0^2) + (n_i/s_i^2)
D
milivella
> "milivella" <milive...@gmail.com> wrote in message
>
> news:fbde0bc1-8737-48d7...@c1g2000yqi.googlegroups.com...
>
> > milivella:
>
> >> But the method falls apart when
> >> applied to Manchester United (league games):
>
> > It was the Bayesian (or "less rough") score, I forgot to write it.
>
> Your formula is interesting
Thanks, my Master!
> but it has little to do with Bayes' rule.
Sorry, Dark Lord of Stats!
> What
> do the unadjusted scores look like?
Here they are, You Ineffable!
G Foster 0.36
G Kuszczak -0.18
G van der Sar -0.21
D Eckersley 1.94
D Vidic 1.29
D Neville 0.73
D O'Shea 0.64
D Evans 0.38
D Rafael 0.27
D Brown -0.04
D Evra -0.16
D De Laet -0.16
D Ferdinand -0.65
M Gibson 2.00
M Welbeck 1.94
M Tosic 1.94
M Fletcher 0.47
M Nani 0.26
M Sung 0.12
M Carrick -0.03
M Martin -0.16
M Possebom -0.53
M Scholes -0.78
M Giggs -0.93
M Anderson -1.03
M Hargreaves -1.75
F Ronaldo 0.64
F Berbatov 0.19
F Macheda 0.10
F Tevez -0.33
F Rooney -0.91
> Or try this:
Your desires are orders, Rum Bottle Fix!
> s_i = player i's standard deviation of the goal difference
How do I tell the Mighty Excel to compute it, dear Limbo Fur Text?!
--
Cheers
milivella
Futbolmetrix
> > do the unadjusted scores look like?
>
> Here they are, You Ineffable!
Hmmm... they still look out of whack. Could you do the calculations
just for games started, as opposed to total appearances? (Or maybe
Ferdinand and Rooney are way overrated...)
>
> Your desires are orders, Rum Bottle Fix!
>
> > s_i = player i's standard deviation of the goal difference
>
> How do I tell the Mighty Excel to compute it, dear Limbo Fur Text?!
My assumption was that you actually had the individual scores for each
game played by each player, but now I'm guessing that you only had the
aggregates (if you're getting your info from footballdatabase.eu). In
that case, you'll have to do with educated guesswork! For example,
assume that the distribution of goals scored and goals conceded by
ManU follows a Poisson distribution, and therefore standard deviation
= sqrt(mean).
Glad to help, Mel Vialli.
milivella
> On Jul 6, 7:16 pm, milivella <milive...@gmail.com> wrote:
>
> > Futbolmetrix:
>
> > > What
> > > do the unadjusted scores look like?
>
> > Here they are, You Ineffable!
>
> Hmmm... they still look out of whack. Could you do the calculations
> just for games started, as opposed to total appearances?
Quick answer: no. :(
My source is
http://www.zerozerofootball.com/
where I find the aggregated stats (wins-draws-losses and goals for-
against) for all the matches that each player has played (even just
for one minute). It would be possible to extract data from the list of
matches played by that player (and so it would be possible to consider
just matches started, or individual scores), though.
> > > s_i = player i's standard deviation of the goal difference
>
> > How do I tell the Mighty Excel to compute it, dear Limbo Fur Text?!
>
> My assumption was that you actually had the individual scores for each
> game played by each player, but now I'm guessing that you only had the
> aggregates (if you're getting your info from footballdatabase.eu). In
> that case, you'll have to do with educated guesswork! For example,
> assume that the distribution of goals scored and goals conceded by
> ManU follows a Poisson distribution, and therefore standard deviation
> = sqrt(mean).
So I'm replacing s_i with the squared root of the goal difference in
the games played by the player, and then computing and posting the
result.
> Glad to help, Mel Vialli.
:)
Now how can I keep posting under this boring "milivella" anymore?
--
Cheers
milivella
milivella
> Futbolmetrix:
>
> > On Jul 6, 7:16 pm, milivella <milive...@gmail.com> wrote:
>
> > > Futbolmetrix:
>
> > > > s_i = player i's standard deviation of the goal difference
>
> > > How do I tell the Mighty Excel to compute it, dear Limbo Fur Text?!
>
> > My assumption was that you actually had the individual scores for each
> > game played by each player, but now I'm guessing that you only had the
> > aggregates (if you're getting your info from footballdatabase.eu). In
> > that case, you'll have to do with educated guesswork! For example,
> > assume that the distribution of goals scored and goals conceded by
> > ManU follows a Poisson distribution, and therefore standard deviation
> > = sqrt(mean).
>
> So I'm replacing s_i with the squared root of the goal difference in
> the games played by the player
Sorry, I meant to ask: have I to replace s_i with sqrt(m_i)?
--
Cheers
milivella
milivella
> Or try this:
>
> m0 = prior mean = 0
> s0 = prior standard deviation = 1 (This is the tricky part...)
> m_i = player i's mean goal difference (with - without)
> s_i = player i's standard deviation of the goal difference
> n_i = number of matches played by i.
>
> (1/s0^2)*m0 + (n_i/s_i^2)*m_i
> Player i's score = -----------------------------
> (1/s0^2) + (n_i/s_i^2)
Daniele (always very kind!) computed the results (using a slightly
different formula), but Ferdinand and Rooney are always among the
worst players.
To check whether this is due to the strength of the opponents faced
(i.e. Ferdinand and Rooney are used against the Chelseas, rested
against the WBAs), I'm computing the GD with and without Rooney, but
this time weighting in opponents' strength. E.g.
Sunderland, when played home, in average had 1.11 goals for vs. 1.32
against.
Sunderland - Manchester United 1-2
When looking at United's performance in this match, instead of
considering 0 goals for - 0 against, we subtract the expected numbers
of goals for/against (i.e. the average number of goals against/for
Sunderland when they played home) from the actual number of goals for/
against:
GF 2-1.32=+0.68
GA 1-1.11=-0.11
I.e. compared to the average EPL team playing against Sunderland away,
United had 0.68 goals-for more and 0.11 goals-against less.
(Indeed, I'll consider Sunderland's averages when playing against all
the *other* EPL teams, excluding the match against United.)
More, I'll post the results considering both every match played by
Rooney (even just 1 minute) and only matches started by him (to check
if few-minute appearances introduce noise).
Of course doing it for just a player is quite useless, but, even if
Kicker format matches played by each player nicely enough, it takes
some time to put everything in a spreadsheet.
--
Cheers
milivella
milivella
> To check whether this is due to the strength of the opponents faced
> (i.e. Ferdinand and Rooney are used against the Chelseas, rested
> against the WBAs), I'm computing the GD with and without Rooney, but
> this time weighting in opponents' strength. E.g.
> Sunderland, when played home, in average had 1.11 goals for vs. 1.32
> against.
> Sunderland - Manchester United 1-2
> When looking at United's performance in this match, instead of
> considering 0 goals for - 0 against, we subtract the expected numbers
> of goals for/against (i.e. the average number of goals against/for
> Sunderland when they played home) from the actual number of goals for/
> against:
> GF 2-1.32=+0.68
> GA 1-1.11=-0.11
> I.e. compared to the average EPL team playing against Sunderland away,
> United had 0.68 goals-for more and 0.11 goals-against less.
>
> (Indeed, I'll consider Sunderland's averages when playing against all
> the *other* EPL teams, excluding the match against United.)
Doing so, United's weighted GD is +44, exactly as United's unweighted
GD. So weighted and unweighted data (about goal difference, not about
goals for and against!) are comparable.
Here are Rooney's data:
Rough data, unweighted:
Rooney playing at least 1 minute: GF 1.70 - GA 0.73 = GD +0.97
Rooney not playing: GF 2.13 - GA 0.25 = GD +1.88
Rooney's provision: -0.91
Rough data, unweighted:
Rooney playing >45 minutes: GF 1.71 - GA 0.83 = GD +0.88
Rooney not playing or playing <45 minutes: GF 1.93 - GA 0.29 = GD 1.64
Rooney's provision: -0.76
Rough data, weighting opponents' strength:
Rooney playing at least 1 minute: GF 0.48 - GA -0.52 = GD +1.00 [i.e.
0.48+0.52]
Rooney not playing: GF 0.81 - GA -0.96 = GD +1.77
Rooney's provision: -0.77
Rough data, weighting opponents' strength:
Rooney playing >45 minutes: GF 0.50 - GA -0.47 = GD +0.97
Rooney not playing or playing <45 minutes: GF 0.63 - GA -0.84 = GD
+1.47
Rooney's provision: -0.50
I.e. Rooney's provision is always negative (i.e. United have better
results without him), even if it's less negative when you consider
that some opponents are tougher than others, and when you count only
matches played by Rooney for more than half.
> Of course doing it for just a player is quite useless, but, even if
> Kicker format matches played by each player nicely enough, it takes
> some time to put everything in a spreadsheet.
Well, maybe it could be done in a quicker way, but anyway it takes
some time, and I don't know whether the first results are inspiring.
--
Cheers
milivella
milivella
> I.e. Rooney's provision is always negative (i.e. United have better
> results without him), even if it's less negative when you consider
> that some opponents are tougher than others, and when you count only
> matches played by Rooney for more than half.
Daniele made me note (thanks!) that Rooney's score is negative
basically because he has not played in the two 5-0s against WBA and
Stoke. Such results can skew things when you consider goal difference
alone, and this is the reason why in the past I've considered a
player's provision as *points* with him minus points without him. (Of
course counting both points and goal difference, as e.g. in the Elo
rating and in Elko's "energy rankings", is the best choice.)
So I'll soon post some data considering points instead of GD.
--
Cheers
milivella
milivella
> So I'll soon post some data considering points instead of GD.
Here they are. You can compare them with the GD data posted before in
this same thread.
Barcelona, rough (= difference in points per match with-without)
G Valdes 2.12
G Pinto -1.89
G Olazabal -2.35
D Marquez 0.81
D Alves 0.60
D Puyol 0.39
D Pique 0.32
D Abidal -0.26
D Caceres -0.44
D Sylvinho -0.48
D Botia -1.32
D Muniesa -2.35
M Xavi 1.40
M Iniesta 0.91
M Busquets 0.68
M Sanchez 0.35
M Toure 0.32
M Keita -0.06
M Hleb -0.37
M Gonzalez -1.32
M Thiago -2.35
M Jeffren -2.42
M Torres -2.42
F Messi 0.88
F Henry 0.82
F Krkic -0.07
F Gudjohnsen -0.33
F Eto'o -0.75
F Pedro -1.14
Barcelona, pseudo-Bayesian (= rough * min
(matches_with,matches_without) )
G Valdes 6.37
G Olazabal -2.35
G Pinto -3.78
D Marquez 12.13
D Pique 4.20
D Puyol 3.93
D Alves 2.41
D Botia -1.32
D Muniesa -2.35
D Abidal -3.40
D Caceres -5.72
D Sylvinho -7.17
M Iniesta 10.92
M Busquets 9.58
M Xavi 4.20
M Toure 4.20
M Sanchez 2.42
M Keita -0.52
M Gonzalez -1.32
M Thiago -2.35
M Jeffren -4.83
M Torres -4.83
M Hleb -7.00
F Henry 7.34
F Messi 6.16
F Krkic -1.09
F Eto'o -1.50
F Gudjohnsen -4.67
F Pedro -6.81
Manchester United, rough
G Foster 0.75
G Kuszczak 0.24
G van der Sar -0.25
D Vidic 1.44
D Eckersley 0.75
D De Laet 0.73
D O'Shea 0.67
D Neville 0.62
D Evans 0.47
D Rafael 0.47
D Brown -0.21
D Evra -0.36
D Ferdinand -0.56
M Gibson 0.77
M Welbeck 0.75
M Tosic 0.75
M Martin 0.73
M Fletcher 0.56
M Carrick 0.28
M Nani 0.19
M Possebom 0.05
M Scholes -0.01
M Sung -0.14
M Anderson -0.31
M Giggs -0.36
M Hargreaves -1.89
F Ronaldo 1.02
F Macheda 0.79
F Berbatov 0.34
F Tevez -0.01
F Rooney -0.11
Manchester United, pseudo-Bayesian
G Foster 1.50
G Kuszczak 0.94
G van der Sar -1.50
D Neville 9.35
D Evans 7.55
D Rafael 7.55
D O'Shea 6.03
D Vidic 5.76
D Eckersley 1.50
D De Laet 0.73
D Brown -1.67
D Evra -3.96
D Ferdinand -7.83
M Fletcher 7.24
M Carrick 3.07
M Gibson 2.31
M Nani 2.23
M Welbeck 1.50
M Tosic 1.50
M Martin 0.73
M Possebom 0.14
M Scholes -0.14
M Sung -1.88
M Hargreaves -3.78
M Giggs -3.96
M Anderson -5.29
F Ronaldo 5.12
F Macheda 3.18
F Berbatov 2.06
F Tevez -0.14
F Rooney -0.87
Quickly comparing these results with the ones based on goal
difference, I don't see any revolution.
---
Now, to the question: what happens if we weight in opponents'
strength? To do it, I used a little trick. Let me show it through an
example, involving United's best result of the season:
- First of all, I considered win = 2 points.
- Under this system, Chelsea away averaged 1.67 points per game (30
points in 18 matches).
- I.e. they conceded in average 0.33 points per game.
- United won against Chelsea at Manchester, so this result is worth 2
- 0.33 = 1.67 points.
I've computed Rooney's score this way, and these are the results:
United when Rooney played at least 1 minute: 0.56 points per game
United when Rooney didn't play: 0.88
Rooney's provision: -0.32
United when Rooney played more than 45 minutes: 0.59
United in all the other matches: 0.70
Rooney's provision: -0.11
Rooney's score remains negative.
--
Cheers
milivella
milivella
> Now, to the question: what happens if we weight in opponents'
> strength? To do it, I used a little trick. Let me show it through an
> example, involving United's best result of the season:
> - First of all, I considered win = 2 points.
> - Under this system, Chelsea away averaged 1.67 points per game (30
> points in 18 matches).
> - I.e. they conceded in average 0.33 points per game.
> - United won against Chelsea at Manchester, so this result is worth 2
> - 0.33 = 1.67 points.
Of course this means that a player playing only against weak teams
can't have more than a given score, because you can't do anything
better than winning.
With goal difference, OTOH, you don't have such a problem, because you
can always score one more goal.
E.g. you play home against Middlesbrough. Even winning, you can't gain
more than 0.33 weighted points, but, even if Middlesbrough's goals
difference in away games is awful (-1.39), winning 5-0 you can gain a
lot of weighted GD.
--
Cheers
milivella
milivella
> But the method falls apart when
> applied to Manchester United (league games):
One possible way to make the results more acceptable to me is if they
show consistence. E.g. if Ferdinand and Rooney have negative score for
the past seasons as well, this could be a sign that we're not just
measuring noise.
I don't know whether it's scientific method, but I want to tell what I
expect from the 2007-08 data before checking the results:
- I'll consider the results not noisy if there will be a good
correlation to 2008-09 results, i.e. if a given player has a similar
score in the two seasons.
- I'm interested in Ferdinand's score more than in Rooney's one,
because (if I'm not wrong) there haven't been many changes in United's
defense, while Berbatov could have changed things in attack. E.g. it
could be the case that Rooney was irreplaceable in 2007-08 - and so
all the matches without him saw United had trouble in scoring - but
had Berbatov as an alternative in 2008-09 - and even without Rooney
United had no problems in finding the net -.
---
[copy-paste, copy-paste, etc.]
---
Rough difference GD with - GD without:
G Kuszczak 0.33
G van der Sar -0.33
G Foster -0.54
D Simpson 2.32
D Pique 0.48
D Ferdinand 0.21
D Evra 0.15
D Vidic 0.03
D O'Shea -0.10
D Silvestre -1.66
D Brown -3.67
M Fletcher 0.82
M Anderson 0.27
M Nani 0.04
M Park -0.04
M Scholes -0.07
M Carrick -0.76
M Giggs -1.11
M Hargreaves -1.11
M Eagles -1.43
F Ronaldo 1.71
F Rooney 1.51
F Tevez 0.59
F Saha 0.01
Pseudo-Bayesian, i.e. previous_score * min
(matches_with,matches_without)
G Kuszczak 2.97
G Foster -0.54
G van der Sar -2.97
D Simpson 6.97
D Pique 4.28
D Evra 0.73
D Ferdinand 0.63
D Vidic 0.19
D O'Shea -1.00
D Silvestre -4.97
D Brown -7.33
M Fletcher 13.09
M Anderson 3.75
M Nani 0.46
M Park -0.46
M Scholes -1.00
M Carrick -5.29
M Eagles -5.71
M Giggs -7.74
M Hargreaves -16.70
F Rooney 16.59
F Ronaldo 6.82
F Tevez 2.35
F Saha 0.10
How to interpret the results? I don't know, but they don't seem good
to me (of course in a different sense the falsification of an
hypothesis is a good result!).
I'll just compare United's best and worst 4-4-2 according in the two
seasons (according to the pseudo-Bayesian formula):
United's best 4-4-2:
2007-08: Kuszczak - Simpson Pique Evra Ferdinand - Fletcher Anderson
Nani Park - Rooney Ronaldo
2008-09: Foster - Neville Evans O'Shea Vidic - Fletcher Gibson Welbeck
Tosic - Ronaldo Berbatov
United's worst 4-4-2:
2007-08: van der Sar - Vidic O'Shea Silvestre Brown - Carrick Eagles
Giggs Hargreaves - Tevez Saha
2008-09: van der Sar - De Laet Brown Evra Ferdinand - Hargreaves Giggs
Scholes Anderson - Tevez Rooney
In both best teams:
- Fletcher (best midfielders in both seasons)
- Ronaldo
In both worst teams:
- van der Sar
- Brown
- Hargraves
- Giggs
- Tevez
Once in the best team, once in the worst one:
- Evra
- Ferdinand
- Anderson (second best midfielder in 07-08, worst one in 08-09)
- Rooney (best forward in 07-08, worst one in 08-09)
- Vidic (actually 5th defenders in 07-08, 4th in 08-09)
- O'Shea
I'd be glad to hear your opinion.
--
Cheers
milivella
milivella
> How to interpret the results? I don't know, but they don't seem good
> to me
But, but. I would like to compare them with *subjective* ratings (e.g.
by journalists). For now, I've checked (at Soccernet) the Actim Index,
that it's an objective stat like the sort of "plus-minus" we're
judging here, but it's based on different data (basically what the
player does on the field). And I've discovered that...
In 2007-08, there were 5 United players in the top 25 Actim Index
positions:
1. Ronaldo
6. Tevez
11. Rooney
14. Ferdinand
17. Brown
The former four have a positive "plus-minus", while Brown has a
negative one.
United player in Actim Index 2008-09:
4. Berbatov
10. Vidic
11. Ronaldo
15. van der Sar
The former three have a positive "plus-minus", while vdS has a
negative one.
What is maybe more interesting is that Tevez, Rooney, Ferdinand are no
more present in the top 25 *and* have a lower "plus-minus" (indeed,
they've passed from a positive one to a negative one); while Vidic and
van der Sar are now present in the top 25 *and* have an higher "plus-
minus" (Vidic from 0.19 to 5.18, vdS from -2.97 to -1.25). So there
*could* be a correlation Actim Index <-> "plus-minus".
---
To verify it, I've checked the United players in the Actim Index
2008-09 top 100 (at the official Premier League site):
1. Berbatov
2. Vidic
3. Ronaldo
4. van der Sar
5. Rooney
6. Evra
7. Carrick
8. O'Shea
9. Ferdinand
10. Fletcher
11. Giggs
I don't know whether they've rated other United players as well, so
I'll compare this ranking with the "plus-minus" one regarding these 11
players alone. Here they are, ranked according their "plus-minus"
score:
1. (+9) Fletcher
2. (+6) O'Shea
3. (-1) Vidic
4. (-1) Ronaldo
5. (-4) Berbatov
6. (+1) Carrick
7. (-3) van der Sar
8. (-2) Evra
9. (-4) Rooney
10. (-1) Ferdinand
11. (==) Giggs
It could be the case that "plus-minus" succeeds in spotting the
obscure hard workers like Fletcher and O'Shea, but it must be
concluded that there isn't any strong correlation Actim Index <->
"plus-minus".
--
Cheers
milivella
Futbolmetrix
> milivella:
>
>> But the method falls apart when
>> applied to Manchester United (league games):
>
> One possible way to make the results more acceptable to me is if they
> show consistence. E.g. if Ferdinand and Rooney have negative score for
> the past seasons as well, this could be a sign that we're not just
> measuring noise.
>
> I don't know whether it's scientific method,
It's actually a very sensible thing to do. One of the greatest discoveries
of sabermetrics (due to Voros McCracken, winner of the Euro 2004 RSS
Sophisticated Prediction Contest) is that pitchers have actually very little
influence on whether balls hit in play become base hits or outs. One of the
main lines of his argument is that there is essentially no correlation in a
pitcher's ability to prevent hits across years. Essentially, it's all due to
luck.
http://www.baseballprospectus.com/article.php?articleid=878
More comments later.
D
milivella
> More comments later.
Can't wait for them.
In the mean time, here are the 20 players with the highest market
value accordint to Transfermarkt
http://www.transfermarkt.co.uk/en/basic/2009/international/marktwerte/tops.html
ranked by "plus-minus" (2008-09 season, of course):
Iniesta 11.85
Kaka 11.58
Messi 9.68
Fabregas 8.81
Xavi 7.09
Villa 5.58
Vidic 5.18
Ronaldo 3.21
Gerrard 2.71
Ibrahimovic 2.17
Torres -0.43
De Rossi -0.70
Lampard -1.78
Eto'o -2.44
Essien -3.50
Robinho -4.20
Ribery -4.52
Rooney -7.27
Ferdinand -9.17
Drogba -10.96
Again, it seems to have some sense (in general, players with negative
scores had a worse season than players with a positive score), but
there are some oddities (hasn't Ribery been fundamental for Bayern
this term?).
--
Cheers
milivella
milivella
> ranked by "plus-minus" (2008-09 season, of course):
Good news, Daniele! I've added the players from 21th to 40th and Diego
tops the ranking! I guess that you now agree that this is the perfect
stat and that the case is closed.
Really, here are top 5 and worst 5:
1. Diego +0.69
2. Iniesta +0.46
3. Henry +0.43
4. Fabregas +0.42
5. Kaka +0.37
...
36. Ronaldinho -0.21
37. Rooney -0.24
38. Buffon -0.31
39. Ferdinand -0.38
40. Drogba -0.48
The figures are different since the last time because I've changed the
formula a bit, so that it now tells how much each player affects the
goal difference in an average match. The formula is now:
(min/(matches-min))*(GD_with-GD_without)
where:
min = min(games_with,games_without)
matches = 38 [34 for the Bundesliga]
This way, the result equals the difference (GD_with-GD_without) if a
player has played exactly half of the matches and skipped the other
half, because this is the situation that gives us most certainty about
his contribution.
--
Cheers
milivella
milivella
> One of the greatest discoveries
> of sabermetrics (due to Voros McCracken, winner of the Euro 2004 RSS
> Sophisticated Prediction Contest) is that pitchers have actually very little
> influence on whether balls hit in play become base hits or outs. One of the
> main lines of his argument is that there is essentially no correlation in a
> pitcher's ability to prevent hits across years. Essentially, it's all due to
> luck.
Things *could* be different for our stat. E.g. Kaka's contribution:
Season Rough ps.-Bayes
2004-05 +1.01 +0.27
2005-06 +0.82 +0.07
2006-07 +0.33 +0.07
2007-08 +1.31 +0.35
2008-09 +1.65 +0.37
Always positive.
Players with a good season above players with a bad season, some
consistency through the years... This stat could be promising.
Unfortunately zerozerofootball.com doesn't have summaries for previous
seasons (and for EPL they start with 2006-07).
--
Cheers
milivella
Futbolmetrix
>
> Things *could* be different for our stat. E.g. Kaka's contribution:
> Season Rough ps.-Bayes
> 2004-05 +1.01 +0.27
> 2005-06 +0.82 +0.07
> 2006-07 +0.33 +0.07
> 2007-08 +1.31 +0.35
> 2008-09 +1.65 +0.37
> Always positive.
>
> Players with a good season above players with a bad season, some
> consistency through the years... This stat could be promising.
> Unfortunately zerozerofootball.com doesn't have summaries for previous
> seasons (and for EPL they start with 2006-07).
I would get rid of the pseudo-Bayes formula. Too much pseudo, too little
Bayes. It's easier to just get rid of players with less than 10 appearances,
because their contribution can't really be estimated precisely.
Unfortunately, all of this is very suggestive. I'd like to see how the +/-
score correlates with some objective measure of performance (actim index,
yahoo fantasy score, newspapers' player ratings?) for a large sample of
players from a whole championship.
D
milivella
> I would get rid of the pseudo-Bayes formula. Too much pseudo, too little
> Bayes. It's easier to just get rid of players with less than 10 appearances,
> because their contribution can't really be estimated precisely.
OK. Thanks for the hint.
If you're patient enough, can you explain me whether the formula used
by IMDB for their rankig of movies
http://www.imdb.com/chart/top
is similar to mine (and, if so, whether it works in that context) or
not? Thanks.
Here is IDB formula:
"The formula for calculating the Top Rated 250 Titles gives a true
Bayesian estimate:
weighted rating (WR) = (v ÷ (v+m)) × R + (m ÷ (v+m)) × C
where:
R = average for the movie (mean) = (Rating)
v = number of votes for the movie = (votes)
m = minimum votes required to be listed in the Top 250 (currently
1300)
C = the mean vote across the whole report (currently 6.8)"
> Unfortunately, all of this is very suggestive. I'd like to see how the +/-
> score correlates with some objective measure of performance (actim index,
> yahoo fantasy score, newspapers' player ratings?) for a large sample of
> players from a whole championship.
"Large" meaning how many players? :/
--
Cheers
milivella
milivella
> I would get rid of the pseudo-Bayes formula. Too much pseudo, too little
> Bayes. It's easier to just get rid of players with less than 10 appearances,
> because their contribution can't really be estimated precisely.
So here is the 2008-09 "rough plus-minus" of some players who has
between 10 and 28 (34 for Bundesliga) appearances:
Arshavin +1.37
Diego +1.11
Iniesta +0.99
Sneijder +0.87
Puyol +0.60
Fabregas +0.52
Fletcher +0.47
Robben +0.29
Adebayor +0.19
Mascherano +0.10
Torres -0.03
Carrick -0.03
Pirlo -0.24
Mutu -0.26
Essien -0.35
Buffon -0.48
Ferdinand -0.65
Drogba -0.73
--
Cheers
milivella
milivella
> So here is the 2008-09 "rough plus-minus" of some players who has
> between 10 and 28 (34 for Bundesliga) appearances:
*24* for the Bundesliga (the logic obviously being that if the number
of matches played with or without that player is less than 10, results
are too skewed).
--
Cheers
milivella
milivella
> Unfortunately, all of this is very suggestive. I'd like to see how the +/-
> score correlates with some objective measure of performance (actim index,
> yahoo fantasy score, newspapers' player ratings?) for a large sample of
> players from a whole championship.
Here is just a very little sample, that is maybe enough to cool us
down. :(
Best and worst player (in the range 10-28 appearances) for each
position, according to La Gazzetta dello Sport's average rating:
Name Rat. +/-
GK
Abbiati 6.46 -0.38
Arthur 5.63 +0.53
DF
Samuel 6.38 +0.32
Lanna 5.28 0.00
MF
Gattuso 6.59 +0.12
Quaresma 5.45 -0.23
FW
Mutu 6.42 -0.26
Cerci 5.45 -1.05
--
Cheers
milivella