Study appoints Robinson worst player of the season

[The Jitty Slitter]

http://www.timesonline.co.uk/tol/sport/football/fink_tank/article1867878.ece


The full rankings are here

By position
http://www.timesonline.co.uk/tol/sport/football/fink_tank/article1852016.ece

Ranked

http://www.timesonline.co.uk/tol/sport/football/fink_tank/article1852016.ece

[United Pumps]

HOW IT WORKS
The Fink Tank Predictor provides forecasts and ranking systems for English and European club football, based on a statistical model of matches based on more than five years of football scores.
In looking at player rankings for this season, the phrase “time-adjusted points” means the number of points the player would have added to an average team in the full season, compared with an average replacement. The points are then adjusted to reflect the amount of time spent on the pitch – minimum 400 minutes.


wut?

[L.G.S]

Just shows too many stats cloud how the player actually was for his team.

Essien sixteenth, probably the best midfielder this season?

Drogba 19th (one off his Prem goal tally), when he's been scoring, assisting, defending and every other possible thing?

[Derrida]

I've read Finkelstein's stuff with a lot of interest. The work they've been doing to try and generate some statistical basis to assess players is very interesting, and reminiscent of folks like Bill James and his progeny in the study of baseball (my other favorite sport), but there are a number of rather significant flaws with his work that make me doubt its real utility (at least as currently constructed) for football.

First, the formulas are hidden or at least only vaguely explained. Not that I have the statistical fortitude to comment one way or another on them, but part of what has made sabermetrics (statistical analysis of baseball) work is the fact that a LOT of people do similar work on subjects and fact-check each other.

Second, Finkelstein writes with a level of certainty that is simply impossible. Even in baseball, where offensive statistics are incredibly easy to compute and analyze, there is a lot of disagreement about who is best and for what reasons. There is more trouble analyzing pitching and FAR more with defense because of the additional layers of complexity that stem from having a group dynamic. In a sport like football where the team dynamic guides every single action it is simply impossible to make statistical judgments that remove context, which is the whole premise of their project. No football player exists in a vacuum, or on a field with 10 "average" players. They have to deal with the particular context they are placed in. It's not station-to-station action like baseball - it's non-stop movement and endlessly complicated chains of action.

That said, I do not mean to deny that there is some ability to measure who did and did not contribute to the goals which actually happened. If their project is imply a backward-looking assessment of what happened, it can probably be pretty accurate. But that is only marginally interesting if it fails to give us an idea about how they will be likely to perform in the future.

Third, a more minor complaint, but the way they calculate their numbers doesn't make sense. They judge players against the average, but then multiply that by playing time. That means that if you're slightly below average and play every minute, you LOSE a lot of points. Now, I'm not sure what reason they have to believe that multiplication will produce a meaningful statistic, but more importantly, this ignores a very important factor: an average player is relatively hard to get. By definition, half the players are worse (actually far more because every team has plenty of guys on the bench who are even worse than their starters).

Judging a player vs. the average is ONE useful metric, but it is almost counterproductive if you don't also judge them vs. a "replacement level" player. One of the most important developments in baseball statistical evaluation in recent decades is precisely that concept: value-over-replacement. If you remove your average player, you don't get to replace him with another average guy - you have to replace him with whoever you can get. When comparing superstars, value-over-average is probably the best metric, but when evaluating other players, VORP is far more useful.

Interesting stuff, nonetheless. I would treat these numbers as one factor in assessing players rather than giving them authority over more subjective judgments. And this comes from someone who has been obsessed with rooting out the parochial Conventional Wisdom in baseball for over a decade. There may come a day when we develop computers sophisticated enough to beat human intuition and pattern-assessment, but I have a very hard time believing we're there yet in football.

[The Jitty Slitter]

Fink. is merely the columnist. The work is done by http://www.dectech.org/

[lanman]

The whole report is absolute rubbish. If anyone believes that Bolton would be significantly better off without Kevin Davies, Wigan without Baines, Charlton without Young or Blackburn without Bentley (each were designated their clubs "worst" players, costing them significant points) then they really don't understand the game. You cannot quantify a players contribution to a team game by using statistics.

[RichardL]

You cannot quantify a players contribution to a team game by using statistics.you can (goals scored is a pretty good one for a striker, for example) but just not these statistics.

Statistics are a valid measurement, but you have to be able to work out what they are a measurement of, or they are useless.

[lanman]

you can (goals scored is a pretty good one for a striker, for example) but just not these statistics.

Statistics are a valid measurement, but you have to be able to work out what they are a measurement of, or they are useless.

Not really. Goals scored, for example, is not truly reflective of a players ability to score as it does not take into account how easy/difficult the chance created for the striker is or the number of attempts on goal a player has or the quality of opposition scored against.

Football is a team game and the only statistics that matter are those avchieved by the team - games won, drawn & lost and goals scored and conceded.

[The Jitty Slitter]

Not really. Goals scored, for example, is not truly reflective of a players ability to score

wow.

[AussieLFCfan]

Any list that rates Fat Frank as the second best player last season in the EPL gets NO recognition or respect from me.

Ridiculous! Honestly, Fat Frank! For Fcuks Sake!:confused:

[Derrida]

you can (goals scored is a pretty good one for a striker, for example) but just not these statistics.

Statistics are a valid measurement, but you have to be able to work out what they are a measurement of, or they are useless.

This is the vital point. This work provides a lot of information about what happened on the pitch, who contributed and who didn't, but it can only tell us as much as we have the ability to discern from it.

While I share the skepticism people have about their results, I think the solution is to focus on the places where they seem most at odds with our subjective perspectives and try to understand what factors contribute to that. It can shed some light on both sides. There are without a doubt problems in their methodology (perhaps some irrevocable problems due to the nature of the game) and this could help figure out what they are. But access to some objective results, produced without the various biases that go into normal evaluation, might give us an idea about the kind of player who is overrated by traditional metrics

Because those subjective biases are real and just as impossible to resolve as the fact that football is composed of endlessly flowing play, making statistical evaluation difficult. For one, human beings simply cannot watch enough players to really see the differences in everyone. There just is not enough time in the day. Even if you watched every Premiership game, you still would have 22 players at a time to track - it's just not possible. Beyond that, everyone has favorites and people they don't like. They support some sides and hate others. They believe that there are some ways of playing that are more "right" than others (defensive, attacking, 4-4-2, 4-3-3, how often should fullbacks move forward, etc. and so on).

Human beings can't help but be subjective (and watching football wouldn't be nearly as fun if this wasn't true) so it can't hurt to add in an objective measure as a counterbalance, as long as we recognize that it is not the "true" assessment, but rather one more piece of the puzzle.

[chrizzah]

These statistical analyses are slightly interesting, but largely irrelevant. Unless you are able to weigh the quality of each pass, save, shot, etc.., these numbers are always going to come up with some ridiculous outcomes that will lead certain BigSoccer participants to make outlandish claims of brilliance for rather pedestrian players and vice versa.

This whole number and quantifying business is one reason I don't think soccer can ever proceed past a certain level of popularity as a spectator sport in the United States. American sports conversations often focus around numbers and soccer just doesn't lend itself to simple quantification like baseball, American football or baseball.

[The Jitty Slitter]

These statistical analyses are slightly interesting, but largely irrelevant.


actually this is incorrect - as their model does have proven predictive value.

So there is truth in it over the long run, but obviously that truth tends to break down in individual games.

[The Jitty Slitter]

This is the vital point. This work provides a lot of information about what happened on the pitch, who contributed and who didn't, but it can only tell us as much as we have the ability to discern from it.

While I share the skepticism people have about their results, I think the solution is to focus on the places where they seem most at odds with our subjective perspectives and try to understand what factors contribute to that. It can shed some light on both sides. There are without a doubt problems in their methodology (perhaps some irrevocable problems due to the nature of the game) and this could help figure out what they are. But access to some objective results, produced without the various biases that go into normal evaluation, might give us an idea about the kind of player who is overrated by traditional metrics

Because those subjective biases are real and just as impossible to resolve as the fact that football is composed of endlessly flowing play, making statistical evaluation difficult. For one, human beings simply cannot watch enough players to really see the differences in everyone. There just is not enough time in the day. Even if you watched every Premiership game, you still would have 22 players at a time to track - it's just not possible. Beyond that, everyone has favorites and people they don't like. They support some sides and hate others. They believe that there are some ways of playing that are more "right" than others (defensive, attacking, 4-4-2, 4-3-3, how often should fullbacks move forward, etc. and so on).

Human beings can't help but be subjective (and watching football wouldn't be nearly as fun if this wasn't true) so it can't hurt to add in an objective measure as a counterbalance, as long as we recognize that it is not the "true" assessment, but rather one more piece of the puzzle.

Agreed.

Interesting that most people's response is simply to scoff based on "what they know" which of course is not really a mathematically valid argument ;)

[Ballpoint]

Nice intelligent thread. (As a newbie, for what it's worth, I must say it sure beats the hell out of 99% of what's on BBC 606.)
There are so many variables and imponderables here. For example, you could evaluate how good a game a defender had by the number of successful tackles he made (among other factors of course) but you'd obviously have to distinguish between a tackle in midfield and a goal-saving tackle in the penalty area. Presumably, fouls committed would statistically count against a player but the reality of football is that fouls sometimes help prevent what would have been a goal (when the resulting free kick is unsuccessful and even in the penalty area if the spot kick is missed) so how do you evaluate that?

[The Jitty Slitter]

Nice intelligent thread. (As a newbie, for what it's worth, I must say it sure beats the hell out of 99% of what's on BBC 606.)
There are so many variables and imponderables here. For example, you could evaluate how good a game a defender had by the number of successful tackles he made (among other factors of course) but you'd obviously have to distinguish between a tackle in midfield and a goal-saving tackle in the penalty area. Presumably, fouls committed would statistically count against a player but the reality of football is that fouls sometimes help prevent what would have been a goal (when the resulting free kick is unsuccessful and even in the penalty area if the spot kick is missed) so how do you evaluate that?

I think a key point is that these are long run studies. So two things fall out of that...

1. One off events become less significant

2. You see what factors are highly predictive.

Of course many factors aren't taken into account

[nicephoras]

wow.

If you have seven chances at goal and convert four are you telling me you're a worse player than someone who has 30 chances and converts 5? While getting chances is a skill in it of itself, there's a limit to how many chances much a pure striker is going to carve out for himself.

[The Jitty Slitter]

If you have seven chances at goal and convert four are you telling me you're a worse player than someone who has 30 chances and converts 5? While getting chances is a skill in it of itself, there's a limit to how many chances much a pure striker is going to carve out for himself.

No - I am saying that this statement is clearly incorrect. Goals scored is of course highly reflective of ability to score over a run of games.

Goals scored, for example, is not truly reflective of a players ability to score

[lanman]

No - I am saying that this statement is clearly incorrect. Goals scored is of course highly reflective of ability to score over a run of games.

So a player 20 goals from 80 chances is a better goalscorer than a player with 15 goals from 30 chances?

Shot conversion is a far better indication of a players ability to score, but even then it does not take into account the difficulty of the chances.

[lanman]

I should also say that quality of opposition should also be factored in. If you just take aggregate goalscoring as a guida then Ali Daie is the best goalscorer to play international football.