GoalImpact – Measuring Players by Goal Difference

On our travels through the Twittersphere we stumbled across a site called GoalImpact.com. It was written in German and had intriguing numbers attached to football players, so immediately we were curious. A quick Google translation later and we were hooked. The site is run by a chap called Jörg Siedel. We’ll leave it to him to explain:

“Everybody has their favourite players – the players they think add most to their team’s success. But this is entirely subjective. Increasingly, statistics are being used to objectify these assessments. Most of these statistics are bottom-up. People count the amount of completed passes, successful tackles, shots on target and many more things. But, while those statistics help to understand the game, I’d argue it that it’s difficult use them to measure the end product of a player on the field. The game of football is too complex to create a complete bottom-up model for it. We cannot easily interpret these kind of statistics to rank player performance without adding at least some subjectivity.

To avoid this, I try to measure a players performance top-down. I simply measure the player’s team’s goal difference when the player is on the field. I don’t look at how he effects his team’s goal difference.

In this raw definition, it would be a very imprecise measure as the team’s performance certainly rests on more factors than just the relative strength of an individual player. His performance also depends on:

  • the players alongside him
  • the strength of the opponent
  • whether he has home advantage
  • red cards shown to either side
  • luck

The first four I can correct explicitly, but luck can only be accounted for by the regression toward the mean and by averaging over many games. By doing these calculations, I obtain an individual score for each player that I call GoalImpact. By definition, GoalImpact is a universal number whereby a defender can be as valuable as an attacker. To make the figure easily understandable, I transformed it to a scale in which the average player has a GoalImpact of 100.”

Differentgame asked Jörg to pick the optimal England squad of players still playing, based on GoalImpact. Here’s what it looks like:


Now to us, that looks a snapshot of the national side over the last decade.

If you take out the “unrealistic” names – those now considered too old or no longer worthy (Scholes, J Cole, Beckham, Ferdinand, Neville, Carragher) the names Jörg replaces them with also look familiar. The next top rated players are the likes of Smalling, Wilshere and Oxlade-Chamberlain.

On this evidence it looks very much like the England squad has and continues to be picked purely on player outcomes than process. That the “best” players have been thrown into the mix with the hope that it gels. It’s worked to an extent, as numerous journeys past the group stages will testify. But the English FA routinely spend big money on luring manager after manager to do the same thing. Jörg, could have run it through his computer in minutes, and he’d probably have done it for free if you asked him nicely. On top of which, he speaks better English than Hodgson, never mind Capello.

If the average GoalImpact player is rated at 100, who’s the best? Cristiano Ronaldo at around 190. Next best is Leo Messi at around 188. So the world’s two greatest players actually do make a huge difference to their team.

We can’t resist asking Jörg what Gareth Bale’s GoalImpact score is. It’s 118. “Bale is not the best player at Spurs at all”, says Jörg. “Despite his talents, his team fails to perform significantly better when he is on the field.”

It looks like  Bale will have to consistently perform for more than a dozen games before he’s mentioned in the same breath as those two.

If you like Jörg’s analysis, you should check out GoalImpact.com and follow him on Twitter and Facebook.

This entry was posted in Sports and tagged , , , , , , , . Bookmark the permalink.

7 Responses to GoalImpact – Measuring Players by Goal Difference

  1. Pingback: GoalImpact – Measuring Players by Goal Difference differentgame | Lync News

  2. bart says:

    this is neat … I’ve been keeping +/- numbers and minutes on the field for the players of a team I follow and have been scratching my head over how to put them into a useful stat … I think I’ll e-mail Mr Siedel to see what else I need in order to come to something like this. Thanks for the heads up on his site and for writing this post!

    what instantly popped into my head was the following … In Holland there are a number of respected journalists who say the following about Luis Suárez (of Liverpool fame, and former Ajax fame) … “you can’t win anything with him on your team” … they think he’s an exciting (and controversial) player who will bring a team lots and lots of goals, maybe even end up being league top scorer, but in the end they feel that he is actually detrimental to the whole team (i.e. the team could score more goals together – and win more – if he wasn’t on the field over an entire season) who are trying to get a league championship.
    He played for Ajax 3,5 seasons (and one at Groningen) scoring a ton of goals … Ajax one became champion in one of those … the last, where he left during the winter transfer window to Liverpool and Ajax won the league on the last day of the season after climbing up the tables at a steady rate (were quite low when he left – for their doing). Oh, Ajax are expected to be in the running to win a title each year. Not winning with the technically most gifted player in the league on the team is strange to say the least (this at a time when no-name teams won the league, e.g. FC Twente, AZ).
    Now, I wonder if this stat, GoalImpact, could indicate if what the journalists think may have a grain of truth. Meaning that Suárez would best be suited for a mid-table team challenging for EL and CL places and not the championship.
    These same journalists said the same about Jerrel (Jimmy Floyd) Hasselbaink (of Chelsea fame).

    thus … how do I quantify what they see so that I can also see it? I only see a gifted player, who scores some crackers … seeing how he affects the team is difficult, as I’m not that game savvy.

    anyway, food for thought really. Again, thanks for the tip, going to read Mr Siedel’s blog now.

  3. Thanks bart

    Am sure jorg will welcome your interest

    I will be comparing all strikers by SPAM at the end of the season

    Also @willtgm has done some analytic work around suarez if you want to check it out

  4. Goalimpact says:

    Hi Bart,

    thanks for your comments and interest in my work. I fully agree that even a striker scoring many goals may be detrimental for the team. My theory here is, that a high number of goals and assists for a single player might also mean that the team is too much centering its efforts on this player. They might play more diverse without him and be more successful.

    Luis Suarez has a Goalimpact of 138.3. That is not bad at all, so he certainly adds to a positive expectation of the team’s goal difference. I’m not very familiar with Suarez, but if you want to win a title, you need very good defense. So, if Suarez adds 0.5 scored goals per game and 0.4 conceded goals, he may actually lower the chance to win a title as compared to a player that adds 0.1 scored goals and minus 0 conceded goals. However, their Goalimpact will be identical. *If* this is the case for Suarez, he would be more suitable for team that is ranked in the mid or lower end of the league. For those teams he would be a real asset. Top teams may be better of with a striker that knows how to play defensive (e.g. pressing), too. Reminder: that is a *if*.


  5. bart says:

    @ both
    Thanks for your replies.
    Would there be any way of combining James Grayson’s Player PAR (Points Above Replacement) and GI to give a type of WAR (Wins Above Replacement) stat as baseball has? e.g. GPAR (Goals & Points Above Replacement)

    @Differentgame … looking forward to the post.

    @Jörg … agree and have sent you an e-mail.


  6. Pingback: The GoalImpact England Team | differentgame

Leave a Reply

Fill in your details below or click an icon to log in:

WordPress.com Logo

You are commenting using your WordPress.com account. Log Out /  Change )

Twitter picture

You are commenting using your Twitter account. Log Out /  Change )

Facebook photo

You are commenting using your Facebook account. Log Out /  Change )

Connecting to %s