Hipsters and their xG 'expected goals' calculus

[Javi]

What is this all about and what is there to learn from this?

 

[VP89]

It's not hipster. I think it's meant to be more useful than just looking at shots on goal by weighting how good a chance it actually is.

 

[pacifictheme]

What is this all about and what is there to learn from this?

Jeff?

 

[Javi]

It's not hipster. I think it's meant to be more useful than just looking at shots on goal by weighting how good a chance it actually is.

So people rate a chance by their own judgement or is there another metric available? If so the mathematics seem misleading.

 

[acnumber9]

They needed to move away from talking about 6's, 8's, 8.5's and double pivots. How would others know they're football experts if they didn't talk about stuff like this?

 

[ROFLUTION]

How accurate is this xG stuff over a whole season?

 

[SqualorVictoria]

It's not hipster. I think it's meant to be more useful than just looking at shots on goal by weighting how good a chance it actually is.

Basically this. Even the biggest critics should at least accept that has a much better information value about the game than having a look at shots / shots on target.

 

[VP89]

So people rate a chance by their own judgement or is there another metric available? If so the mathematics seem misleading.

The expected goal value (EGV) drives the XG stat. It's calculated by looking at proximity to goal, how many defenders are around the player and so on.

 

[Invictus]

Expected Goals is computed to turn each goal scoring attempt into a number between 0 and 1, representing the odds of that attempt producing a goal.

The easiest attempt to explain is a penalty.

Typically, penalties are awarded around 0.75 ExpG, based on historic conversion rate. A penalty is the easiest attempt to classify, since it’s a situation isolated from play, with a standard spot for taking it.

You can expand the number of situations and run separate regression models for other situations such as open play shots, headers, direct free-kicks, corners, throw-ins, fast breaks, rebounds from a keeper save, rebounds from the woodwork etc.

Essentially you can use a database (Opta) that will allow you hundreds of thousands of simulations to come up with an Expected Goal value for each situation.

And in a low scoring game like football, you can have a much xG than the opponent and not win. That's why in turn statistical models are not necessarily accurate predictors of match outcomes.

https://www.redcafe.net/threads/xg-...l-spinoff-thread-ft-babu-of-rawk-fame.432919/

*LOCK*