xG Discussion

[davieG]

Put it here!

[nnfox]

Can someone explain exactly what xG is?  

[Ric Flair]

Is this yet another new iphone, Apple have released?

[Jobyfox]

5 minutes ago, nnfox said:

Can someone explain exactly what xG is?  

It’s something I put in my development plan at work and never achieve 

[Dahnsouff]

For each scoring opporunity in the game

- Given the position from it was taken

- The type of action (shot, header)

- The nature of how the ball was received

- Proximity of opposition players

 

You can gauge how historically likely are you to score

 

 

 

 

[Dahnsouff]

Or (possibly) more factually, as Opta states

 

[Raw Dykes]

2 minutes ago, nnfox said:

Can someone explain exactly what xG is?  

xG is a measure of how many goals a team is expected to score based on the number and quality of the chances it created. Let's say a team gets an xG rating of 2.5 in a match. They may have created 3 clear cut, one-on-ones. They may have had 25 hopeful long-range shots. Based on the probability of each chance ending up with a goal, on average, a team would end up scoring 2.5 times with the chances they made.

 

If they actually scored more than 2.5 goals in that match, then the scorers did well with little to work with. If they scored fewer than 2.5 goals, then they were wasteful.

 

xG assigns a rating to each type of chance, based on the likelihood of it being converted. A penalty might be around 0.8, for example. An attempted lob from the halfway line might be around 0.05 or something.

 

It's basically a more useful version of shots taken/shots on target stats.

[martyn]

It's an interesting stat, and has showed previously it's a reasonably reliable indicator that a teams form will drop off if their results are exceeding xG stats.

 

Would be really interesting if they could start filtering in quality of players involved in the chance (both outfield player, and goalkeeper).

[stripeyfox]

It is the method used by Spurs fans to demonstrate how they rightfully won the PL in 2016.

 

[Fox92]

I don't get it. Why do we see this on programmes like MOTD now. They weren't goals so it's irrelevant.

[Nicolo Barella]

If we look at our xG so far, it doesn't really tell us anything we don't know - we've been unsustaimably clinical and need to up our chance creation. Now, obviously Vardy scores on low probability all the time, that's what makes him so good, but increasing chance creation will ease our reliance on him and improve his performance too.

[Aus Fox]

5 minutes ago, martyn said:

It's an interesting stat, and has showed previously it's a reasonably reliable indicator that a teams form will drop off if their results are exceeding xG stats.

 

Would be really interesting if they could start filtering in quality of players involved in the chance (both outfield player, and goalkeeper).

Can you give any examples other than Manchester United, for whom we didn’t require Xg to tell us they were about to fall quickly.

[Woof]

A point that has to be born in mind with this measure is that in Statistics, "expected" simply means average.  So as said above it measures how many goals on average would be scored from the positions created in a game.  It does not take into account the individual a chance falls to, so the best strikers (like JV) consistently outscore xG and vice versa.

 

A very detailed data set is available for every game played in the premier league and this goes back some years.  The way the measure is calculated has got more sophisticated over the years taking into account more and more factors and there are now several different versions of it and continuing debate over which of these is best and over ways of improving it.  I'd say it is a useful indicator, especially when considered over a long run of games but of course like all statistical measures has to be interpreted with caution.

[egg_fried_rice]

6 minutes ago, Fox92 said:

I don't get it. Why do we see this on programmes like MOTD now. They weren't goals so it's irrelevant.

Same reason you see stats like shots, tackles, corners, fouls, and bookings. To provide additional context.

[messerschmitt]

Given that players aren't taken into consideration, a penalty will be the same Xg for any team in any game? what about a corner. does it have an Xg before its taken or when anyone heads or kicks it towards goal? Where does the Xg start being judged from, If a player gets played through from forty yards and runs towards the goal is the XG when he gets the ball or when he shoots?

[egg_fried_rice]

12 minutes ago, martyn said:

It's an interesting stat, and has showed previously it's a reasonably reliable indicator that a teams form will drop off if their results are exceeding xG stats.

 

Would be really interesting if they could start filtering in quality of players involved in the chance (both outfield player, and goalkeeper).

I don't think they should. In fact, it's the gap between expected goals/saves and actual goals/saves that indicate the quality of a player. Players like Vardy & Aguero consistently outperform xG - this tells us they are excellent finishers which is evident to everyone who watches them.

[egg_fried_rice]

2 minutes ago, messerschmitt said:

Given that players aren't taken into consideration, a penalty will be the same Xg for any team in any game? what about a corner. does it have an Xg before its taken or when anyone heads or kicks it towards goal? Where does the Xg start being judged from, If a player gets played through from forty yards and runs towards the goal is the XG when he gets the ball or when he shoots?

Yes, penalties are the same and are not included in the normal xG stats. Corners will have a statistical number on the likelihood of being scored before they are taken (it's very low iirc), but the xG is taken from where the player meets the ball once it arrives. xG is judged from where and when the shot is taken.

[filbertway]

Big fan of xG, much more useful than any other stat related to shooting out there.

[Poznan34]

 

 

 

[yorkie1999]

So, it's really a tool for gamblers, or nerds.

[egg_fried_rice]

The fella in that video is Duncan Alexander - well worth following on twitter if you're a fan of stats ranging from insightful to bizarre.

 

 

 

[Poznan34]

Nerds like Stoke City manager Nathan Jones 

 

 

 

[daventry_fox]

Why are penalties the same? Say for example you have been awarded a penalty. On the pitch you have (prime) Matt le tissier with a record of 48/49

And Michael Owen (prime) with a record of 2/8 who do you want taking the penalty? XG is all about historics so surely le tissier should have higher xG from the spot than Owen. And surely you would pick your penalty taker based on this 

[egg_fried_rice]

Just now, daventry_fox said:

Why are penalties the same? Say for example you have been awarded a penalty. On the pitch you have (prime) Matt le tissier with a record of 48/49

And Michael Owen (prime) with a record of 2/8 who do you want taking the penalty? XG is all about historics so surely le tissier should have higher xG from the spot than Owen. And surely you would pick your penalty taker based on this 

Because it's about the quality of chance, not the quality of player taking it. xG is about averages.

 

It's not really a useful stat for penalties, but you would pick Tis over Owen because he would outperform the xG on penalties whereas Owen would under perform based on the figure you provided.

 

It's more about measuring the quality of chances created in open play. If you have a high xG and a low amount of goals it tells you your finishing is poor and vice versa. Obviously anyone can tell you a pot shot from 30 yards is less likely to go in than a one on one inside the box, but it helps in statistical mapping and analysis. It also helps you look at an individual player's strengths and weaknesses.

 

[Guest Cujek]

its a load of bollocks.