Xg from WBA game

What is meant by assist type?

Does it cover the difference between first time finishes and those when those where the ball is under control, for example?

The infamous Ronnie Rosenthal miss remains the worst I've seen imho, as he had time to control the ball once past the keeper.

That, to me, makes it worse than any number of first time efforts when the ball is flashing across the 6 yard box, even if they closer in with no one to beat.
Assist type would be through ball, cross, rebound for example.

And no to the first time shot, v controlling touch. That is handled, in theory, by the proximity of opposing players.
 
People will just quote statistics off various websites. I am very sceptical of their validity. There are a number of different xg providers. We don't know what variables they are using to calculate the xg. Ive read some are just based on where the player is when he takes the shot.
I'd also bet that companies like the BBC and sky will be using the very cheapest stats they can buy.

Xg is also supposedly taken from thousands of games. Applying it to one game is worthless. In thousands of games the bumps and anomalies get smoothed out.

The human brain is a marvelous thing and is more than capable of watching a football match and determining who had the better chances. The biggest problem we have is bias.

I certainly thought west brom had a decent number of chances. xg also doesn't take into account the ball flashed across the goalie, or the simple pass they failed to make to put a man through. On another day we could have been beat. Just like on another day we would beat derby and the mackams. That's football.
 
I tend to look at goal attempts as a rough guide of who was best etc.

I think we have had more attempts than the opposition in every league game this season, possibly they only one that wasn't was Cardiff away.

We should have picked up more points than we have.
 
I think most people misunderstand statistics to be honest and certainly misunderstand probability and its attempt to quantify uncertainty. As soon as you attribute a number to something, the vast majority of people think that is an ‘answer’ and either take it as gospel or, alternatively, try and prove it to be wrong.

xG does what it does. It shows the likely probability each of the shots taken in a match resulting in a goal. Each shot (and its goal probability) is independent from every other shot, which is why you get different results from both an initial shot and any rebound attempt.

If you take Maja’s effort last night, the stats must show that when a cross is met by a forward in that position, it results in a goal 83% of the time. Of course, that means that it doesn’t result in a goal 17% of the time.

Some of that 17% will be due to incredibly bad finishing or unbelievable saves/defending, but some of it will also be due to situations such as yesterday, when the forward is off balance, stretching and barely getting a touch to the ball.

My only problem with xG is that it’s not really mathematically sound to simply add those probabilities together and then use the result to say this is how many goals each team should have scored. That’s not how probability works. It’s not how goals work either.
 
My only problem with xG is that it’s not really mathematically sound to simply add those probabilities together and then use the result to say this is how many goals each team should have scored. That’s not how probability works. It’s not how goals work either.
Indeed and particularly when there are multiple shots in one attack phase involved, as with the best will in the world, you can't score both the initial shot and a follow up from a rebound / save / block.
 
My only problem with xG is that it’s not really mathematically sound to simply add those probabilities together and then use the result to say this is how many goals each team should have scored. That’s not how probability works. It’s not how goals work either.
You're right about that, and football is one of those odd occasions where the outcome of one event can change the following events, and even prevent them from happening altogether.

If a team scores and goes 1-0 up then they might take less risk and have less chances etc, so xG after scoring would probably be lower, purely down to there being less shots.
 
You're right about that, and football is one of those odd occasions where the outcome of one event can change the following events, and even prevent them from happening altogether.

If a team scores and goes 1-0 up then they might take less risk and have less chances etc, so xG after scoring would probably be lower, purely down to there being less shots.
Billy is saying a bit more than that.

In a game I picked randomly from last night salzburg v brest(OK it wasn't random). The score was 1-4 with xG of 1.62 and 2.02 brest did much better than their xg. Some days they will do much worse.. But over a season, if the stats any good it should be quite close to goals scored.

Boro last season xg 69,scored 70 goals. Let's look at another example of last night's opponents, West Brom last season xg 61 goals 70.


Looking at a single game xg is a blunt instrument as are all stats. Over a long enough period teams will, generally revert to mean.
 
Billy is saying a bit more than that.

In a game I picked randomly from last night salzburg v brest(OK it wasn't random). The score was 1-4 with xG of 1.62 and 2.02 brest did much better than their xg. Some days they will do much worse.. But over a season, if the stats any good it should be quite close to goals scored.

Boro last season xg 69,scored 70 goals. Let's look at another example of last night's opponents, West Brom last season xg 61 goals 70.


Looking at a single game xg is a blunt instrument as are all stats. Over a long enough period teams will, generally revert to mean.
Oh yeah, I get that, that was a given. It's just less variance with more stats. A single game isn't expected to be very accurate, but could be reviewed quite easily to see how it compares to xG.

I wouldn't trust it just by looking at stats, for a single game, but if I've seen the game it's easier to get a grasp of how close to reality it was, if you can remove personal bias.
 
If you can put up with the fact that this guy’s American, here’s a decent explanation of why xG doesn’t really work for an individual match.

 
I think xG does work for a single match. It tells you how many goals you would expect to score from your chances but it doesn't give an indication of what the standard deviation would be. You shouldn't expect to score the same amount as your xG in every game. Each game will have a variation where sometimes you score more and sometimes you score fewer.

If you have 6 identical shots with a total xG of 1 that would be the same as rolling a die 6 times trying to get a 6. Some days you could get 6 sixes and some days you could get 0 sixes. It's the same with football. Some days everything goes in and some days nothing does.

Over a larger number of games, like a season, you should expect to get a similar number to your total xG over the full sample size.
 
I think xG does work for a single match. It tells you how many goals you would expect to score from your chances but it doesn't give an indication of what the standard deviation would be. You shouldn't expect to score the same amount as your xG in every game. Each game will have a variation where sometimes you score more and sometimes you score fewer.

If you have 6 identical shots with a total xG of 1 that would be the same as rolling a die 6 times trying to get a 6. Some days you could get 6 sixes and some days you could get 0 sixes. It's the same with football. Some days everything goes in and some days nothing does.

Over a larger number of games, like a season, you should expect to get a similar number to your total xG over the full sample size.
If the xg model says that from X place in the box there is a 90% chance of scoring this is based on various factors evening out over a long duration.

Let's say in our single game the team is playing very defensively and everytime the attacker shoots from that position in reality he has next to no chance of scoring as there are five defenders in the way. Yet the xg will be very high as usually there will not be so many defenders in the way.

You are better off turning to the bloke next to you and saying, which team do you think has had more chances.
 
If the xg model says that from X place in the box there is a 90% chance of scoring this is based on various factors evening out over a long duration.

Let's say in our single game the team is playing very defensively and everytime the attacker shoots from that position in reality he has next to no chance of scoring as there are five defenders in the way. Yet the xg will be very high as usually there will not be so many defenders in the way.

You are better off turning to the bloke next to you and saying, which team do you think has had more chances.
That wouldn't be an xG of 0.9 because of the position of the defenders.
 
I think xG does work for a single match. It tells you how many goals you would expect to score from your chances but it doesn't give an indication of what the standard deviation would be. You shouldn't expect to score the same amount as your xG in every game. Each game will have a variation where sometimes you score more and sometimes you score fewer.

If you have 6 identical shots with a total xG of 1 that would be the same as rolling a die 6 times trying to get a 6. Some days you could get 6 sixes and some days you could get 0 sixes. It's the same with football. Some days everything goes in and some days nothing does.

Over a larger number of games, like a season, you should expect to get a similar number to your total xG over the full sample size.
Don't forget xG is based on the average player not an individual player. So some team will constantly score above their xG as they have players capable of outscoring the average players. Some teams will score below their xG because their players aren't as capable.

Man City comfortably outscore their xG each season.
Everton struggle to meet their xG

It's not all probability, it's also down to ability, and to some extent the quality of the GK a team on facing
 
Back
Top