New striker?

Sadly I'm not sure even that's true.

As far as I'm aware xG has a problem comparing players who play in possession based sides with those who counter attack.

Some xG calcs take into account the number of defenders in the area (although I don't think they yet map exact positions), and others try and add multipliers to account for fast breaks, but I don't believe any gets close to being perfect.

This is important for City, as we tend to play against teams that will not only have defenders packed in the box, but they will be about as solid as a defence can be. They'll spend long periods in their exact positions, knowing where there team mates are, and as a result are hard to break down (hence why so much of City's probing play is trying to draw people out of position). Now consider how teams defend against counter attacks - how many times do we see 5 or 6 defenders back in the box but still an easy goal? It's all about the players being off balance, not in a set position, facing the wrong way etc. (Liverpool for example have been exceeding their xG by miles under Klopp, because, while they do have a lot of possession, they are more open, and very good at creating chaotic situations - exactly the kind that xG struggles with).

So, the same shot from a City player - say on the penalty spot - is often likely to much more difficult than it would be for a typical Spurs or Dortmund player (not sure why I chose those teams). The more complex xG calculators will try and compensate, but they can't get it perfect.

So, while it might be useful to compare xG between players in the same team, it's very difficult to compare between teams, especially if they have different styles.

I use both understat and fbref for xG.
Both are great in their own way BUT fbref is IMO the most robust. As you are many of the stats on fbref are powered by statsbomb and they are the fastest growing provider selling data analysis to professional clubs.

Here is the description of how they calculate xG

"Very simply, xG (or expected goals) is the probability that a shot will result in a goal based on the characteristics of that shot and the events leading up to it. Some of these characteristics/variables include:

  • Location of shooter: How far was it from the goal and at what angle on the pitch?
  • Body part: Was it a header or off the shooter's foot?
  • Type of pass: Was it from a through ball, cross, set piece, etc?
  • Type of attack: Was it from an established possession? Was it off a rebound? Did the defense have time to get in position? Did it follow a dribble?
Every shot is compared to thousands of shots with similar characteristics to determine the probability that this shot will result in a goal. That probability is the expected goal total. An xG of 0 is a certain miss, while an xG of 1 is a certain goal. An xG of .5 would indicate that if identical shots were attempted 10 times, 5 would be expected to result in a goal.

There are a number of xG models that use similar techniques and variables, which attempt to reach the same conclusion. The model that FBref uses is provided by StatsBomb.
What sets StatsBomb's xG model apart from others is their use of freeze frames. A freeze frame is the location of all players on the pitch at the moment the shot was taken. Was the goalkeeper in position? Was it an open goal or were there a number of defenders between the shooter and the goal? Was the shooter being pressured? Was it a 1v1 situation with the keeper?"

The full article is here:


Based on over performance of xG it is easy to argue that Son and Kane are better finishers than Aguero/Ronaldo because over a reasonable time-frame (understat is best for this to see quickly for the last 6 years or so). In non penalty xG terms Ronaldo has been marginally below each of the last 5 seasons and Kun is marginally ahead. Meanwhile Son and Kane have some big positive numbers.

How to explain this? Kun is the best all-time Premier League striker because consistently he gets off more shots per game on average in the danger areas than anyone else in the PLover the last 10 years via his awareness of space, his strength, his short back lift etc etc. His conversion rate is only just above average BUT thorugh his high shot count he scores more goals. Pure and simple it is a numbers game. With the exception of 12/13 when he was poor, his shot p90 has been in the range of 4.22 to 5.24. His conversion rate per shot has averaged 18.13% with just 3 seasons at 20%+ or 1 goal p 5 shots.

Ronaldo? Average 2+ goals p90 more than Kun. So more goals!
 
I use both understat and fbref for xG.
Both are great in their own way BUT fbref is IMO the most robust. As you are many of the stats on fbref are powered by statsbomb and they are the fastest growing provider selling data analysis to professional clubs.

Here is the description of how they calculate xG

"Very simply, xG (or expected goals) is the probability that a shot will result in a goal based on the characteristics of that shot and the events leading up to it. Some of these characteristics/variables include:

  • Location of shooter: How far was it from the goal and at what angle on the pitch?
  • Body part: Was it a header or off the shooter's foot?
  • Type of pass: Was it from a through ball, cross, set piece, etc?
  • Type of attack: Was it from an established possession? Was it off a rebound? Did the defense have time to get in position? Did it follow a dribble?
Every shot is compared to thousands of shots with similar characteristics to determine the probability that this shot will result in a goal. That probability is the expected goal total. An xG of 0 is a certain miss, while an xG of 1 is a certain goal. An xG of .5 would indicate that if identical shots were attempted 10 times, 5 would be expected to result in a goal.

There are a number of xG models that use similar techniques and variables, which attempt to reach the same conclusion. The model that FBref uses is provided by StatsBomb.
What sets StatsBomb's xG model apart from others is their use of freeze frames. A freeze frame is the location of all players on the pitch at the moment the shot was taken. Was the goalkeeper in position? Was it an open goal or were there a number of defenders between the shooter and the goal? Was the shooter being pressured? Was it a 1v1 situation with the keeper?"

The full article is here:


Based on over performance of xG it is easy to argue that Son and Kane are better finishers than Aguero/Ronaldo because over a reasonable time-frame (understat is best for this to see quickly for the last 6 years or so). In non penalty xG terms Ronaldo has been marginally below each of the last 5 seasons and Kun is marginally ahead. Meanwhile Son and Kane have some big positive numbers.

How to explain this? Kun is the best all-time Premier League striker because consistently he gets off more shots per game on average in the danger areas than anyone else in the PLover the last 10 years via his awareness of space, his strength, his short back lift etc etc. His conversion rate is only just above average BUT thorugh his high shot count he scores more goals. Pure and simple it is a numbers game. With the exception of 12/13 when he was poor, his shot p90 has been in the range of 4.22 to 5.24. His conversion rate per shot has averaged 18.13% with just 3 seasons at 20%+ or 1 goal p 5 shots.

Ronaldo? Average 2+ goals p90 more than Kun. So more goals!

I'd be interested to know how that affects their stats over time (It's a tech that's developing, which is why I caveated it so much).

I read an article from just a couple of years ago which said they were working towards accurate player positions, but even if they perfect it, then isn't it likely historical stats would still be based on older, less sophisticated systems?

It might be interesting to see how those over/under performers become less so as the years progress.

Comparisons of the very best players is so subjective. I don't know if it's true, but not long ago I read that City scored less with Aguero in the team, compared with Jesus. Juve also scored less goals after Ronaldo joined them. While getting off lots of shots is often a sign of a good player, it can also be the sign of a player hogging chances (or in a team with less goal scorers, is the ball being funnelled through to them more often?). When I compared Sterling with Ronaldo and Kane recently someone mentioned that the latter two took more shots from outside the box which had little chance of success (and didn't usually succeed). Not only does that bump up the number of shots they take, but it's actually wasteful, and likely takes chances away from other players.

I find all this stuff fascinating, but seems like the more you know, the less some of it means :)
 
i dont think we will go back for Kane now but would be interested to see what happened if we put that £100m bid we made for Kane into a January one for Haaland? 120m euros ish to Dortmund for a player they lose for 75m euros just months later?
 
I'd be interested to know how that affects their stats over time (It's a tech that's developing, which is why I caveated it so much).

I read an article from just a couple of years ago which said they were working towards accurate player positions, but even if they perfect it, then isn't it likely historical stats would still be based on older, less sophisticated systems?

It might be interesting to see how those over/under performers become less so as the years progress.

Comparisons of the very best players is so subjective. I don't know if it's true, but not long ago I read that City scored less with Aguero in the team, compared with Jesus. Juve also scored less goals after Ronaldo joined them. While getting off lots of shots is often a sign of a good player, it can also be the sign of a player hogging chances (or in a team with less goal scorers, is the ball being funnelled through to them more often?). When I compared Sterling with Ronaldo and Kane recently someone mentioned that the latter two took more shots from outside the box which had little chance of success (and didn't usually succeed). Not only does that bump up the number of shots they take, but it's actually wasteful, and likely takes chances away from other players.

I find all this stuff fascinating, but seems like the more you know, the less some of it means :)

Lots of points there!

FWIW the current advanced Stasbomb xG model was worked backwards for the historic data on FBref. Presumably they can/will do the same thing as the accuracy evolves.

Aguero v Jesus - I'm not sure if true but, if it was, correlation doesn't imply causation because there could be many reasons.

Ronaldo - i heard a pod recently from James Horncastle and the combination of playing everything to suit his requirements coupled with his greed meant the team overall wasn't performing as well. They wanted him out and not just for £££ reasons.

Ronaldo's long range shots do obviously reduce his conversion rate on a goal per shot basis but are reflected in his total NPxG with low probability. However, even if they were deducted his shots p90 stats would still be very high.
The same holds good for Kane albeit a lesser extent.

I've an old chart looking at the goalscoring heavy weights Kane, Salah, Aguero and Sterling for 17/18 - to 19/20 seasons. Over 30% of Kane's shots were from outside the box compared to Salah 23%, Sterling 20% and Aguero 18%.
If hypothetically all of Kane's long range shots were eliminated his total non-penalty goals would still be lower than Sterling and Aguero for those 3 seasons!

Boring to many and probably most on here, but I totally agree it's fascinating. The growth of data available in the public domain has grown enormously in recent times but it only scratches at the surface when drilling down deeper with the paid-for feeds. I'm in a small pro-sports trading syndicate (posh for gambling :) ) and we enlist and pay on an occasional basis for the services of a pro-data analyst to model a multitude of bespoke scenarios to assist in making some dollar. I would argue that the more you know (with context) the more it means :)
 
i dont think we will go back for Kane now but would be interested to see what happened if we put that £100m bid we made for Kane into a January one for Haaland? 120m euros ish to Dortmund for a player they lose for 75m euros just months later?
I couldn’t agree more with this. However, I think it is highly unlikely and would depend on how Dortmund is performing domestically and whether they progress in the CL. In theory, it makes a lot of sense for Dortmund to cash in a few months earlier…
 
i dont think we will go back for Kane now but would be interested to see what happened if we put that £100m bid we made for Kane into a January one for Haaland? 120m euros ish to Dortmund for a player they lose for 75m euros just months later?
Now that would be the interesting.
 
i dont think we will go back for Kane now but would be interested to see what happened if we put that £100m bid we made for Kane into a January one for Haaland? 120m euros ish to Dortmund for a player they lose for 75m euros just months later?
We've got nowt to lose, lets do it, you never know
 
i dont think we will go back for Kane now but would be interested to see what happened if we put that £100m bid we made for Kane into a January one for Haaland? 120m euros ish to Dortmund for a player they lose for 75m euros just months later?
I believe the buyout becomes active in Jan.
 

Don't have an account? Register now and see fewer ads!

SIGN UP
Back
Top
  AdBlock Detected
Bluemoon relies on advertising to pay our hosting fees. Please support the site by disabling your ad blocking software to help keep the forum sustainable. Thanks.