Expected assists - xA. Is there really any point?

changingman said:
Must admit I'm a bit of a stats geek and whilst the methodology is up for debate, xA could be a very useful stat.

As mentioned above it's effectively normalising the "assist" statistic, which could be impacted by poor finishing. A player might create 100 clear cut chances but a striker only finishes one of them.

I might be mistaken but if I were the coaching staff I'd be looking to develop passages of play which result in the player with the highest xA (presumably McGree) playing with the player with the highest xG.
Click to expand...
Ok, I understand what you mean.

So let's look at some other stats of the same ilk.

Forss has xA of 0.9 and has 4 assists. So what does it tell us about him on the flip side? He's rubbish at creating chances but has just been really lucky? Many say attacking footballers should be all about 'end product' (ie what has been delivered) not hypotheticals. Forss also has 7 goals but his xGOT (goals on target) is only 4.0 and his xG is 5.1. What does this all tell us about Forss?

Crooks xA is 1.8, but has 5. Another one whose crap but just gets really lucky scoring and creating goals? Should we get shot of him for someone who has a higher xG+xA per 90 and doesn't rely on luck?

I'd rather just judge players on what they've actually done. I don't know exactly how Brentford went about their 'moneyball' model but I'd really be interested to know if xA, xG, xGOT, xA per 90 or xG per 90 played any significant part in it.
 
I do like this stat, but as ever, it has it's flaws.

Barlaser's cross for Akpom's first on Saturday is clearly an assist, but Forss' pass to McGree back at Norwich away (when he scored that wonder goal) is also classed as an assist, which for me, didn't exactly create that goal as such.

It all boils down to the quality of the finisher of who's receiving the pass.

The conversion rate of the person receiving the pass is the most crucial when looking at 'assists' - this is why xA is a good metric as it removes the quality and the conversion rate of the finisher.
 
changingman said:
Must admit I'm a bit of a stats geek and whilst the methodology is up for debate, xA could be a very useful stat.

As mentioned above it's effectively normalising the "assist" statistic, which could be impacted by poor finishing. A player might create 100 clear cut chances but a striker only finishes one of them.

I might be mistaken but if I were the coaching staff I'd be looking to develop passages of play which result in the player with the highest xA (presumably McGree) playing with the player with the highest xG.
Click to expand...
Why, surely if we're playing this daft game wouldn't you pair highest xA with player with highest shot conversion?
In your example above the useless striker would have a very high xG and the 100 clear cut chance creator would have very high xA. Yet they were **** together.
 
atypical_boro said:
Ok, I understand what you mean.

So let's look at some other stats of the same ilk.

Forss has xA of 0.9 and has 4 assists. So what does it tell us about him on the flip side? He's rubbish at creating chances but has just been really lucky? Many say attacking footballers should be all about 'end product' (ie what has been delivered) not hypotheticals. Forss also has 7 goals but his xGOT (goals on target) is only 4.0 and his xG is 5.1. What does this all tell us about Forss?

Crooks xA is 1.8, but has 5. Another one whose crap but just gets really lucky scoring and creating goals? Should we get shot of him for someone who has a higher xG+xA per 90 and doesn't rely on luck?

I'd rather just judge players on what they've actually done. I don't know exactly how Brentford went about their 'moneyball' model but I'd really be interested to know if xA, xG, xGOT, xA per 90 or xG per 90 played any significant part in it.
Click to expand...
Spot on.
 
atypical_boro said:
Ok, I understand what you mean.

So let's look at some other stats of the same ilk.

Forss has xA of 0.9 and has 4 assists. So what does it tell us about him on the flip side? He's rubbish at creating chances but has just been really lucky? Many say attacking footballers should be all about 'end product' (ie what has been delivered) not hypotheticals. Forss also has 7 goals but his xGOT (goals on target) is only 4.0 and his xG is 5.1. What does this all tell us about Forss?

Crooks xA is 1.8, but has 5. Another one whose crap but just gets really lucky scoring and creating goals? Should we get shot of him for someone who has a higher xG+xA per 90 and doesn't rely on luck?

I'd rather just judge players on what they've actually done. I don't know exactly how Brentford went about their 'moneyball' model but I'd really be interested to know if xA, xG, xGOT, xA per 90 or xG per 90 played any significant part in it.
Click to expand...
Not all assists are equal. Goals is a bit different because realistically goals can only be scored from a certain range and usually deliberately. An example of how assists aren't all equal is in 1986 Hector Henrique made a 5 yard pass to Maradona in his own half who was surrounded by 2 players. Maradona went a dribbled through the England team and scored one of the greatest goals ever. That is very different from a defence splitting pass that leaves it on a plate for a striker to tap in an open goal. You would obviously score more goals creating chances that could be tap ins than you could making 5 yard passes in your own half and hoping your striker can go on a 50 yard dribble past the entire defence on his own.

Those stats tell us that Forss has made passes where the attacker still had a lot of work to do (McGree goal against Norwich, Chuba goal v Sheff Utd and McGree goal against Blackburn). Compare that to last season where Jones was regularly getting to the byline and putting them across the 6 yard box for tap ins. Forss isn't there for his assists but for his ability to make runs and finish chances. He's a good finisher which is why he is scoring more than he would be expected to with his chances.
 
Nano said:
Not all assists are equal. Goals is a bit different because realistically goals can only be scored from a certain range and usually deliberately. An example of how assists aren't all equal is in 1986 Hector Henrique made a 5 yard pass to Maradona in his own half who was surrounded by 2 players. Maradona went a dribbled through the England team and scored one of the greatest goals ever. That is very different from a defence splitting pass that leaves it on a plate for a striker to tap in an open goal. You would obviously score more goals creating chances that could be tap ins than you could making 5 yard passes in your own half and hoping your striker can go on a 50 yard dribble past the entire defence on his own.

Those stats tell us that Forss has made passes where the attacker still had a lot of work to do (McGree goal against Norwich, Chuba goal v Sheff Utd and McGree goal against Blackburn). Compare that to last season where Jones was regularly getting to the byline and putting them across the 6 yard box for tap ins. Forss isn't there for his assists but for his ability to make runs and finish chances. He's a good finisher which is why he is scoring more than he would be expected to with his chances.
Click to expand...
Thanks, good explanation. I still think just watching a player a few times in the flesh would give you most of what you need though, wouldn't it?
 
Nano said:
Not all assists are equal. Goals is a bit different because realistically goals can only be scored from a certain range and usually deliberately. An example of how assists aren't all equal is in 1986 Hector Henrique made a 5 yard pass to Maradona in his own half who was surrounded by 2 players. Maradona went a dribbled through the England team and scored one of the greatest goals ever. That is very different from a defence splitting pass that leaves it on a plate for a striker to tap in an open goal. You would obviously score more goals creating chances that could be tap ins than you could making 5 yard passes in your own half and hoping your striker can go on a 50 yard dribble past the entire defence on his own.

Those stats tell us that Forss has made passes where the attacker still had a lot of work to do (McGree goal against Norwich, Chuba goal v Sheff Utd and McGree goal against Blackburn). Compare that to last season where Jones was regularly getting to the byline and putting them across the 6 yard box for tap ins. Forss isn't there for his assists but for his ability to make runs and finish chances. He's a good finisher which is why he is scoring more than he would be expected to with his chances.
Click to expand...
Great explanation, in effect it's about the "quality" of chance the assist leads to.

Has anyone got a link to the databy the way?
 
atypical_boro said:
Thanks, good explanation. I still think just watching a player a few times in the flesh would give you most of what you need though, wouldn't it?
Click to expand...
I guess that is part of the purpose of the metric though, isn't it? You can theoretically compare hundreds of players on a database in minutes, rather than watching one player over a few games over a more significant period. Clubs undoubtedly will do both though.
 
Magic Man said:
I guess that is part of the purpose of the metric though, isn't it? You can theoretically compare hundreds of players on a database in minutes, rather than watching one player over a few games over a more significant period. Clubs undoubtedly will do both though.
Click to expand...
Well yeah I suppose, but I guess it depends how the data is used.

I bet loads of these younguns nowadays in their late teens will be claiming 'such and such is mint' cos he has a high xG+xAper90 without actually bothering to sit down and watch them though.

I think with this post I may have officially transitioned into middle age. Get me a sports car.
 
JackG said:
I suggest you address your queries to Tony Bloom (Brighton) and Mathew Benham (Brentford). They've successfully turned two lower-league clubs into top-half PL sides by understanding the usefulness of data. They've also made a fair few bob for themselves.
Click to expand...
I agree with you but the £450m + spent by Tony Bloom has presumably helped too.

I think Brentford is *possibly* the more interesting story of the two TBH, though the press do seem to have a thing about Brighton.
 
Also one thing to note on the ‘statistical’ approach of Brighton, Brentford, Southampton et al is that they’re only really doing what clubs like Benfica, PSV, Sporting Lisbon have been doing for decades; scouting in unusual or unlikely places and polishing up rough diamonds.

They deserve a lot of credit for the way they’re going about their business but you see a lot of hyperbole around the way they work. What people like Matthew Benham have done is bring a more Continental approach to the English game in terms of hiring and firing.

They’ve nailed a system down and started recruiting to for specific players to fill specific roles in the team, who can then be sold at a profit.
 
You must log in or register to reply here.
Back
Top