Topic: Statistics and Analytics - insight into our performance

[ElstonGunn]

Nevertheless, the point he's making is that Messi's reduced scoring is because he isn't positioning himself the same as he did before, no?

No. Last year his ExpG per shot was 0.152. This year it's 0.149. So the positions he's taken the shots from are essentially identical.

11tegen11's argument is that Messi's difference in scoring efficiency is simply a function of not finishing as well this year as last--last year he was a historically good finisher, this year below average--which he thinks is just random variation.

[PhaseOfPlay]

No. Last year his ExpG per shot was 0.152. This year it's 0.149. So the positions he's taken the shots from are essentially identical.

11tegen11's argument is that Messi's difference in scoring efficiency is simply a function of not finishing as well this year as last--last year he was a historically good finisher, this year below average--which he thinks is just random variation.

Okay, but the conclusion is the same, yes? That instead of looking at their goal numbers, look at where they take their shots from. Have I got that right?

[ElstonGunn]

Okay, but the conclusion is the same, yes? That instead of looking at their goal numbers, look at where they take their shots from. Have I got that right?

If you're trying to assess whether you should sign a player, yes, look at where he takes his shots from rather  than how well he finishes his chances. Goals are an amalgam of those two factors--so they're not irrelevant, just maybe less so than the positions alone.

[PhaseOfPlay]

If you're trying to assess whether you should sign a player, yes, look at where he takes his shots from rather  than how well he finishes his chances. Goals are an amalgam of those two factors--so they're not irrelevant, just maybe less so than the positions alone.

Okay, glad I figured it out, because I had to read his article 3 times

Here's the thing.

Scouts already know that.

[ElstonGunn]

Okay, glad I figured it out, because I had to read his article 3 times

Here's the thing.

Scouts already know that.

And yet they signed Aspas! 

(His goals were mostly from improbable locations.)

[PhaseOfPlay]

And yet they signed Aspas! 

(His goals were mostly from improbable locations.)

Not to get into that direction, but using one signing to refute that is shaky ground, is it not?

The suggestion seems to be that Aspas was signed based on his number of goals, and not any visual observation of what it is he was actually doing.

[ElstonGunn]

Not to get into that direction, but using one signing to refute that is shaky ground, is it not?

The suggestion seems to be that Aspas was signed based on his number of goals, and not any visual observation of what it is he was actually doing.

Ha, not trying to refute your point with just Aspas, or really at all. Notice the winky face. Nor do I doubt that scouts generally have a good idea of what to look for, or that Aspas was signed without watching him. I do think there are a lot of players who put up statistical red flags in this area, that theoretically could be picked up by scouting (Aspas as one example), that get signed anyway.

Stats analysis is likely most valuable as a way to avoid bad signings than to make great signings, and this kind of reasoning would have told you not to buy Aspas, or to pay 25 million for Benteke, etc.

I do also think that scouts could use a lot more help from this kind of analysis than they probably get. If you'd just followed Ted Knutson's transfer advice this summer--a random financial analyst who likes to bet on and write about football in his free time--you would've been better off than  70%+ of clubs.

[PhaseOfPlay]

Ha, not trying to refute your point with just Aspas, or really at all. Notice the winky face. Nor do I doubt that scouts generally have a good idea of what to look for, or that Aspas was signed without watching him. I do think there are a lot of players who put up statistical red flags in this area, that theoretically could be picked up by scouting (Aspas as one example), that get signed anyway.

Stats analysis is likely most valuable as a way to avoid bad signings than to make great signings, and this kind of reasoning would have told you not to buy Aspas, or to pay 25 million for Benteke, etc.

I do also think that scouts could use a lot more help from this kind of analysis than they probably get. If you'd just followed Ted Knutson's transfer advice this summer--a random financial analyst who likes to bet on and write about football in his free time--you would've been better off than  70%+ of clubs.

But here's the question - the same metrics that scouted Sturridge were also used to scout Aspas. So what could cause the difference in performances between the two?

[ElstonGunn]

But here's the question - the same metrics that scouted Sturridge were also used to scout Aspas. So what could cause the difference in performances between the two?

No idea. I do know that Sturridge for 12m was off the charts by any stats analysis, whereas Aspas was a big meh at best.

[PhaseOfPlay]

The problem is that when someone comes up with the following conclusions:

basically arguing that it's impossible to identify finishing skill, at least over a one-season sample, and that a scout should only pay attention to the quality of positions the striker gets into. Here's the extremely convincing graphs:

Then some explanation is necessary. And here's why - there is a huge place for statistics in football. However, statistical analysis (for better or worse) HAS been used in football since at least the late 50's. So it's not like there's a new path to forge. What statistical analysis needs to do is be approachable to footballing IQ, and not dismissive of it, because statements like the ones I've mentioned above have a massive danger of backfiring and damaging the perception of SA in the eyes of traditional football methods. So a statement such as "ONLY pay attention to the quality of the positions the striker gets into" needs more than a graph. It needs proof. Because positioning is relatively easy to teach. Finishing isn't. Why is that?

[ElstonGunn]

The problem is that when someone comes up with the following conclusions:

Then some explanation is necessary. And here's why - there is a huge place for statistics in football. However, statistical analysis (for better or worse) HAS been used in football since at least the late 50's. So it's not like there's a new path to forge.

What people and a few teams have started doing in the last few years is fundamentally different from what came before. It's seems to me there is a new path to forge and the clubs that are more open to it and on the cutting edge will gain a significant (though likely temporary) advantage.

What statistical analysis needs to do is be approachable to footballing IQ, and not dismissive of it, because statements like the ones I've mentioned above have a massive danger of backfiring and damaging the perception of SA in the eyes of traditional football methods. So a statement such as "ONLY pay attention to the quality of the positions the striker gets into" needs more than a graph. It needs proof. Because positioning is relatively easy to teach. Finishing isn't. Why is that?

"A graph" is the best proof you can ever have for anything like this. It's actually a devastating graph. The only way it could be better proof (other than improving the ExpG metric he's using, which is undoubtedly possible) is if the graph encompassed more data. What kind of "proof" are you imagining?

I don't understand what you're driving at with the last question. The "why" seems irrelevant here.

Let me also add that we've got to be careful about equating "ExpG per shot" with "positioning." The quality of the shots you take is affected by a number of individual characteristics, including dribbling ability, positioning, shot selection, pace, first touch etc.

[PhaseOfPlay]

What people and a few teams have started doing in the last few years is fundamentally different from what came before. It's seems to me there is a new path to forge and the clubs that are more open to it and on the cutting edge will gain a significant (though likely temporary) advantage.
"A graph" is the best proof you can ever have for anything like this. It's actually a devastating graph. The only way it could be better proof (other than improving the ExpG metric he's using, which is undoubtedly possible) is if the graph encompassed more data. What kind of "proof" are you imagining?

I don't understand what you're driving at with the last question. The "why" seems irrelevant here.

Let me also add that we've got to be careful about equating "ExpG per shot" with "positioning." The quality of the shots you take is affected by a number of individual characteristics, including dribbling ability, positioning, shot selection, pace, first touch etc.

How does that tie in with "ONLY pay attention to positioning" and "IMPOSSIBLE to identify finishing skill"?

[ElstonGunn]

How does that tie in with "ONLY pay attention to positioning" and "IMPOSSIBLE to identify finishing skill"?

If I said "only pay attention to positioning," in the sense of only player movement, I mispoke/typed. What I mean is (repeating what I understand 11tegen11's argument to be) is that you should only pay attention to the positioning of where the shots were taken from, statistically, which is a product of much more than just a player's off the ball movement. That is, what ExpG did the player produce?

I don't see that it ties in at all with the difficulty of identifying finishing skill (and note that every single time I've mentioned this I've specifically added the caveat that it's not impossible to find, just that it only shows up as more than noise in a giant sample). That is, unless you're defining finishing skill broadly to encompass factors that affect where the shot was originally taken from, they don't relate to each other at all.

[PhaseOfPlay]

If I said "only pay attention to positioning," in the sense of only player movement, I mispoke/typed. What I mean is (repeating what I understand 11tegen11's argument to be) is that you should only pay attention to the positioning of where the shots were taken from, statistically, which is a product of much more than just a player's off the ball movement. That is, what ExpG did the player produce?

I think I understand now - so we eliminate shots from bad positions, and what we're left with is the more true measure of the strikers ability, then? In other words, get rid of the "noise"?

I don't see that it ties in at all with the difficulty of identifying finishing skill (and note that every single time I've mentioned this I've specifically added the caveat that it's not impossible to find, just that it only shows up as more than noise in a giant sample). That is, unless you're defining finishing skill broadly to encompass factors that affect where the shot was originally taken from, they don't relate to each other at all.

The question I ask is what is being defined as "skill"?

[ElstonGunn]

I think I understand now - so we eliminate shots from bad positions, and what we're left with is the more true measure of the strikers ability, then? In other words, get rid of the "noise"?

Not quite, I'm not being very clear. We don't eliminate any shots. Let me try again.

So ExpG (or Expected Goals) is a new-fangled stat that assigns every shot in a game a number based on how likely it is for an average player to score from that shot. By far the most important factor in determining the likelihood of scoring from a particular shot is the position it was taken from--a shot from the six yard box is much easier to score than from the halfway line. Other factors go into it as well, though--it's easier to score with your feet than your head, easier to score off a throughball than a cross, etc.

So if a shot from the penalty spot off a throughball with your feet is scored 50% of the time, it is assigned an ExpG of 0.50. And that's done for every shot.

Now of course a striker is unlikely to score exactly as often as ExpG predicts. So let's say Sturridge takes 10 shots over the course of 2 games, with a total ExpG of 1.5 goals. This implies that each shot had an average ExpG of 0.15 (that is, a 15% chance of scoring). That 0.15 number can be described as his Expected Goals per Shot, or ExpG/S.

In reality, however, he scores 2 goals. That means that he's scoring at a higher rate than an average finisher would have with the same chance. We can find the exact rate by dividing 2 goals by 1.5 expected goals to get 1.33, showing that his finishing for those 10 shots was 33% above average. Sounds good.

So we've got 4 stats.

One is his Expected Goals per 90 minutes, which in this case is .75.
              (That is we dived his total ExpG (1.5) over two games by 2 (the number of games)).

Second is his ExpG/shot, which is .15.

Third is the total goals he scored, 2.

And fourth is the quality of his finishing (goals per expected goal), 33% above average.

What the post I linked to argues is this:
A) The quality of finishing (i.e. G/ExpG) is over a non-giant sample all noise. The distribution of finishing quality, as defined in this way, is random. So we don't give Sturridge any credit for finishing 33% above average.

B) The really important stat is ExpG/90. That is, what we care about is how many shots (especially, but not only, good shots) the player takes. So Sturridge gets credit in this example for his ability to create so many shots that he's expected to get .75gp90 (a great rate), but doesn't get any credit for finishing those chances at a higher rate than we expected.

The question I ask is what is being defined as "skill"?

The ability to "create" (or take shots that have) high Expected Goals. In less statty language, the ability to create lots of shots and/or shots that are very likely to be converted.

[PhaseOfPlay]

Not quite, I'm not being very clear. We don't eliminate any shots. Let me try again.

So ExpG (or Expected Goals) is a new-fangled stat that assigns every shot in a game a number based on how likely it is for an average player to score from that shot. By far the most important factor in determining the likelihood of scoring from a particular shot is the position it was taken from--a shot from the six yard box is much easier to score than from the halfway line. Other factors go into it as well, though--it's easier to score with your feet than your head, easier to score off a throughball than a cross, etc.

So if a shot from the penalty spot off a throughball with your feet is scored 50% of the time, it is assigned an ExpG of 0.50. And that's done for every shot.

Now of course a striker is unlikely to score exactly as often as ExpG predicts. So let's say Sturridge takes 10 shots over the course of 2 games, with a total ExpG of 1.5 goals. This implies that each shot had an average ExpG of 0.15 (that is, a 15% chance of scoring). That 0.15 number can be described as his Expected Goals per Shot, or ExpG/S.

In reality, however, he scores 2 goals. That means that he's scoring at a higher rate than an average finisher would have with the same chance. We can find the exact rate by dividing 2 goals by 1.5 expected goals to get 1.33, showing that his finishing for those 10 shots was 33% above average. Sounds good.

So we've got 4 stats.

One is his Expected Goals per 90 minutes, which in this case is .75.
              (That is we dived his total ExpG (1.5) over two games by 2 (the number of games)).

Second is his ExpG/shot, which is .15.

Third is the total goals he scored, 2.

And fourth is the quality of his finishing (goals per expected goal), 33% above average.

What the post I linked to argues is this:
A) The quality of finishing (i.e. G/ExpG) is over a non-giant sample all noise. The distribution of finishing quality, as defined in this way, is random. So we don't give Sturridge any credit for finishing 33% above average.

B) The really important stat is ExpG/90. That is, what we care about is how many shots (especially, but not only, good shots) the player takes. So Sturridge gets credit in this example for his ability to create so many shots that he's expected to get .75gp90 (a great rate), but doesn't get any credit for finishing those chances at a higher rate than we expected.

The ability to "create" (or take shots that have) high Expected Goals. In less statty language, the ability to create lots of shots and/or shots that are very likely to be converted.

So where do opposition defenders and goalkeepers come into it? Does shot quality change by level of goalkeeper or number of defenders in between the ball and the goal, even if the position is a "good" position?

What about quality if the final ball? The outcome of a shot is as much a function of the quality of the final delivery as it is the shot itself. How is that factored in?

What about the number of touches before a shot? The majority of goals are scored from one touch, so how does a greater number of touches before a shot get accounted for?

What about psychology? How is that accounted for? A player gets into good positions, receives good final passes, takes a good first touch, but has no confidence in scoring and takes their eye off the ball at the last minute. How is that accounted for?

[ElstonGunn]

So where do opposition defenders and goalkeepers come into it? Does shot quality change by level of goalkeeper or number of defenders in between the ball and the goal, even if the position is a "good" position?

What about quality if the final ball? The outcome of a shot is as much a function of the quality of the final delivery as it is the shot itself. How is that factored in?

What about the number of touches before a shot? The majority of goals are scored from one touch, so how does a greater number of touches before a shot get accounted for?

What about psychology? How is that accounted for? A player gets into good positions, receives good final passes, takes a good first touch, but has no confidence in scoring and takes their eye off the ball at the last minute. How is that accounted for?

None of those things are included for various reasons.

The first and second would be in a perfect world, and probably will in the future as we collect more data. Numbers of touches could be included, but I don't think there's an evidence that it affects the likelihood of it turning into goal except that crosses are harder to score off (which is included). In principal, there's no way to include the last one I don't think, but I really can't imagine what benefit it would provide. Psychology is going to be part of the player's finishing skill, not how likely they should be to score with the shot, so it wouldn't affect ExpG.

That ExpG is still perfectly without those additions is proven by a combination of that first graph, which shows that it's repeatable, and all the research that indicates it correlates with team success better than anything you'd look at traditionally. Not perfect, but a quantum leap ahead of what came before.

As for how not taking those things into account has to do with the finishing skill, the point for me, once again, isn't that you can prove finishing isn't a skill, you can just indicate it's an unimportant skill. It is theoretically possible that the issue here is just that defensive pressure and goalkeeper positioning aren't accounted for. In that sense nothing is proven, and the field is of course still developing. I think it's much more likely that we're talking about a skill that affects tiny, tiny margins and so is only the reason for a goal being scored or not in a relatively small number of cases, making it appear to be nonexistent in a season-sized samples.

[PhaseOfPlay]

None of those things are included for various reasons.

The first and second would be in a perfect world, and probably will in the future as we collect more data. Numbers of touches could be included, but I don't think there's an evidence that it affects the likelihood of it turning into goal except that crosses are harder to score off (which is included). In principal, there's no way to include the last one I don't think, but I really can't imagine what benefit it would provide. Psychology is going to be part of the player's finishing skill, not how likely they should be to score with the shot, so it wouldn't affect ExpG.

Goal chances decrease the more touches a player takes before the shot. Plenty of the previous studies of the game have shown this, so it's incredibly relevant. Interesting that this model doesn't take this into account.

That ExpG is still perfectly without those additions is proven by a combination of that first graph, which shows that it's repeatable, and all the research that indicates it correlates with team success better than anything you'd look at traditionally. Not perfect, but a quantum leap ahead of what came before.

Team success is one thing, but the statement it is making is flying in the face of the knowledge of the game on an individual. With that being the case, the burden of proof for it's veracity is on this model, I think.

As for how not taking those things into account has to do with the finishing skill, the point for me, once again, isn't that you can prove finishing isn't a skill, you can just indicate it's an unimportant skill.

So by this token, if we can coach a player to get into the right areas of the box, we can create a goalscorer. Would that be a fair conclusion from this model? If finishing is indeed an unimportant skill, then really anyone can score a consistently high rate of goals, yes?

It is theoretically possible that the issue here is just that defensive pressure and goalkeeper positioning aren't accounted for. In that sense nothing is proven, and the field is of course still developing. I think it's much more likely that we're talking about a skill that affects tiny, tiny margins and so is only the reason for a goal being scored or not in a relatively small number of cases, making it appear to be nonexistent in a season-sized samples.

So again, if skill is only a tiny factor in goalscoring, any club can theoretically make a killing on the transfer market by buying up small market players from any position, and coaching them to arrive in the box at certain areas, and they will score goals, correct?

[iamrobk]

So where do opposition defenders and goalkeepers come into it? Does shot quality change by level of goalkeeper or number of defenders in between the ball and the goal, even if the position is a "good" position?

What about quality if the final ball? The outcome of a shot is as much a function of the quality of the final delivery as it is the shot itself. How is that factored in?

What about the number of touches before a shot? The majority of goals are scored from one touch, so how does a greater number of touches before a shot get accounted for?

What about psychology? How is that accounted for? A player gets into good positions, receives good final passes, takes a good first touch, but has no confidence in scoring and takes their eye off the ball at the last minute. How is that accounted for?

If I understand everything else correctly, these aren't taken into account statistically, but this is where actual scouting and observation can help.

[PhaseOfPlay]

If I understand everything else correctly, these aren't taken into account statistically, but this is where actual scouting and observation can help.

Which is basically how things are done now anyway. My issue is not without how it could be put to use, but with the idea that finishing is an unimportant skill, and is less important than positioning. I really want to see how that is justified.

[sauril]

This is a really interesting conversation and I'm enjoying reading the discussion. Hope I'm not sticking my nose in too far 

I'm a relative newcomer to following football at this detail but have been watching for years. (I'm Canadian, but lived in England as a child, and my (English) father instilled a love of the game).

Regarding the discussion on finishing skill, I think there's a related stat that comes from NHL hockey (like I said, I am Canadian) that is relevant here. It's called shooting percentage, and is pretty much exactly what you would expect - the number of goals / number of shots on goal.  Now because the number of shots is so much higher (25-30 per team per game x 82 games), shooting %'ge with the top offensive players becomes a real statistic because the sample size is high enough.

The point that I'm getting to is that many of the people who are doing advanced stats in hockey talk about "regression to the mean", where you look at a player who is doing really well - is their shooting percentage much higher than the historical value? If so, they're probably just on a lucky streak where the puck is going in the net a lot. Don't give them a long term contract based on that year. One thing that must be noted is that some players have hugely higher career shooting percentagethan others - and this is where I'm leading...

It's obvious to anyone who watches the game that some players are better finishers than others and that it's a skill. Give Suarez or Sturridge the chance that Coutinho created for Henderson in the Arsenal game (the one he chipped wide) and there's a much greater chance of it going in. So that being said...

Once the number of shots becomes statistically significant, it must be possible to take finishing into account. What's this number? How old does a striker have to be for him to take enough shots? And if you start tracking it from when they are younger, can you get more information to go along with all the other stats to make more informed decision

[rickardinho1]

This is a really interesting conversation and I'm enjoying reading the discussion. Hope I'm not sticking my nose in too far 

I'm a relative newcomer to following football at this detail but have been watching for years. (I'm Canadian, but lived in England as a child, and my (English) father instilled a love of the game).

Regarding the discussion on finishing skill, I think there's a related stat that comes from NHL hockey (like I said, I am Canadian) that is relevant here. It's called shooting percentage, and is pretty much exactly what you would expect - the number of goals / number of shots on goal.  Now because the number of shots is so much higher (25-30 per team per game x 82 games), shooting %'ge with the top offensive players becomes a real statistic because the sample size is high enough.

The point that I'm getting to is that many of the people who are doing advanced stats in hockey talk about "regression to the mean", where you look at a player who is doing really well - is their shooting percentage much higher than the historical value? If so, they're probably just on a lucky streak where the puck is going in the net a lot. Don't give them a long term contract based on that year. One thing that must be noted is that some players have hugely higher career shooting percentagethan others - and this is where I'm leading...

It's obvious to anyone who watches the game that some players are better finishers than others and that it's a skill. Give Suarez or Sturridge the chance that Coutinho created for Henderson in the Arsenal game (the one he chipped wide) and there's a much greater chance of it going in. So that being said...

Once the number of shots becomes statistically significant, it must be possible to take finishing into account. What's this number? How old does a striker have to be for him to take enough shots? And if you start tracking it from when they are younger, can you get more information to go along with all the other stats to make more informed decision

Good post. As you alluded to football doesn't really lend itself to statistics as well as american sports due to the isolated incidents during a game that aren't as repeatable as in hockey, basketball, or baseball to name a few. Sample size will always be an issue in football, and because "finishing" can be done in so many different ways in football in countless scenarios its really difficult to quantify compared to other sports (eg. lay-up %, points "in the paint", field goal %, 3 pointer %, RBI's, etc..)

With football being such a big business though and more and more scrutiny being placed on player transfers and scouting than ever before teams are looking to gain an edge in any way they can, which has seen the birth of companies like Opta to increase the statistical understanding we have of football teams as well as the individual players themselves. Statistics like pass completion percentage (pass accuracy) and aerial duels won (to name two) wouldn't really have been mentioned a couple of years ago but are becoming increasingly more frequently used in football these days.

[BabuYagu]

MCofA had an article about the value of through balls and calculated that any shot taken from any position is at least 2 to 3 times more likely to result in a goal if the preceding pass was a through ball. Liverpool have approximately twice as many through balls as any other team in the Premier League.

This one factor will screw with the shot position data quite a bit. A through ball to Sturridge who beats an advancing keeper from just outside the area (e.g. the chip against Everton) isn't neccessarily a good value shot. However, it does have a high probability in a goal. Likewise, a corner which is headed over the bar from inside the 6 yard box would be a good value shot but isn't very likely to result in a goal due to being under pressure, stretching to reach the ball, unable to keep it down etc.

My point is that surely the type of player you are and the types of chances your side creates for you has a much greater impact than where shots are taken from. Andy Carroll for example will rarely have a chance 1-on-1. Most of his chances will be him fighting with a defender to get to a ball that has been lofted into the box. His chances may be in good shooting positions but he has almost all the work to do himself. Whereas Gerrard and Coutinho has been serving goals up on plates for Sturridge in the last few games. Sturridge has the pace and intelligence to get behind the defence. Carroll does not. Carroll has the stregth to outmuscle defenders and win a fair share of balls that you would normally say are 70/30 in favour of the defence.

[U13]

The last page has blown my mind a bit but it's been very interesting.

The positioning vs finishing ability model is interesting but from my point of view I find it difficult to get on board with because it doesn't seem to accommodate how open a skill finishing is.

as important as the position of where a goal scoring opportunity is, surely the difficulty of a goal scoring opportunity can be largely effected by the defenders around you, the position of the keeper, the flight of the ball you're connecting with etc?

From what I can see that hasn't been taken into consideration by the model used in that article or am I interpreting it wrongly?

[PhaseOfPlay]

The last page has blown my mind a bit but it's been very interesting.

The positioning vs finishing ability model is interesting but from my point of view I find it difficult to get on board with because it doesn't seem to accommodate how open a skill finishing is.

as important as the position of where a goal scoring opportunity is, surely the difficulty of a goal scoring opportunity can be largely effected by the defenders around you, the position of the keeper, the flight of the ball you're connecting with etc?

From what I can see that hasn't been taken into consideration by the model used in that article or am I interpreting it wrongly?

Bingo

[Halcyon Lissome]

It could very well be that the variance in the skill of finishing is insignificant compared to the effective difference in positions where shots are taken. Or simply put; it might be that where you take a shot is far more important than the difference between forwards' finishing capabilities; and that's why they're taken as more important.

[PhaseOfPlay]

It could very well be that the variance in the skill of finishing is insignificant compared to the effective difference in positions where shots are taken. Or simply put; it might be that where you take a shot is far more important than the difference between forwards' finishing capabilities; and that's why they're taken as more important.

So then we can train goalkeepers to be top finishers would be the conclusion, as long as we can get them into the right positions?

[Halcyon Lissome]

So then we can train goalkeepers to be top finishers would be the conclusion, as long as we can get them into the right positions?

I don't understand the point of the question; but if you have a goalkeeper who is adept at getting into the right positions and taking the right shots to a higher degree than your forwards then they should be looking at a change of career. Ultimately, a goalkeeper or a forward is just a human being with a set of traits they're good at, which are usable in a context like football.

It's not that they're just running to a spot and shooting from it; but it encapsulates the other skills (possibly game reading, dribbling, technique, passing, etc) that allows the player to get into those spots that makes it important.

If we said a player that scores the most goals should be played as forward, the statement would be uncontroversial. But the game is a bit more complex and if we could just find that player we would. Taking it a step back then, you look at shot location and the percentages involved with scoring goals from those positions; that are most likely to lead to goals. If Elston is right and they're some of the most significant predictors, then it makes sense to concentrate on such a stat rather than a subjective term like finishing ability.

[PhaseOfPlay]

I don't understand the point of the question; but if you have a goalkeeper who is adept at getting into the right positions and taking the right shots to a higher degree than your forwards then they should be looking at a change of career. Ultimately, a goalkeeper or a forward is just a human being with a set of traits they're good at, which are usable in a context like football.

The point of the question is that there is a lot more to goalscoring than positioning. Merely getting players into positions doesn't guarantee goals. Why is that?

What separates an average goalscorer from a good goalscorer from a great goalscorer, if they all get into the same positions to shoot?

[Halcyon Lissome]

The point of the question is that there is a lot more to goalscoring than positioning. Merely getting players into positions doesn't guarantee goals. Why is that?

What separates an average goalscorer from a good goalscorer from a great goalscorer, if they all get into the same positions to shoot?

I added more in the previous post, but what I added is probably more appropriate to this post so I'll just paste it again here:

It's not that they're just running to a spot and shooting from it; but it encapsulates the other skills (possibly game reading, dribbling, technique, passing, etc) that allows the player to get into those spots that makes it important.

If we said a player that scores the most goals should be played as forward, the statement would be uncontroversial. But the game is a bit more complex and if we could just find that player we would. Taking it a step back then, you look at shot location and the percentages involved with scoring goals from those positions; that are most likely to lead to goals. If Elston is right and they're some of the most significant predictors, then it makes sense to concentrate on such a stat rather than a subjective term like finishing ability.

ADD: there's also little need to look at extreme outliers: if a player is getting into the best shooting positions but isn't scoring; then you're really not going to worry about such a player. It's also not about 'guaranteeing' goals as nothing does that; merely looking at the increased probability of scoring. This is a look at the greater/wider picture. Ultimately, it doesn't matter how good you are at finishing if you don't get into position either.

[PhaseOfPlay]

I added more in the previous post, but what I added is probably more appropriate to this post so I'll just paste it again here:

It's not that they're just running to a spot and shooting from it; but it encapsulates the other skills (possibly game reading, dribbling, technique, passing, etc) that allows the player to get into those spots that makes it important.

If we said a player that scores the most goals should be played as forward, the statement would be uncontroversial. But the game is a bit more complex and if we could just find that player we would. Taking it a step back then, you look at shot location and the percentages involved with scoring goals from those positions; that are most likely to lead to goals. If Elston is right and they're some of the most significant predictors, then it makes sense to concentrate on such a stat rather than a subjective term like finishing ability.

ADD: there's also little need to look at extreme outliers: if a player is getting into the best shooting positions but isn't scoring; then you're really not going to worry about such a player. It's also not about 'guaranteeing' goals as nothing does that; merely looking at the increased probability of scoring. This is a look at the greater/wider picture. Ultimately, it doesn't matter how good you are at finishing if you don't get into position either.

The position the Elston is taking is that finishing skill is unimportant, and position is everything, and that scouts should look at where a striker takes their shots from rather than their goal total or conversion rates.

Do you agree with that premise?

[Halcyon Lissome]

The position the Elston is taking is that finishing skill is unimportant, and position is everything, and that scouts should look at where a striker takes their shots from rather than their goal total or conversion rates.

Do you agree with that premise?

As far as I've comprehended his posts, he isn't saying that. He isn't saying that it's unimportant in that sense that you can disregard it completely but that in comparison to the shooting analysis in question it is not nearly as significant to even worry about at this level.

[PhaseOfPlay]

As far as I've comprehended his posts, he isn't saying that. He isn't saying that it's unimportant in that sense that you can disregard it completely but that in comparison to the shooting analysis in question it is not nearly as significant to even worry about at this level.

Am I misinterpreting this statement then?

As for how not taking those things into account has to do with the finishing skill, the point for me, once again, isn't that you can prove finishing isn't a skill, you can just indicate it's an unimportant skill.

[Halcyon Lissome]

Am I misinterpreting this statement then?

I think your reply shows that you're taking it so literally that you're arguing with an extreme example. It's basic sense that no matter how often a player gets into a great shooting position, if he can't hit the target then he's useless there. So some degree of finishing ability is relevant. He's just saying that at the top level the difference between finishing ability in most forwards is negligibly important in comparison to when you look at their shooting choices - or their ability to get into good spots to shoot. In that sense, it's so much more important and repeatable to look at those instances, than the relatively small difference finishing ability can make.

So when you're scouting to get a player; it's better to look at their shooting positions (if they qualify as a decent scorer to begin with) than to pick players on their perceived finishing ability.

[BreakfastPercy]

The last page has blown my mind a bit but it's been very interesting.

The positioning vs finishing ability model is interesting but from my point of view I find it difficult to get on board with because it doesn't seem to accommodate how open a skill finishing is.

as important as the position of where a goal scoring opportunity is, surely the difficulty of a goal scoring opportunity can be largely effected by the defenders around you, the position of the keeper, the flight of the ball you're connecting with etc?

From what I can see that hasn't been taken into consideration by the model used in that article or am I interpreting it wrongly?

I think that's the whole point!

I'll see if I can explain the Messi example better (if not for you it might help someone else), and if I'm wrong anyone please chime in.

Shots are categorized based on where they are taken from.

Given the number of shots Messi has taken, and from what location/category they were taken from, he's expected to have scored 11 goals this season. In fact any player taking those shots from those locations would be expected to score 11 goals, because on average that's what the data says players do.

Messi has actually scored 9 goals, so has converted less than your average player is expected to by the model. Compare that the year before when he converted way more than your average player and you have this wild fluctuation. In which case, statistically, how does Messi's conversion/finishing give any real indicator of whether he's a good striker? One year he's amazing, one year he's below average.

Unlike conversion rates though...

It turns out that players with a high ExpG per 90 in one season, are also the players with a high ExpG per 90 in the next season.

It's a bit like the old 'proper' football adage about having to be in the right place to miss them. You can't bank on Sturridge scoring everything in sight next season because- like you say- what about "the defenders around you, the position of the keeper, the flight of the ball you're connecting with." Statistically what you can expect though is that Sturridge will keep getting himself into those areas which is where he will continually set himself apart and that is something that statistically is a lot easier to predict and therefore scout.

(I hope I've gotten that right, sorry if not)

[RedDot]

First-time poster here. Have found the discussion really interesting so far. Have a few of my layman thoughts to (humbly) add to the conversation  I feel that the proposed model is useful in some ways, but perhaps there needs to be more qualifying as to what kind of insights the ExpG statistic can produce. 

On ExpG, I would think that strikers with a consistently high score, are benefiting from a combination of their team's skill at creating high-probability chances (precise throughballs, interplay of passes) and their own talent at getting on the end of such chances (movement, timing of runs).

On finishing, I believe that the ExpG score does assign some form of value to the striker's skill. The exact calculations are private at this point (i believe) so its hard to tell just how valid they are. Examples might be (just guessing) whether a shot was placed or blasted and how many defenders were bypassed or beaten on the way to goal.

If seen in that light, ExpG might be a good indicator of a striker's real quality (less the ability of his team to create quality chances). If a striker is outperforming his ExpG significantly for one season, that might indicate the presence of a hot streak (and the opposite is true). This is good for scouts in that they can detect help us to detect the presence of a potential dud (one-season wonder) or a hidden gem (a good striker whose output and value has fallen due to an unlucky season).

Other thoughts to consider… if a striker regularly outperforms his ExpG for more than two or three seasons, does that suggest the presence of an unaccounted factor or skill set that is allowing him to get goals more often? I reckon that there still needs to be further testing on a larger number of samples in order to verify the strength of this system. But it's an interesting model.

[jooneyisdagod]

Great chat going on lads. Keep it up. Just wanted to point out that the reason the model finds finishing is an 'unimportant' skill is because of the fluctuations between different years which in turn could be influenced by tactics, how the player got into the position, the positions of the defenders and the goalkeeper etc. But could it also be that the difference is negligible simply because the shots statistics were obtained for mostly strikers that have actually been trained for their finishing skill since they were kids ? So what I'm trying to say is given most top flight strikers are top flight strikers by virtue of having some attributes, physical, mental and tactical and have been trained to harness, the difference in finishing ability itself is so negligible when compared to the ability to get into the regions where shots carry a greater chance of a goal. So perhaps including a goalkeeper's shooting stats if it is possible would indicate a clear difference in 'finishing ability' than including only strikers who are more or less 'maxed out' ? Of course there will always be outliers but that's to be expected with stats.

[PhaseOfPlay]

I think your reply shows that you're taking it so literally that you're arguing with an extreme example. It's basic sense that no matter how often a player gets into a great shooting position, if he can't hit the target then he's useless there. So some degree of finishing ability is relevant. He's just saying that at the top level the difference between finishing ability in most forwards is negligibly important in comparison to when you look at their shooting choices - or their ability to get into good spots to shoot. In that sense, it's so much more important and repeatable to look at those instances, than the relatively small difference finishing ability can make.

So when you're scouting to get a player; it's better to look at their shooting positions (if they qualify as a decent scorer to begin with) than to pick players on their perceived finishing ability.

Okay, that's much clearer.

Now my next question - What influences positioning?

[Endoe]

Okay, that's much clearer.

Now my next question - What influences positioning?

The quality of the finishing, of course




 

[PhaseOfPlay]

The quality of the finishing, of course