The Stats Thread

[GozoCanary]

You must log in or register to see media

Scary in a way, because Emi is almost certainly going at the end of the season whatever happens, and Pukki is not getting any younger. Also, the sum of their two parts is almost certainly greater than simply adding their individual contributions together. As the last few games show, they often look made for each other.

I guess we should enjoy it while we can, and worry about the future when the future arrives.

 

[robbieBB]

Vrancic had 2 chances in the 1st half . The 1st one was perhaps a 0.14 chance. The 2nd was more like 0.4 chance where he dragged it left form 12 yards out . Clearly enough chances to be more than 1 up at half time and does it really matter if DF said should instead of could ?

Mario's first chance was from the edge of the D, and was an 0.05/1 in 20 chance. The one from 15 yards out was the 0.14 chance.
Does it matter whether he said "should" or "could"? What DF said is only important if someone attaches importance to it, in this instance carrowcanario in his post [HASHTAG]#136[/HASHTAG] above.

 

[mike555]

Mario's first chance was from the edge of the D, and was an 0.05/1 in 20 chance. The one from 15 yards out was the 0.14 chance.
Does it matter whether he said "should" or "could"? What DF said is only important if someone attaches importance to it, in this instance carrowcanario in his post [HASHTAG]#136[/HASHTAG] above.

Fair enough Robbie. But Mario 2nd chance was better than 0.14 imo. Should have at least been on target and was a poor miss as he was completely unmarked .

 

[robbieBB]

Fair enough Robbie. But Mario 2nd chance was better than 0.14 imo. Should have at least been on target and was a poor miss as he was completely unmarked .

I think we are all inclined to over-estimate the xG value of chances. I'll have another look at that Mario chance to see why Infogol rate it 0.14. It's rated only slightly higher on Experimental 3-6-1, somewhere between 0.15 and 0.2.

 

[carrowcanario]

Mario's first chance was from the edge of the D, and was an 0.05/1 in 20 chance. The one from 15 yards out was the 0.14 chance.
Does it matter whether he said "should" or "could"? What DF said is only important if someone attaches importance to it, in this instance carrowcanario in his post [HASHTAG]#136[/HASHTAG] above.

Who decides Mario's first chance was an 0.05/1 in 20 chance and the other ws a 0.14 chance. What is their knowledge / experience of football at this level, Mario's abilities. Is the chance specific to Mario or an average Championship player or an average person.

 

[carrowcanario]

Does it matter whether he said "should" or "could"? What DF said is only important if someone attaches importance to it, in this instance carrowcanario in his post [HASHTAG]#136[/HASHTAG] above.

Semantics are irrelevant. DF probably knows the Norwich players more than anyone, seeing them day in , day out in training. Therefore if he thinks we should have done better with the chances we had, then I give this much more weight than anyone else, possibly with the exception of the player themself. You can normally tell when a player thinks they should have done better by the way they react. If they don't think they have a good chance to score then they shouldn't be taking the shot. If they are taking shot thinking they have a 0.1 chance of scoring then I would seriously question their decision making.

 

[DHCanary]

Who decides Mario's first chance was an 0.05/1 in 20 chance and the other ws a 0.14 chance. What is their knowledge / experience of football at this level, Mario's abilities. Is the chance specific to Mario or an average Championship player or an average person.

The numbers quoted in this thread seem to roughly align with Infogol, who state:

"As the game progresses, data is continuously collected for every action that takes places during the game, most notably the attempts on goal. This data is primarily the x, y location on the field where an attempt originates, whether the attempt is made with the foot, the head or another body part and a descriptive assessment of which type of phase of play preceded the attempt.

These numerical inputs are then used to calculate the likelihood that the attempt at scoring will be successful"

The core of the calculation is "what % of similar chances have been converted in the past", and I think they tend to use players in the same division for this. Otherwise you'd see xG data in the lower leagues that suggests everyone is underperforming, because they're being compared to a sample including Messi, etc.

 

[carrowcanario]

The numbers quoted in this thread seem to roughly align with Infogol, who state:

"As the game progresses, data is continuously collected for every action that takes places during the game, most notably the attempts on goal. This data is primarily the x, y location on the field where an attempt originates, whether the attempt is made with the foot, the head or another body part and a descriptive assessment of which type of phase of play preceded the attempt.

These numerical inputs are then used to calculate the likelihood that the attempt at scoring will be successful"

The core of the calculation is "what % of similar chances have been converted in the past", and I think they tend to use players in the same division for this. Otherwise you'd see xG data in the lower leagues that suggests everyone is underperforming, because they're being compared to a sample including Messi, etc.

So if a player misses their first shot for whatever reason, say the ball bobles on the ground, then this will effect each subsequent chance. The criteria is so limited to almost to be worthless.

If similar chances taken in the past are taken into account, this supports the argument that missing chances is a bit of a theme for us.

It is effectively an algorithm based on very limited data and which is supposed to accept this as definitive as opposed to the opinion of DF of other professionals in the game.

The factors mentioned above don't seem to take in some very obvious things that would impact a players ability to score i.e. A player is 8 yards out with only the GK to beat v a player who is 8 yards out with perhaps 12 players between them and the goal which is absurd, even in the case of the players being 8 yards out with only the GK to beat there are so many variables i.e what does the GK do both before and a the point of shooting, does the GK make an exceptional save, is the ball on the players stronger / weaker foot, what effort has the player had to expend to get into that position.

I get that stats can be useful to focus a coaches attention on particular things what I don't accept is that a stat is a better judge of if a player should have scored than a professionals view, particularly one that works day in, day out with a player.

 

[robbieBB]

You must log in or register to see images

The ball, played across by Pukki, who has followed it in, has just left Mario's foot, and is hidden by the video "Play" arrow. If you visualise a cone formed by lines from each goalpost out to where Mario strikes the ball, there is a blocking defender near the apex of the cone. The keeper is central and slightly advanced, and two defenders are running in to further narrow the corridor to the keeper's right. Not sure, but I think the ball actually went through the legs of the defender closest to the keeper.
(The image is a just a screen shot, by the way)

 

[Canary Rob]

Yes, this is the message from the stats, rather than "poor finishing". Goals come from creating big chances for someone; it doesn't matter that much who has the chance.

That’s absurd. It matters very much if the chance falls to Harry Kane or Niklas Bendtner

 

[RiverEndRick]

That’s absurd. It matters very much if the chance falls to Harry Kane or Niklas Bendtner

This sums up my reservations about any preoccupation with stats in that it changes what we see and how we see it. Football is a flowing and unpredictable game and for me that is the attraction. When it gets down to a 0.14 analysis of a chance I'm reminded of Wordsworth's lines:

"Our meddling intellect
Mis-shapes the beauteous forms of things:--
We murder to dissect."

 

[robbieBB]

That’s absurd. It matters very much if the chance falls to Harry Kane or Niklas Bendtner

When it comes to scoring goals, Kane's record is clearly superior to Bendtner's. Taking the Premier League for example, Kane has a goals to game ratio of 1:1.5 against Bendtner's 1:4.25. The question is, though: what is it about Kane that makes him so much more prolific as a goal scorer than Bendtner was? Was Bendtner in some sense "less clinical" than Kane? Given the same chance, does Kane convert it more frequently than Bendtner? If that was the case you might expect Kane to be a much better penalty taker than Bendtner, whereas in fact Bendtner's career penalty success rate was higher than Kane's currently is.
So, rather than it "mattering very much who the chance falls to", maybe what really matters is how good particular players are at engineering high value chances for themselves (with the assistance of team mates obviously). This is what actually distinguishes the players who regularly feature high in the goal scoring charts, the Lewandowskis, C. Ronaldos, Messis, Kanes etc. (and yes, the Pukkis too).
If you take Pukki as an example, a large proportion of his goals come from relatively high value chances created by his running onto central through balls. He isn't actually particularly "clinical" -- much to DF's frustration on Saturday.

 

[Canary Rob]

Ok so you’re making a point that it is a complicated set of factors that dictate whether someone is finishing clinically or not.

Well I think we can agree on that. But it sort of highlights why people can take the view that our finishing was poor and you raising the xG stat as supposedly indicating it wasn’t is rather going round in circles.

Ultimately, the stats don’t tell the true story. If we can agree on that, then let’s just let it lie there - none of us let the stats completely dictate our views. Subjectively, I thought our finishing was poor and has been for much of the season (though Pukki is coming into form I think). Subjectively, I also see our defence as having significantly improved. The stats show certain insights as to certain assessments of chances and defensive work, and they temper my views in some ways, but they won’t change them. Nor should they for the types of reasons you outline above.

 

[Canary Spring]

If "clinical" simply means "score a lot of goals", or "score more goals than most other teams", then of course there are "clinical teams". Likewise "clinical strikers"

No it doesn't have to do with quantity. I didn't say that. It's about success on a per shot basis. We could take the definition further but that should suffice. Saying that there are no clinical strikers; and pointing out, as the article does, that very few strikers ever beat the xG average says to me that all strikers are of nearly equal talent when taking shots. I don't believe that. I think xG and xGA are better used as team stats than for individuals.

 

[robbieBB]

Listed below are 12 players who I think we might all agree are among the top strikers/goal scorers in world football. The table compares their goal tally with their individual xG over the last six and a half seasons (i.e. this season so far, plus the previous six completed ones). The players are ranked by the amount their goal tally exceeds their cumulative xG. The figures are from Understat.

You must log in or register to see images

* Zlatan's figures are for one season fewer than the others, i.e. five and a half seasons.

What the table illustrates is how the best goal scorers in world football don't earn that accolade by hugely out-scoring their xG year after year; they score the goals their xG says they should score, i.e. more or less matching their xG. Even the very best, Messi, manages to exceed his xG by just 5 to 6 goals per season on average. Lewandowski, on the other hand, just manages to break even.

Coming next, a look at how, on that measure, our goal scorers are doing so far this season.

 

[carrowcanario]

Clinical meaning = very efficient and without feeling; coldly detached.

So I would say a clinical striker is one that can control their emotions and remain composed and unflappable when they get a opportunity to score. The other part is very efficient so a key measure for me would be the difference between attempts to score and goals scored, be it a ratio, percentage etc. I can understand the view that xG & xGA may be a more useful metric for a team rather than individual players. I think there are a lot of variables that xG doesn't factor in that effect a players chance score and more importantly whether the decision to have a go was the right one or not. Like most things the best way to test a theory is to look at the extremes. So for me if a player takes a 100 shots at goal in a season and only scores 1 goal, so a 1% conversion rate, then this is not efficient as for me this means for example if they took a shot and missed then they potentially have given away possession in the oppositions third 99 time in 100. This type of data would be a trigger to look much more closely at each of those chances and evaluate why their conversion rate wasn't better. This would involve a professional looking at a recording of each chance and completing an in depth analysis of not just the player, but the whole attacking phase. When DF for example says after a game that we should have scored more this is effectively what he has done albeit by watching the chances in real time rather than via CCTV. Any stat or indeed a professional opinion is just a starting point for investigation rather an end point, like a result. I would equally want to take a much more in depth look at a player who had a 100% conversion rate, i.e. Why are they scoring every time, what's happening when they get in the final third and not scoring, could they take a bit more of a risk and miss a few chances, but ultimately score more goals etc, etc.

https://www.transfermarkt.co.uk/championship/chancenverwertung/wettbewerb/GB2/saison_id/2020
Caution: this is transfermarkt, so accuracy of figures not guaranteed!
The conversion rates given are Shots/goals. On that basis, our conversion rate of 7.1% puts us 14th in the CR table. Transfermarkt don't give a separate rate for SOT/goals, but they do give the SOT figures so rates are easily calculated. Haven't done the complete table but for comparison:
Brentford: 14.4%
Reading: 13%
Swansea: 12.5%
Watford: 11.6%
Norwich: 9.6%

(NB The German for SOT is "Schüsse auf das Tor", literally translated by Transfermarkt as "Shots at the goal".)

So the table you provide gives us a conversion rate of 7.1% (14th in the table), where the average for the league is 7.5%. Now we are top of the league well clear of 2nd place Brentford by 7 points despite them scoring 14 more goals than us, with a conversion rate of 10.2%, indeed all our main competitors, except Watford have a conversion rate of 9.1% or higher, even Watford have a rate of 8.2%. I would expect for a team in our position in the league to be well above average and much higher than 14th. The average conversion rate for teams in the top 6, excluding us, our competitors, is 9.24%, which doesn't compare very favourably with our 7.1%. If our conversion rate was that of the average of the rest of the top 6 we would have scored 61 goals, when we have only scored 47. So for me it is perfectly reasonable for fans using everyday football language to say that we are not as clinical (efficient) as our main rivals in taking our chances. Why this is happening is a whole new can of worms.

Clearly in addition to a considerable amount on anecdotal evidence there is data to support the non clinical view that requires further investigation. It is not sufficient to brush those concerns aside by for example saying our xG is comparable so there isn't an issue, there is.

 

[robbieBB]

Now we are top of the league well clear of 2nd place Brentford by 7 points despite them scoring 14 more goals than us, with a conversion rate of 10.2%, indeed all our main competitors, except Watford have a conversion rate of 9.1% or higher, even Watford have a rate of 8.2%. I would expect for a team in our position in the league to be well above average and much higher than 14th. So for me it is perfectly reasonable for fans using everyday football language to say that we are not as clinical (efficient) as our main rivals in taking our chances. Why this is happening is a whole new can of worms, but clearly there is data to support the non clinical view that requires further investigation, rather than push those concerns aside by saying our xG is ok so there isn't an issue.

I didn't say there is no issue. I am simply questioning the suggestion that the issue is lack of clinical finishing where. by "finishing", I mean execution of the actual shot. Like you, I think we should be looking elsewhere, for instance at decision making and lack of patience, e.g. taking low xG shots versus re-cycling to try and create a significantly better chance.
There have been several instances where people have criticised Pukki for failing to convert chances that they feel he "should" have converted. But if you look at the xG of those chances, more often than not they turn out to be 0.25 xG or worse, i.e 1 in 4 or worse; there's no "should have scored" about it.

 

[Canary Rob]

Listed below are 12 players who I think we might all agree are among the top strikers/goal scorers in world football. The table compares their goal tally with their individual xG over the last six and a half seasons (i.e. this season so far, plus the previous six completed ones). The players are ranked by the amount their goal tally exceeds their cumulative xG. The figures are from Understat.

You must log in or register to see images

* Zlatan's figures are for one season fewer than the others, i.e. five and a half seasons.

What the table illustrates is how the best goal scorers in world football don't earn that accolade by hugely out-scoring their xG year after year; they score the goals their xG says they should score, i.e. more or less matching their xG. Even the very best, Messi, manages to exceed his xG by just 5 to 6 goals per season on average. Lewandowski, on the other hand, just manages to break even.

Coming next, a look at how, on that measure, our goal scorers are doing so far this season.

“Only” 5 or 6 goals a season??

Are you mad?

But as has been said repeatedly, clinical also means getting in the position to score and making the right decision to shoot (or to pass). The XG stat doesn’t show that.

We are all agreed (I think) that the XG stat doesn’t of itself demonstrate clinical finishing. So I’m struggling to understand what you are saying because you seem to be agreeing?

 

[Canary Rob]

I didn't say there is no issue. I am simply questioning the suggestion that the issue is lack of clinical finishing where. by "finishing", I mean execution of the actual shot. Like you, I think we should be looking elsewhere, for instance at decision making and lack of patience, e.g. taking low xG shots versus re-cycling to try and create a significantly better chance.
There have been several instances where people have criticised Pukki for failing to convert chances that they feel he "should" have converted. But if you look at the xG of those chances, more often than not they turn out to be 0.25 xG or worse, i.e 1 in 4 or worse; there's no "should have scored" about it.

Oh I see, so you are arbitrarily narrowing down the meaning of clinical finishing to make a point against which nobody is raising a dispute.

I think the key point being missed here Robbie is that unlike you, I (and I can’t speak for everybody, but I think quite a few others) do not think xG is a measure of much more than passing interest on a match by match and blended season basis. It is far too flawed and simplistic.

That’s not to say I don’t find it interesting and useful, but it does mean I treat it with extreme scepticism because the factors are far more complex than just measuring percentage chances of scoring from any given position (eg it doesn’t factor in the movement or spin of the ball, we don’t have the tech to assess that, the scores are different with different tech which highlights its unreliability). I’d hope we can all agree that xG is a very simplistic measure for a complex situation?

So you making your point by reference to xG ducks the fact that I am not convinced by xG as useful. So it sort of fails before it’s even started, I’m afraid.

 

[robbieBB]

Oh I see, so you are arbitrarily narrowing down the meaning of clinical finishing to make a point against which nobody is raising a dispute.

On the contrary, I think you are "arbitrarily" extending the use of "clinical finishing" to include factors which aren't to do with the execution of the shot itself.
The more general point is that precision in the use of language is a key factor in the growth of knowledge and understanding in any field. It doesn't matter what the "normal" use of "clinical finishing" is. What matters is developing a terminology which allows important distinctions to be explicitly drawn, e.g. the difference between shot execution on the one hand, and choice of when to shoot on the other.