The Stats Thread

[robbieBB]

@lisboncanary
Reply forthcoming

By the way, a dedicated Stats thread is not a novel feature on this board. One was started way back, I think in 2013, after I had enquired whether sufficient people were interested to make it worthwhile. The answer was Yes! so a thread was duly started. Anderson & Sally's book The Numbers Game: Why Everything You Know About Football is Wrong had recently been published and it was evident that something important was happening in British football of which fans generally were largely unaware. At the same time our club began to take football analytics seriously; it was an important element in Chris Hughton's drive to modernise the football infrastructure and make it fit for purpose for a top level club.

 

[robbieBB]

Has there been any attempt to change the situation where a ball that thumps off the crossbar or post is not a shot on target, and yet a weak effort that goes straight into the goalkeeper's hands and is close to a back pass is a shot on target?

Not that I am aware of. Nor do I think it likely. The salient feature of a shot on target (SOT) is that a goal would ensue from it were the shot not to be saved by the keeper or a “last line” defender with the keeper beaten (Opta). A SOT weakly hit straight at the keeper would be a goal were the keeper not to save it; a shot that hits the woodwork and bounces clear would not. (Shots that hit the woodwork and deflect into the goal are SOTs by the way.)

I’m interested in why you raise this particular issue. Is it that you think a well struck shot that is narrowly off-target has more value (in some sense) than a weakly struck shot that’s on-target?

 

[DHCanary]

You must log in or register to see media

EFL stats on twitter has been doing some interesting comparison of Championship players. The linked thread highlights the areas Skipp in particular has excelled.

The previous CB thread really contrasted Hanley and Gibson quite nicely.

 

[GozoCanary]

Not that I am aware of. Nor do I think it likely. The salient feature of a shot on target (SOT) is that a goal would ensue from it were the shot not to be saved by the keeper or a “last line” defender with the keeper beaten (Opta). A SOT weakly hit straight at the keeper would be a goal were the keeper not to save it; a shot that hits the woodwork and bounces clear would not. (Shots that hit the woodwork and deflect into the goal are SOTs by the way.)

I’m interested in why you raise this particular issue. Is it that you think a well struck shot that is narrowly off-target has more value (in some sense) than a weakly struck shot that’s on-target?

I guess I raise it (as I also raised the issue of the wonderful through-ball compared to merely being the last person to touch the ball before the scorer) to underline the difficulty of turning something in the messy thing called reality into a binary digit. What you say is true - that the weakly struck shot would have gone into the goal if the keeper or a defender hadn't stopped it - but to make the shot the equal of a wonderful shot that hits the underside of the crossbar and somehow bounces back out, or a wonderful shot that the goalie somehow miraculously saves - seems to me to miss out something. The type of shot that hit the crossbar might go in 50% of the time with even a tiny difference in the contact made; the weakly struck shot will only go in 2% of the time when the goalie commits a howler.

EDIT: In theory, for example, a midfielder who never does anything other than make five-yard passes to one of his fellow midfielders could have a successful pass rate of 98% and look like a star compared to another midfielder who attempts defence-splitting passes that often don't work out. I'm aware that the stats people are trying to make their calculations more sophisticated to allow for things like this, but I would question whether this is possible without creating something so byzantine that it is almost impossible to comprehend. Then we're back in the area of subjective judgement because we somehow have to try to balance (in my example) the values of the safe-and-steady guy against his more adventurous colleague. And stats can't do that.

 

[robbieBB]

@lisboncanary
Question: what makes the shot that "hits the underside of the crossbar and somehow bounces back out, or ......... that the goalie somehow miraculously saves" wonderful? What I mean is, "hitting the underside of the bar and somehow bouncing out" could describe all sorts of different shot, from beautifully struck half-volleys to scuffed mis-hits. What would such a shot have to be like for you to call it "wonderful"? What is it that makes that sort of shot "wonderful"?

Here's the point. Shots differ in all sorts of ways, and may vary in value (of some kind) accordingly. But trying to refine the SOT metric may not be the appropriate way to give recognition to such differences. The SOT metric is what it is; it doesn't pretend to be anything else. It simply measures shots on target (as per the definition given). Of course, there is the further question of what the use of that particular metric is.

(I haven't forgotten the other questions you asked. Taking them one at a time, though some points will obviously apply to more than one.)

 

[robbieBB]

The type of shot that hit the crossbar might go in 50% of the time with even a tiny difference in the contact made; the weakly struck shot will only go in 2% of the time when the goalie commits a howler.

Now we're talking! It looks to me as if what you are really getting at is the need to differentiate between SOTs. All goals are equal, but not all SOTs are equal when it comes to resulting in a goal. The probability of a SOT resulting in a goal varies with the "savability" of the shot, and a major factor affecting savability is the point within the area of the goal the ball would cross the goal line. One stat I saw some time ago claimed that the probability of a goal resulting from a shot heading towards the centre of the goal is 12%, as opposed to 50% for a shot heading towards the corners or close to the upright.
Going back to the use of the SOT metric, if what you want is e.g. a measure of striking quality, then a player's SOT score is a better metric than raw shots taken. But two players might have identical SOT scores while having very different Goals totals; shot placement would likely be a major factor in accounting for the difference.
Note, by the way, that I haven't at any point resorted to terms like "wonderful"

 

[GozoCanary]

Now we're talking! It looks to me as if what you are really getting at is the need to differentiate between SOTs. All goals are equal, but not all SOTs are equal when it comes to resulting in a goal. The probability of a SOT resulting in a goal varies with the "savability" of the shot, and a major factor affecting savability is the point within the area of the goal the ball would cross the goal line. One stat I saw some time ago claimed that the probability of a goal resulting from a shot heading towards the centre of the goal is 12%, as opposed to 50% for a shot heading towards the corners or close to the upright.
Going back to the use of the SOT metric, if what you want is e.g. a measure of striking quality, then a player's SOT score is a better metric than raw shots taken. But two players might have identical SOT scores while having very different Goals totals; shot placement would likely be a major factor in accounting for the difference.
Note, by the way, that I haven't at any point resorted to terms like "wonderful"

You've made much of my use of 'wonderful', Robbie, but I'm not sure the difference between that and '50% for a shot heading towards the corners ... ' signifies a great deal more than a semantic distinction. Fans will say that Vrançic's free kick was 'wonderful', while a statistical analysis will say that it went into the top corner as per your 50%. I accept that 'wonderful' was a mistake on my part, though, because it allowed you to move the argument onto a dubious dichotomy between my subjective, emotional response and your scientific one: I become pure Emotion and you become pure Reason. I will try to avoid subjective language in future. I congratulate you on your rhetorical agility, however.

Your two responses since my last post seem to accept that SOT is a blunt instrument and requires more sophistication. I return to my earlier question, therefore, of how that can be done without making the measures required so complex that they either bore or confuse people. And how can it be done without the reintroduction of some kind of subjective evaluation (e.g. how does it allow for a deflected shot that goes into your 50% area - was it 'heavily' deflected and who or what decides this)?

One more question if I may. I assume that you believe that SOT is a meaningful measure despite its limitations in the same way that the various statistical averages can be meaningful even though there are outliers: in our case, that the odd shot which hits the woodwork or heavily deflects does not significantly change the aggregated data. Am I right in my assumption? If so, was I right to use the verb 'believe' because we are talking about an act of faith which cannot be proved, or do you think its usefulness as a measure can be clearly demonstrated?

EDIT: We have focused on SOT but I also want to raise the issue of how stats can evaluate a goalie, which seems even more problematical, but I'll save that for a future post.

 

[robbieBB]

You've made much of my use of 'wonderful', Robbie, but I'm not sure the difference between that and '50% for a shot heading towards the corners ... ' signifies a great deal more than a semantic distinction.

There's actually a quite fundamental point here though. Here's a different example for illustration. The new variant Sars-Cov-2 virus has proved to be between 50% and 70% more transmissible than the original, on account of a mutation to a more efficient mechanism for attaching to and gaining access to cells. Now one may marvel at such a development and the wondrous way in which Nature (in this case in the form of viruses) works, but terms like "marvellous" or "wonderful" have no place in the scientific explanation, and hence understanding, of the development itself. The latter is all to do with the mechanism by which the virus binds to ACE2 receptors on the surface of cells etc. etc. Virologists are human beings; they have emotions like the rest of us. But the language of emotion plays no part in the explanatory theories they develop, which in turn are fundamental to our understanding of viruses. The aim of football analytics is to develop an analogous understanding of "how football works". This being still in its early stages, it remains an open question as to what the fundamental explanatory concepts will turn out to be. But you can get a sense of the direction in which things are moving if you just consider how the concept of space and the use made of it on the field of play has assumed centre stage.

 

[robbieBB]

Your two responses since my last post seem to accept that SOT is a blunt instrument and requires more sophistication. I return to my earlier question, therefore, of how that can be done without making the measures required so complex that they either bore or confuse people. And how can it be done without the reintroduction of some kind of subjective evaluation (e.g. how does it allow for a deflected shot that goes into your 50% area - was it 'heavily' deflected and who or what decides this)?

Yes SOT is a "blunt instrument". Taken on its own, it is of limited usefulness, though not entirely useless. I'd say the same applies to the majority of the metrics commonly referred to as "football stats". Another obvious example would be Possession; we are all familiar with the saying "It's not how much possession you have that matters, it's what you do with it". But even a basic stat like raw %Possession can yield interesting general conclusions; e.g. there's a statistical basis for DF's insistence on his teams playing a possession-based game.

And so yes, the answer is more, and more complex, data. This is how all science-based knowledge and understanding develop. What it does mean is that most of us are going to be reliant on better equipped and more expert expositors if we want to deepen our knowledge of the beautiful game. Phil McNulty's next but one successor as the BBC's chief football writer will need to more resemble the BBC's chief medical correspondent than just another version of Dave Freezer. Even now we could do with a trained football analyst sitting alongside Chris Goreham in the Carrow Road commentary box. Of course, not everyone actually wants to "deepen" their knowledge and understanding of the game; after all, there is an implication that what you might call "football lore" is just that, comparable to e.g. weather lore. Prior to the development of meteorology, such generalisations as "Red sky at night, shepherds's delight" filled the vacuum; they survive because there is usually a grain of truth in them. But if you really want to understand what that red sky has to do with tomorrow's weather, you need to delve into atmospheric pressure systems etc.

Re. subjectivity, why would you think that "delving deeper" into football through analytics will "reintroduce some kind of subjective evaluation"? Does the same apply to the atmospheric physics which explains the relationship between tonight's red sky and tomorrow's fine weather? Actually I don't think the example you give is to do with "subjective evaluation" at all; it's the problem (familiar in all kinds of contexts) of drawing a line in a continuum. The basis on which you do it is a matter of how productive it is (explanatorily, or in some other sense) to draw the line in one particular place rather than another. Here's an analogy. You have a blood test for suspected anaemia. Your GP reports that your red blood cell count is below normal, which he then elaborates on by saying the normal range is between X and Y, while your counts is Z. Where the boundaries of normality are set requires a decision, but not a subjective one, being based e.g. on the objective consequences for people's health if their counts persist in being outside that range.

 

[Farked19]

Anyone remember who the Sheffield United player was back in the 70's who wrote about the role of luck in the life of a professional footballer? He was quite famous at the time as being one of the few who had been to Unversity.

 

[SuffolkCanary]

Taken from the NCFC.central Instagram page:
Daniel Barden is the youngest Norwich GK to keep a clean sheet since Declan Rudd in 2009 v Paulton Rovers.
It’s also only the third time since 2013 that a Norwich goalkeeper has made 7 or more saves without conceding (Gunn v Chelsea 2018 / Krul v Everton 2019).
Barden looked calm and assured in goal against Coventry, it was a good performance from the youngster.

 

[Golden Eadie 2]

Taken from the NCFC.central Instagram page:
Daniel Barden is the youngest Norwich GK to keep a clean sheet since Declan Rudd in 2009 v Paulton Rovers.
It’s also only the third time since 2013 that a Norwich goalkeeper has made 7 or more saves without conceding (Gunn v Chelsea 2018 / Krul v Everton 2019).
Barden looked calm and assured in goal against Coventry, it was a good performance from the youngster.

These are facts SC, not stats. People will get angry with you for posting on the wrong thread.

 

[carrowcanario]

Taken from the NCFC.central Instagram page:
Daniel Barden is the youngest Norwich GK to keep a clean sheet since Declan Rudd in 2009 v Paulton Rovers.
It’s also only the third time since 2013 that a Norwich goalkeeper has made 7 or more saves without conceding (Gunn v Chelsea 2018 / Krul v Everton 2019).
Barden looked calm and assured in goal against Coventry, it was a good performance from the youngster.

Barden looks like a great prospect. Trouble is the best thing he can do now for his development is to go out on loan and play every week, as he's not going to get regular game time with us.

 

[robbieBB]

I assume that you believe that SOT is a meaningful measure despite its limitations in the same way that the various statistical averages can be meaningful even though there are outliers: in our case, that the odd shot which hits the woodwork or heavily deflects does not significantly change the aggregated data.

Yes, just as there are centenarians who have smoked through their adult lives and have never been struck down with lung cancer. But the point that some differences don't matter as regards the value of a particular metric applies in other ways too. Take your example of Assists, where you compare a typical Buendia through ball creating a chance for Pukki, with someone who makes a routine pass to a team mate who then manufactures a great solo goal for himself. A case in point, Oliveira's late equaliser against Wolves in the 2:2 draw at Molyneux in 2017/18. Harrison Read made a straightforward cross field pass from wide on half way, to Nelson unmarked in the centre circle. He ran forward a few yards and fired a shot from way out which evaded John Ruddy:

You must log in or register to see media

Is it necessary to refine the definition of an Assist to exclude such cases in order for the metric to be useful in the way intended? No, the metric as currently defined successfully identifies players who consistently create chances from which goals are scored; in addition, Buendia is just as likely as any other player to benefit from being credited with an "accidental" assist.
The Assist metric could, if necessary, be refined, e.g. by using the xG of the chance created, but in reality quality of chance is already implicit. A player who creates only low quality chances, such as shots from distance, will record fewer assists than one like Emi who regularly creates better quality chances, simply because fewer goals ensue.
Re. "If so, was I right to use the verb 'believe' because we are talking about an act of faith which cannot be proved, or do you think its usefulness as a measure can be clearly demonstrated?", usefulness has to be demonstrable. Someone may have a hunch, indeed most discoveries in science come from hunches -- though hunches are generally born out of prior knowledge and experience, rather than being pure shots in the dark. What matters is that the hunch is shown to be correct.

 

[1950canary]

May I just say without wishing to be too unkind - and I stress this is purely just my own opinion - that the contents of this thread is the biggest load of old b*****ks that I have ever read in my life.

 

[robbieBB]

Pearls before swine 1950, pearls before swine -- and wow!, just look at the size of the herd

 

[Golden Eadie 2]

May I just say without wishing to be too unkind - and I stress this is purely just my own opinion - that the contents of this thread is the biggest load of old b*****ks that I have ever read in my life.

Way to get "likes" 1950. Unfortunately I am the idiot who suggested this thread, but it seems one poster, guess who, is able to spend 18 hours a day formulating mindlessly boring "stats" to have his own private glorification party.

 

[DUNCAN DONUTS]

Pearls before swine 1950, pearls before swine -- and wow!, just look at the size of the herd

You are simply a misguided contrarian on a football message board of a team that's doing well .

You clearly like to play the pantomime villain but not sure what kicks you get out of it.

 

[mike555]

Way to get "likes" 1950. Unfortunately I am the idiot who suggested this thread, but it seems one poster, guess who, is able to spend 18 hours a day formulating mindlessly boring "stats" to have his own private glorification party.

Well, according to Robbie, most of us are not cultured enough to appreciate his pearls of wisdom. And this is not my interpretation of what he said , it is the exact meaning of what he said. I agree with you that this Stats thread has given him the opportunity to "have his own glorification party"
Not the first time he has resorted to quoting from the Bible (Mathew 7.6)

What's the meaning of the phrase 'Pearls before swine'?
Items of quality offered to those who aren't cultured enough to appreciate them.

 

[DUNCAN DONUTS]

Well, according to Robbie, most of us are not cultured enough to appreciate his pearls of wisdom. And this is not my interpretation of what he said , it is the exact meaning of what he said. I agree with you that this Stats thread has given him the opportunity to "have his own glorification party"
Not the first time he has resorted to quoting from the Bible (Mathew 7.6)

What's the meaning of the phrase 'Pearls before swine'?
Items of quality offered to those who aren't cultured enough to appreciate them.

Biblical