Div or Dom?

[VladimirIlyich]

The one goal he did score was a good finish, to be fair.

Did his best to put it over the bar though. Seriously though,Solanke looked way off the level needed and he got a good few minutes last season too. Ings,on the other hand,didn't appear to get too much game time to the non statto eye and despite looking pretty poor he still looked much better than Solanke,ie could do a job for a mid table PL club. Solanke's development would need to be phenomenal to challenge for a starting place within the next season or two. I hope he does that big step up but like Paulie D said 'I just dont see it.'

[The Woolster]

We do! It’s very fucking obvious.

 

You may think you do, but really there are a load of cognitive biases clouding your judgement.

 

I remember last year at the start of the season when some people didn't think Salah was a good finisher. I mean, they obviously formed an opinion on too small a sample, didn't they...?

 

He's far from our best finisher, he's very wasteful.

 

However, his movement and positioning is excellent meaning he gets shitloads of opportunities and still scores plenty of goals.

 

If he was a clinical finisher we wouldn't have been able to afford him.

[lifetime fan]

 

You may think you do, but really there are a load of cognitive biases clouding your judgement.

 

I remember last year at the start of the season when some people didn't think Salah was a good finisher. I mean, they obviously formed an opinion on too small a sample, didn't they...?

 

 

Nope.

[magicrat]

I would stick with big Div.  He has at least shown he is capable of playing well and being dangerous in the premier league. 

 

I'd love to be proven wrong on Solanke but I just can't see it.  When a young lad comes you can usually see he's got an eye for goal, he's got pace, he's got a mercurial touch.  Just don't see it 

 

He's no Ngog that's for sure

[The Woolster]

Nope.

 

You sticking to your guns that Salah is not a good finisher?

[VladimirIlyich]

You sticking to your guns that Salah is not a good finisher?

Its a bit of a loaded question as Salah tends to be both maker and finisher of a lot of his goals. He is a very good finisher but could be even better if he worked on his right foot more.

[lifetime fan]

You sticking to your guns that Salah is not a good finisher?

 

His goal record says he is.

 

Your eyes tell you he could be more clinical.

 

That’s the problem with stats.

[Captain]

When stattos claim something, it is usually (hopefully) backed up by evidence...

 

You wish! Stattos should stick to counting goals and corner kicks.

[Numero Veinticinco]

The trick is to know he could finish without stats whilst he wasn’t finishing all of his chances. Since then he has clearly been clinical.

[The Woolster]

Someone tell Michael Edwards, quick!

[The Woolster]

His goal record says he is.

 

Your eyes tell you he could be more clinical.

 

That’s the problem with stats.

 

My eyes are fine, but I think you should have gone to Specsavers

[VERBAL DIARRHEA]

Solanke.

[lifetime fan]

My eyes are fine, but I think you should have gone to Specsavers

Show me a breakdown of his shots to goals ratio from the beginning of the season to say the end up September and from then onwards.

[3 Stacks]

Players who have the knack of being in good positions and who have great movement often get undue criticism for their finishing because they create so many chances and it's just impossible to finish most of those at the highest level.  Some examples are Salah, Cavani and Sterling, two of which are brilliant and the other is very good.

 

That's why you also need stats because they rightfully tell you that those three are very good, whereas just watching them makes you feel that they are wasteful.

[LFC 6 Times]

Sterling is a poor finisher but like Mo’ the volume of chances and quality of chances he gets is a result of his movement which is his best attribute and the quality of the team he plays in.

[Numero Veinticinco]

Comparing Mo with Sterling is a crime against humanity.

[The Woolster]

Show me a breakdown of his shots to goals ratio from the beginning of the season to say the end up September and from then onwards.

 

What are you expecting it to show?

 

And why September, why not October or November

[TheHowieLama]

Might be one for the amnesty thread but I think Salah has just had the season of his life -- to say there were not concerns about his finishing when he arrived is silly.

[Red House]

Didn't Mo have the highest number of big chances missed in the league last season? So, it's hardly ludicrous to say he had some finishing issues in the first part of the season despite his ridiculous goal tally. It definitely got better as the season went on but it's undeniable that he had some Sterling-like open goal misses early on in the season. In any case, Mo's greatest attribute has always been his off the ball movement rather than his finishing. 

[The Woolster]

He was also the player who had the most big chances last season. Total big chances had and total big chances missed are highly correlated. Big chances are also a small sample stat, but he converted a little under 50% of his big chances, compared to the average of a little under 40%, so looks ok to me.

 

But thats the whole point, he missed some easy chances at the start of the season, that does not mean he's not a good finisher, it just means he missed some easy chances. He didn't go from a bad finisher to a good finishrt over the space of a month or 2. The term 'luck' really refers to random distribution.

[Red House]

When stattos claim something, it is usually (hopefully) backed up by evidence...

 

You think Sturridge is a better finisher, I think Sturridge is a better finisher, you'll be hard pressed to find anyone who doesn't think that, but last season Sterling scored 21% of his shots, and Sturridge only scored 8% of his. The season before Sterling converted 11% of his shots, Sturridge again converted only 8%. On average, roughly 10% of all shots are converted, so the last 2 seasons, Sturridge has been below average.

 

This what Ne Moe means with regards to luck, not that all players have the same finishing ability, but that in small samples of shots, you can't tell if someone is a good finisher or not because goals are not evenly distributed amongst those shots. I still think Sturridge is a good finisher despite his poor finishing over the last 2 seasons.

 

Last season Solanke had 23 shots, we have no idea from that amount if he is a good finisher or not unless he had scored a load of goals, and even then there would be luck involved.

 

I just can't see how you can make that claim. All that can be concluded from the argument you've proposed is that the operationalised variable you have used for finishing (i.e. shot conversion) cannot predict whether a player is actually a good finisher or not from that particular sample of shots. What you most definitely cannot say is that the same stands for Captain or lifetime fan or as we may call it the eye test. Simply because shot conversion (or even its evolved form, xG added) does not equate to finishing skill as they are referring to. In fact, it is a highly stripped down proxy for finishing skill.

 

For illustrative purposes, the average informed football fan would not require much mental effort to distinguish a composed finish from a rushed one, or to identify a flukey finish. On the other hand, you would have a hard time building any statistical model that could distinguish between these with similar accuracy. That is because the statistical modeller can only model from what is quantifiable, which means leaving out all relevant aspects that cannot be quantified like composure, body language etc.... (I am a doctoral researcher in cognitive and affective neuroscience, so I face the same issue everyday having to quantify unobservable psychological constructs). How the fuck do you objectively integrate whether a player shat himself in front of goal in your model? It is hardly surprising then that these stripped down measures appear vulnerable to fluctuations over time. But, the brain does not seem to be limited by this quantification problem. If anything, modelling confidence and body language is part of its built-in evolutionary toolkit. So, the human brain can and does derive far more information from each of these 23 shots than its reductionistic quantified equivalent can. In other words, Captain, lifetime fan or any football fan are likely to form significantly better informed opinions from this sample of shots than conversion rate is likely to do.

 

*It is also worth noting that the brain is itself a predictive machine, and one that in all likelihood uses what seems to approximate bayesian computation (see the bayesian brain hypothesis or some of Karl Friston’s work). I am not sure about the But, given the small sample size issue you are mentioning, I would not be surprised if bayesian inference modelling would represent a superior alternative as it does a better job of accounting for uncertainty. This again would suggest to me that the ‘eye test’ would outperform conversion rate models in terms of predictive value.

 

Everything about the eye test would suggest that finishing skill is a fairly stable trait that appears relatively unaffected by aging effects (bit like everyone could see that even a fat old Fowler was still as cool a finisher as always despite the dwindling numbers). When you see such a discrepancy, you’ve got to start asking yourself what exactly your measure is actually measuring. So, I would strongly contend that the 'luck' and repeatability issues that have been highlighted here are due to the poor content/construct validity of your operationalised variable rather than finishing skill per say, as it would be defined by the average football fan.

[LFC 6 Times]

I'm not comparing the two, I'm saying that Sterling's best attribute is his movement coupled with his searing pace like Mo'. But as a footballer Mo' is absolutely streets ahead in every department. My strongest retort to any daft cunt claiming Salah's a one season wonder is citing the fact that not only was he top goal scorer, and has posted great numbers throughout his career but he also missed the most "big chances". Mo' scores fuck loads of goals because he's so fucking brilliant off the ball and will continue to get a huge volume of chances.

 

His finishing gradually improved his in his time here, it was notably poor at the start of the season but from December onwards it was absolutely elite and as good as any Liverpool player I've ever seen. Some of the chances he missed off the top of my head: Watford opening day skied over the bar with his right foot, Arsenal at home 0-0 ball across the face Straight at Cech from 3 yards out reminiscent of Benteke at the Emirates August 2015, opening goal rebound miss vs Leicester in the 3-2, a litany at the etihad in the 0-5 prior Mane's sending off, Dragging it wide against Manc cunts after De Gea's save on Matip, Huddersfield missed pen, Miss away at Spartak after great counter and then at Huddersfield rounding the keeper to kick into side netting.

[aws]

I reckon Mo usually needs to miss the first chance to get his eye in. After that he's deadly.

[VladimirIlyich]

I reckon Mo usually needs to miss the first chance to get his eye in. After that he's deadly.

This is a fair point but he also began to convert his first chances a lot too last season so his confidence became sky high too. Interesting season coming up with what he did last season and I think a fully fit Mane will chip in even more too. We do need a 4th front man who is a good back up for Firmino/Salah in terms of raw finishing as we are still relying on them for almost all of our goals and staying injury free. Shaqiri is a decent back up for Mo and Mane but a Sturridge type Firmino back up still needs to be found.

[The Woolster]

I just can't see how you can make that claim...

 

Nice one, love it, one of my favourite posts on here for a long time!

 

I think its worth pointing out that I'm just a guy who did a few stats modules at uni and dabbled in a bit of footy analytics when they first came out. I pretty much got out of the game when they people with proper stats backgrounds started to get involved. I am very open to the fact I may be wrong on some things!

 

What you most definitely cannot say is that the same stands for Captain or lifetime fan or as we may call it the eye test. Simply because shot conversion (or even its evolved form, xG added) does not equate to finishing skill as they are referring to. In fact, it is a highly stripped down proxy for finishing skill.

 

 

My claim is that from such a small sample of shots (a debate could be had about what is small, but I think 23 shots is), you cannot tell for sure if a player is good at finishing or not, and I believe that stands for the eye test just as much as the statistical test.

 

What the eye test will tell you is how good they were at finishing those particular chances, but if you try to extrapolate how well they will finish the next chance, the next 10 chances, the next 23 chances, or the next 100 chances, I believe there is a good chance you will be very wrong.

 

The eye test is also only an opinion, someone else may well have a different opinion. Now everyone has a right to their own opinion, and I am not saying that anyone's opinion on Solanke's finishing is wrong, what I am arguing against their conviction in their own opinion.

 

My eye test told me that whilst Solanke could have done better with some of his chances, he was no Jesus Navas and was a little unfortunate with some of his opportunities and does not seem to me like he is a bad finisher. Who is right, me or Lifetime? The conversion rate would certainly suggest Lifetime at the moment, but we will see.

 

For illustrative purposes, the average informed football fan would not require much mental effort to distinguish a composed finish from a rushed one, or to identify a flukey finish.

Agreed. However I would say that the majority of shots are in the spectrum of instinctive/reactive due to pressure being applied by defenders, the amount of shots where the shooter has the time to be composed (I guess there is a debate on what composed means) would be relatively few, and the amount of flukey finishes very small.

 

On the other hand, you would have a hard time building any statistical model that could distinguish between these with similar accuracy. That is because the statistical modeller can only model from what is quantifiable, which means leaving out all relevant aspects that cannot be quantified like composure, body language etc.... (I am a doctoral researcher in cognitive and affective neuroscience, so I face the same issue everyday having to quantify unobservable psychological constructs). How the fuck do you objectively integrate whether a player shat himself in front of goal in your model? It is hardly surprising then that these stripped down measures appear vulnerable to fluctuations over time.

I agree again, I am really not a fan of the single game expected goal maps that have become quite prevalent. Some expected goal models can be very different from each other for single games for instance. but over a larger sample, this would come out in the wash.

 

But, the brain does not seem to be limited by this quantification problem. If anything, modelling confidence and body language is part of its built-in evolutionary toolkit. So, the human brain can and does derive far more information from each of these 23 shots than its reductionistic quantified equivalent can. In other words, Captain, lifetime fan or any football fan are likely to form significantly better informed opinions from this sample of shots than conversion rate is likely to do.

Here we start to differ, but then I am not a researcher of cognitive neuroscience, so please correct me if I am wrong!

 

Is this not where the cognitive biases come in? The brain derives an immense amount of information, and some of the cognitive biases come about because the brain attempts to simplify how it processes that information to come to a decision.

 

For instance memory. There was a study looking at the recall of football managers of the key events of a game, they were able to recall 60% of those event accurately, meaning they got 40% wrong, and this was from right after the game. If the judgement on the finishing was taken after watching all 23 chances in a row I might agree with you, but we are talking 23 shots taken over a whole season, how many of those shots do people actually remember, let alone the finer details of each shot?

 

People also tend to mainly remember key events don't they? In football, this often means goals. I don't have a study to back this up, but anecdotally it seems to me that people, on average, overestimate the probability that a goal should be scored from any given chance. How many times do you here the phrase "he should have scored that"? Quite often those comments come from chances that are, on average, scored around 50% of the time. Its like telling someone who called tails and the coin lands on heads and telling them they should have got it right. If my opinion is correct, could this not then bias any opinion about finishing ability if their prior about the quality of the opportunity is already wrong?

 

As for the confidence and body language, I think people can put too much significance into those things. For instance, if we take free kicks, I reckon if you were to ask a load of people who the best free kick takers are, a decent proportion would have Ronaldo high in their list. He has possible scored more free kick goals than anyone else in the last 10 years, and as above, goals are what people tend to remember. I am not sure if you could get anyone who had better, more confident body language when taking a free kick either. But since he's been at Madrid, his conversion rate of free kicks is 7.3%. The average conversion rate of free kicks I think is somewhere between 6.5% and 7%. The best free kick takers (more small sample issues here though...) convert free kicks in the mid-teens.

 

 

*It is also worth noting that the brain is itself a predictive machine, and one that in all likelihood uses what seems to approximate bayesian computation (see the bayesian brain hypothesis or some of Karl Friston’s work). I am not sure about the But, given the small sample size issue you are mentioning, I would not be surprised if bayesian inference modelling would represent a superior alternative as it does a better job of accounting for uncertainty. This again would suggest to me that the ‘eye test’ would outperform conversion rate models in terms of predictive value.

 

Don't know if you have seen, but one of the more recent public research does in fact use Bayesian inference, and it does get good results with smaller samples, however the error margins are very large and only a few players have above average finishing based on the confidence interval https://statsbomb.com/2017/07/quantifying-finishing-skill/

 

Everything about the eye test would suggest that finishing skill is a fairly stable trait that appears relatively unaffected by aging effects (bit like everyone could see that even a fat old Fowler was still as cool a finisher as always despite the dwindling numbers). When you see such a discrepancy, you’ve got to start asking yourself what exactly your measure is actually measuring. So, I would strongly contend that the 'luck' and repeatability issues that have been highlighted here are due to the poor content/construct validity of your operationalised variable rather than finishing skill per say, as it would be defined by the average football fan.

I do believe that finishing ability exists, as I said, I think that Sturridge is a better finisher than Sterling. But I also believe that finishing ability plays are far smaller part in goal scoring than chance quality. Sterling has converted a lot more of his chances over the past few seasons, why? Because he's had much better chances. If he was a good finisher, he would have scored even more!

 

If Solanke continues to get himself into good positions, I believe that he will start scoring goals. That is unless he really is truly bad, as in Benteke bad...