Topic: Statistics and Analytics - insight into our performance

[JackWard33]

What I found interesting was the frequency distribution graph. I've taken the liberty of adding a rough area graph to simulate a histogram distribution and to push the point more.



Our mode has deviated from the 0-1 goal which implies there's a higher probability of a win. Of course both Nessy's and Jack's points have merit and it's not like we will categorically state "Look, correlation...therefore cause".  But it' a pattern nevertheless and warrants pointing out.

There might be a lot of noise here, but to paraphrase Nate Silver it doesn't mean there isn't any signal in the noise.

The inevitable question though is what are you trying to prove?
It's not possible to draw a conclusion because there's so many problems with the initial premise (no control - what would have happened if Coutinho had played in the matches he didnt, no way to identify whats causing the uplift if there is one - is it Coutinho or is it playing 4 attackers not 3 for example and so on)

[JCB]

The inevitable question though is what are you trying to prove?
It's not possible to draw a conclusion because there's so many problems with the initial premise (no control - what would have happened if Coutinho had played in the matches he didnt, no way to identify whats causing the uplift if there is one - is it Coutinho or is it playing 4 attackers not 3 for example and so on)

And have any of us drawn any conclusions, tried to prove anything or said anything to that effect. We're all totally aware that this may or may not be incidental or that it may or may not be a pattern/trend. I've made no assertions here.

When looking at stats we can only point out the data and let the viewer make his own decisions based on the data available. You've quite clearly made yours - ie the data should be dismissed - but trust me I've not made one either way. 

[BassTunedToRed]

Do you see any merit in these with 1 player vs without 1 player stats / metrics

No not particularly, but equally I do find them kind of fascinating. Why has our form been a lot better without Coutinho? I don't have a definitive answer, and there won't be one, but it does fascinate me as it's so counter-intuitive.

[JackWard33]

And have any of us drawn any conclusions, tried to prove anything or said anything to that effect. We're all totally aware that this may or may not be incidental or that it may be a pattern/trend. I've made no assertions here.

When looking at stats we can only point out the data and let the viewer make his own decisions based on the data available. You've quite clearly made yours - ie the data should be dismissed - but trust me I've not made one either way. 

I'm not confronting you so you don't need to be defensive I'm challenging the premise of ALL with player vs without player statistics
If you don't challenge the premise of data you just end up a) wasting time b) giving equal or undue waiting to data results that don't merit it

so when you say 'let the viewer make his own decisions' my post was genuinely asking what decisions can we make from it?
the obvious implication is that it seeks to show  'if we a better team without Coutinho in it or not' .... and my point is that's not a provable or disprovable question.

Your post 'there's noise here but there may be a signal' is the exact problem. Yeah there may.... there may not... it may be that the signal is the opposite of what the "data" shows...

I'm open to being convinced there is merit to trying to answer the better with/without question which is why I asked Bass in the first place... but right now it doesn't look like it given the metrics people are using to try and show it

[the 92A]

While it is hard to draw anything as to why, what it does do is put to bed the argument that Liverpool are nothing without Phil and we'll fall apart without him.

[JCB]

I'm not confronting you so you don't need to be defensive I'm challenging the premise of ALL with player vs without player statistics
If you don't challenge the premise of data you just end up a) wasting time b) giving equal or undue waiting to data results that don't merit it

so when you say 'let the viewer make his own decisions' my post was genuinely asking what decisions can we make from it?
the obvious implication is that it seeks to show  'if we a better team without Coutinho in it or not' .... and my point is that's not a provable or disprovable question.

Your post 'there's noise here but there may be a signal' is the exact problem. Yeah there may.... there may not... it may be that the signal is the opposite of what the "data" shows...

I'm open to being convinced there is merit to trying to answer the better with/without question which is why I asked Bass in the first place... but right now it doesn't look like it given the metrics people are using to try and show it

I didn't mean to sound defensive, I just wanted to state equivocally that at no point where we (me and Bass) trying to prove anything. I'm sorry if I misinterpreted you but your question read as such... I try to avoid making conclusions in any data i publish and I always try and let the data do the talking as it can be quite easy to mislead people with Stats if one wanted to.

Incidentally I think the Null hypothesis of the author here wasn't "Are we better off without Coutinho?", but rather "Are we worse off without him?" Something that a lot of fans were bemoaning and predicting when he left.

His data -given the sample size - in my opinion, seems to suggest that the answer is No. But as I have already stated and will do so again, your points as well as Nessy's have merit. 

[canadianscraggledog]

The last 19 league games since our Spurs loss mini table

Man city     15-2-1 47 points
Liverpool    13-5-1 44 points
MU             12-3-4 39 points
Chelsea      11-4-4 37 points
Spurs         10-5-4 35 points

We are the form team since the early part of the year besides City though we are not far behind.

[groove]

You have to think of all possible outcomes though don't you? That our results would've been even worse without Coutinho and our results since January could've been even better with him. Ultimately, for me, it's hard to see how we're not at least slightly worse off without him but these have to come from footballing reasons. We absolutely shouldn't trust with/without stats because the sample size is never big enough within a reasonable time frame. ie when you do get a big enough sample size (it is probably 100s of games by the way for a stat like this) the whole team that the player plays in has changed, the opponents teams have changed and even the game itself has probably slightly evolved that comparing recent results to results 100s of games ago is pointless.

[Chavasse1917]

The last 19 league games since our Spurs loss mini table

Man city     15-2-1 47 points
Liverpool    13-5-1 44 points
MU             12-3-4 39 points
Chelsea      11-4-4 37 points
Spurs         10-5-4 35 points

We are the form team since the early part of the year besides City though we are not far behind.

Results since Christmas would have us first with Spurs and City behind us, apparently!

[JCB]

You have to think of all possible outcomes though don't you? That our results would've been even worse without Coutinho and our results since January could've been even better with him. Ultimately, for me, it's hard to see how we're not at least slightly worse off without him but these have to come from footballing reasons. We absolutely shouldn't trust with/without stats because the sample size is never big enough within a reasonable time frame. ie when you do get a big enough sample size (it is probably 100s of games by the way for a stat like this) the whole team that the player plays in has changed, the opponents teams have changed and even the game itself has probably slightly evolved that comparing recent results to results 100s of games ago is pointless.

That's the problem. The sample size is never going to be big enough AND all internal/external factors remain consistent for you to make an informed decision. We will never have the ideal circumstances to answer anything specific with a degree of confidence. You are then left with 2 options, either:

Don't make any definitive assertions with the data and leave that to the viewer to judge; or
change the question (the Null Hypothesis) to one that can be answered and take a hit...

For example the question shouldn't be: "Are we better without Coutinho?" but "Are our results better in the games we've played without Coutinho". That way you're just assessing results and that is quantifiable. Of course you can still argue about all the external/internal influences that affect the outcome here (teams played, form, players available, etc) but as long as your sample size is big enough and you try not to attribute cause, correlation and effect and are implicit with what you're trying to assess then you're just quantifying facts.

I've often wondered what a sufficient sample size should be when it comes to football stats when you only have a 38 game season. A  lot of the models you see out there run Monte Carlo simulations. This is where you simulate conditions and run it thousands of times. This way you can take advantage of the central limit theorem and a larger data set of values to then predict with a little more certainty, but when assesing empirical data I'm still unsure. That's why it's better to analyse more granular data, ie instead of analysing games or goals, analyse shots or passes. The larger number gives you a greater degree of confidence. 

[BassTunedToRed]

The last 19 league games since our Spurs loss mini table

Man city     15-2-1 47 points
Liverpool    13-5-1 44 points
MU             12-3-4 39 points
Chelsea      11-4-4 37 points
Spurs         10-5-4 35 points

We are the form team since the early part of the year besides City though we are not far behind.

Bear in mind that Liverpool's best rolling 19 game points haul in the PL era is 48. 972 'half seasons', and this isn't far short of the best.

[Roger Federer]

Didn’t we get 44 points in the first 19 games of last season? It’s very good obviously, but to win the title you need to keep that up for a whole season, or even better it, considering how many points City will get. Last year Chelsea got 93 as well.

[BassTunedToRed]

Didn’t we get 44 points in the first 19 games of last season? It’s very good obviously, but to win the title you need to keep that up for a whole season, or even better it, considering how many points City will get. Last year Chelsea got 93 as well.

We got 43. This is the first time Klopp has hit 44.

[Roger Federer]

We got 43. This is the first time Klopp has hit 44.

Ah, right. Cheers.

[canadianscraggledog]

Bear in mind that Liverpool's best rolling 19 game points haul in the PL era is 48. 972 'half seasons', and this isn't far short of the best.

This mini table gives me confidence we will finish strongly and finish 2nd or 3rd.

We go 7-2-1 and finish on 80 points.

[BabuYagu]

This mini table gives me confidence we will finish strongly and finish 2nd or 3rd.

We go 7-2-1 and finish on 80 points.

To be needing less than 2 points per game to finish with 76 points (and therefore almost certainly CL football) is a good spot to be in.

Chelsea, on the other hand, would need 2,4 points per game to better that which would equate to almost 92 points in a season. Possible for sure but a big, especially given the clear problems at the club right now. Just cannot see them stringing that run together winning at least 8 games from 10, if not more.

[Geormajesty]

Michael Caley's progressive passes/runs stats are back!

[Geormajesty]

Oxlade-Chamberlain consistently comes out really well in stats, clear to see why we were so keen on him.

[Yiannis]

When you try so hard to stand out.

What on earth was that? Really. A couple of other sites that use xG placed that chance, naturally, to plus 90% possibility.

[PhaseOfPlay]

When you try so hard to stand out.

What on earth was that? Really. A couple of other sites that use xG placed that chance, naturally, to plus 90% possibility.

Looks like their circles are preset sizes, so perhaps they're operating each shot with a set limit?

[PhaseOfPlay]

Shots on target per goal up to Game 5 (excluding today's game).

Looks like Chelsea are well-earning their position right now. Us and City being just below average, which makes sense given the comments on both teams right now, and their inability to "click" up front just yet:



Save Percentage up to Game 5 (excluding today's game). Essentially, Alisson is earning his corn:

[BrandoLFC]

Shots on target per goal up to Game 5 (excluding today's game).

Looks like Chelsea are well-earning their position right now. Us and City being just below average, which makes sense given the comments on both teams right now, and their inability to "click" up front just yet:



Save Percentage up to Game 5 (excluding today's game). Essentially, Alisson is earning his corn:

Chelsea's finishing is running hot.  I would bet it won't continue, they're 4th in xG and I don't think the eventual better teams they play will allow them to volume shoot as they are right now.

[afc tukrish]

Looks like their circles are preset sizes, so perhaps they're operating each shot with a set limit?

That's what she said?

[Byrneand]

Shots on target per goal up to Game 5 (excluding today's game).

Looks like Chelsea are well-earning their position right now. Us and City being just below average, which makes sense given the comments on both teams right now, and their inability to "click" up front just yet:



Save Percentage up to Game 5 (excluding today's game). Essentially, Alisson is earning his corn:

With regards Chelsea, though you would argue that mean reversions would see their performance deteriorate from here rather than improve.

As we all know there are also a whole host of other factors such as opposition etc.

[PhaseOfPlay]

With regards Chelsea, though you would argue that mean reversions would see their performance deteriorate from here rather than improve.

As we all know there are also a whole host of other factors such as opposition etc.

Why would they deteriorate? Champion contenders usually convert at a rate of below 3:1. Do you think they're going to get that bad?

[PhaseOfPlay]

Chelsea's finishing is running hot.  I would bet it won't continue, they're 4th in xG and I don't think the eventual better teams they play will allow them to volume shoot as they are right now.

It's not really their shot volume, it's their shot accuracy. What they get on target, goes in at a better than average rate, but not too much better. They are certainly capable of sustaining that over a season.

[BrandoLFC]

It's not really their shot volume, it's their shot accuracy. What they get on target, goes in at a better than average rate, but not too much better. They are certainly capable of sustaining that over a season.

More than a 3rd of their goals have been speculative long range efforts that rarely go in even when on target.  I think we've had this conversation before but I don't really recall how it ever ended.  In this instance Chelsea have shot more than LFC with less clear cut chances and a lower xG but have scored more goals.  On target or not I can't see how that is sustainable especially as the players in question either have no history of this conversion rate or are years removed from a similar one.  If they have no history of ever performing at that level then a 5 game sample size is something that makes you think it's sustainable?  It's like Andros Townsend scoring 2 times in a row from his patented 30 yard left footer and then thinking he's finally turned the corner when instead lady luck just smiled on him in that instance and he still takes really shitty shots that hurt his team more than it helps.

[PhaseOfPlay]

More than a 3rd of their goals have been speculative long range efforts that rarely go in even when on target.  I think we've had this conversation before but I don't really recall how it ever ended.  In this instance Chelsea have shot more than LFC with less clear cut chances and a lower xG but have scored more goals.  On target or not I can't see how that is sustainable especially as the players in question either have no history of this conversion rate or are years removed from a similar one.  If they have no history of ever performing at that level then a 5 game sample size is something that makes you think it's sustainable? It's like Andros Townsend scoring 2 times in a row from his patented 30 yard left footer and then thinking he's finally turned the corner when instead lady luck just smiled on him in that instance and he still takes really shitty shots that hurt his team more than it helps.

I'm not going off a 5-game sample size, I'm going off general trends for the top 6 clubs, who tend to score at a slightly better than 3:1 rate over the course of a single season, and over multiple seasons, barring outliers. It's a game of skill, tactics and chance, and the skill and tactics part is what holds off regression to the mean (which - although I'm definitely not a stats expert - generally occurs over long periods of time?)

[BrandoLFC]

I'm not going off a 5-game sample size, I'm going off general trends for the top 6 clubs, who tend to score at a slightly better than 3:1 rate over the course of a single season, and over multiple seasons, barring outliers. It's a game of skill, tactics and chance, and the skill and tactics part is what holds off regression to the mean (which - although I'm definitely not a stats expert - generally occurs over long periods of time?)

But your current example is a 5 game sample size and why wouldn't a player's conversion rate history take more precedence over a general trend for teams?  Adding Ronaldo to Juve, would you expect Juve to shoot and convert exactly as last year or change because of Ronaldo now being on the team? 

For tactics, Chelsea tactics so far seem to be speculative long range shooting and not an ability to consistently break down defenses for high quality chances.  Therefore I would state that based on their current sample size I don't see how their conversion rate is sustainable regardless of what top 6 clubs generally achieve.  Certainly not to where you can say they are on LFC or ManC's level regardless of what the table says right now.

Now vs. Cardiff they did create a lot of high quality chances with Giroud starting instead of Morata.  If they are able to consistently repeat that I would agree they should be able to sustain what they are doing.  Can they repeat that type of performance is the question?  Is it because Giroud makes their offense work better or is it because Cardiff are really bad?  We don't know yet though Cardiff being really bad probably plays a good part in it.

Either way I'm looking forward to our game against them to really get a measure of where they are.  Based on the stats so far they're good but certainly not title contending good, guess we'll see.

[PhaseOfPlay]

But your current example is a 5 game sample size and why wouldn't a player's conversion rate history take more precedence over a general trend for teams?  Adding Ronaldo to Juve, would you expect Juve to shoot and convert exactly as last year or change because of Ronaldo now being on the team? 

For tactics, Chelsea tactics so far seem to be speculative long range shooting and not an ability to consistently break down defenses for high quality chances.  Therefore I would state that based on their current sample size I don't see how their conversion rate is sustainable regardless of what top 6 clubs generally achieve.  Certainly not to where you can say they are on LFC or ManC's level regardless of what the table says right now.

Now vs. Cardiff they did create a lot of high quality chances with Giroud starting instead of Morata.  If they are able to consistently repeat that I would agree they should be able to sustain what they are doing.  Can they repeat that type of performance is the question?  Is it because Giroud makes their offense work better or is it because Cardiff are really bad?  We don't know yet though Cardiff being really bad probably plays a good part in it.

Either way I'm looking forward to our game against them to really get a measure of where they are.  Based on the stats so far they're good but certainly not title contending good, guess we'll see.

Good points, but just to be clear, I was very careful to add the context of the numbers being "right now" and for performances in general to be incomplete "just yet", so I'm not making any predictions off the numbers - I'm just saying that right now, Chelsea have earned their spot in the top three, City and us are not performing as expected but we're not too far off, and to add now - the lower clubs that are running good numbers will probably drop off within the next 5 games - although there's always one that manages to be an outlier.

[Byrneand]

Why would they deteriorate? Champion contenders usually convert at a rate of below 3:1. Do you think they're going to get that bad?

Yes - I do think they will deteriorate.

1) The team really isn't that good.
2) They are scoring at a rate 5% better than the best ratio over the last decade of champions (city 2014 at 2.33).
3) They are scoring at a rate 18% lower than the average of the last decade winner's ratio (2.68).
4) Out of the 14 goals they've scored, 3 have been penalties. if they get one more penalty all season then they will have had more penalties than 5 of the last 10 teams to win it (included in the above stats) got all season. For context, they get 0.6 penalties per game so far this year. That would mean they get 22 in a season.... The previous record for a champion being 11 (Chelsea in 2010). I'm sure another team got more in a single season and didn't win it... but 22 sounds high.
5) 1 of the 14 goals was an own goal. City got 3 OG last year, Chelsea got 1 in 17, Leicester got 0 in 16, Chelsea got 1 in 15. (as a side note looking into this there were 3 OG's in one game Southampton vs Sunderland all for Southampton were they won 8-0. Perhaps that might need to be investigated by the betting police!!). There have usually been about 30 total across all clubs in a year. So Chelsea have got their fair share of luck.
6) They've played easy teams (3 of them at home), hence they have +2 in APLT vs our +6. They will play against teams that can defend.
7) They've had a tonne of long shots that have crept in (abnormal) as highlighted by Brando
The team really isn't that good.

If you strip out 2 of their 14 goals (take your pick or whatever reason), their ratio goes down to 2.58.

So to answer your question:

a) Why would they deteriorate? The above reasons are abnormal. Particularly the penalties.
b) I'm willing to sell the bet that they end up the season with a ratio above 2.5. So at least a 15% deterioration.

[Nessy76]

Interesting stuff. Of course, penalties are not just random events. Teams can win more penalties by playing more football in the opposition box. We got a ton of them with Suarez, because the only way to stop him was to foul him.
Similar with own goals, they are a lot more likely when you are putting defenders under pressure and the ball is moving quickly around the area, both of which are often the result of good aggressive play.
In this case it does seem like the numbers are disproportionate, but without any analysis of how the side is getting into these situations, it's difficult to say if or to what extent they are genuinely overperforming.

[Byrneand]

Interesting stuff. Of course, penalties are not just random events. Teams can win more penalties by playing more football in the opposition box. We got a ton of them with Suarez, because the only way to stop him was to foul him.
Similar with own goals, they are a lot more likely when you are putting defenders under pressure and the ball is moving quickly around the area, both of which are often the result of good aggressive play.
In this case it does seem like the numbers are disproportionate, but without any analysis of how the side is getting into these situations, it's difficult to say if or to what extent they are genuinely overperforming.

Agreed. Same with own goals in that it's pressure.

I don't know what the most penalties received in any year is (knock yourself out whoever wants to add that detail). Looking at Liverpool in 2013/2014, we scored 10 penalties that year. We probably missed one or two too. Along with the 11 from Chelsea in 2010 it's hard to imagine peak penalties being much higher, even when considering trend of getting more from referees in general. VAR would be the step change in that ultimately.

[Yiannis]

Regarding Chelsea, small sample and all but  their xG so far points to what some lads above have said and reminiscent of their league title under Conte.

 

https://understat.com/team/Chelsea/2016

[hesbighesred]

Interesting stuff. Of course, penalties are not just random events. Teams can win more penalties by playing more football in the opposition box. We got a ton of them with Suarez, because the only way to stop him was to foul him.

Yes and no. It should be the case, of course, but the most penalties last season? Crystal Palace. We had fewer all season at Anfield than fucking Spurs did, despite having Salah who has all the qualities you would expect of a penalty winning machine.

From Paul Tomkins' work it seems like one of the best ways of getting penalties is to a) be British and/or b) have a British manager. It's notable that we had so many penalties in that season, when we also had our most British team in a long time.

[BassTunedToRed]

FWIW - Chelsea's conversion rate of 45% for shots on target *should* cool off, else they'd be breaking largely new ground - very few teams have broken 40% in recent times.

[Scottymuser]

But your current example is a 5 game sample size and why wouldn't a player's conversion rate history take more precedence over a general trend for teams?  Adding Ronaldo to Juve, would you expect Juve to shoot and convert exactly as last year or change because of Ronaldo now being on the team? 

As for this question, you'd expect Juve to shoot *far* more than the previous season, and convert at a lower % rate, and probably score about the same number of goals, due to he who fannies about and dives around being there (and his historically below average/poor ratio at converting shots to goals getting out weighed by shooting more than any other player in the game). 

And what a surprise, so far, this is exactly what you'd expect - this season in 4 matches so far, Juve have taken 85 total shots (he who fannies about and dives around 32 of them) at an average of 21.25 shots per match, scoring 9 goals at a acc % of 31.8% and a goals/shot % of 10.6%.  Scale that up to 38 matches, that's around 800-10 shots total, 85ish goals. 

Compare that to last season where they had in the league 554 shots total, 207 on targer with a 37.4% acc rate, and a conversion rate of 15.5% for their 86 goals.    For reference, Icardi had the most shots of any Juve player last year with 114, scoring 22 goals (19.3%) vs in La Liga (2358 mins) - he who fannies about and dives around last year from 2293 had 178 shots and 26 goals, so 14.6% conversion rate.  Weighting that by min, Dybala shot every 20 mins 40 secs on the pitch, Ronaldo every 12 mins 53 secs. 

[BrandoLFC]

As for this question, you'd expect Juve to shoot *far* more than the previous season, and convert at a lower % rate, and probably score about the same number of goals, due to he who fannies about and dives around being there (and his historically below average/poor ratio at converting shots to goals getting out weighed by shooting more than any other player in the game). 

And what a surprise, so far, this is exactly what you'd expect - this season in 4 matches so far, Juve have taken 85 total shots (he who fannies about and dives around 32 of them) at an average of 21.25 shots per match, scoring 9 goals at a acc % of 31.8% and a goals/shot % of 10.6%.  Scale that up to 38 matches, that's around 800-10 shots total, 85ish goals. 

Compare that to last season where they had in the league 554 shots total, 207 on targer with a 37.4% acc rate, and a conversion rate of 15.5% for their 86 goals.    For reference, Icardi had the most shots of any Juve player last year with 114, scoring 22 goals (19.3%) vs in La Liga (2358 mins) - he who fannies about and dives around last year from 2293 had 178 shots and 26 goals, so 14.6% conversion rate.  Weighting that by min, Dybala shot every 20 mins 40 secs on the pitch, Ronaldo every 12 mins 53 secs.

That was my point, I already knew the answer to the question or I wouldn't have stated it as I'd look like a dumbass if I was wrong.  I think you got Icardi mixed up with Dybala there.

[betterthandirt]

saw this posted on reddit and liked it and would share it here, quite statty but some conclusions on how good Milner has been and how much further we are able to push our big 3 forward due to progression ability of Keita/Milner and the fullbacks.

https://saturdaysoncouch.wordpress.com/2018/09/20/manchester-united-vs-liverpool-ball-progression-analysis-the-jesse-lingard-effect/

[clinical]

Come on, own up which Muppet said VVD was slow. He clocked faster speed than Mbappe in our match. Was fastest player.