Topic: Statistics and Analytics - insight into our performance

[Klippity Klopp]

This is interesting.

Sometimes laying things out like that can help analyse things.

Thought this was interesting on Palace and although they haven't scored yet in the league they have a higher XG than Chelsea who have scored 12 actual goals more:

[supaerheraw]

https://cartilagefreecaptain.sbnation.com/2017/9/29/16374122/epl-week-7-preview

Interesting read. Brief:

We allow very few attempts at goal, but concedes most goals per attempt (nothing new there).

We have many attempts at goal (second to Spurs), but aren't that effective (nothing new there).

We've met defensively solid teams thus far, and Newcastle looks to be another tough challenge this weekend.

[riismeister]

Pretty extraordinary that Palace have been so unlucky+ineffective that the graph has to start at negative 2.

[BabuYagu]

Sometimes laying things out like that can help analyse things.

Thought this was interesting on Palace and although they haven't scored yet in the league they have a higher XG than Chelsea who have scored 12 actual goals more:

Makes you also wonder where Newcastle would be with a more clinical finisher than Joselu for them. Put Morata in that team, what would happen?

Yes I noticed that on Chelsea. I still think they have problems being masked by the ridiculous form or Morata and things like Alonso putting the ball straight through Lloris. Little things like that early in the season can often hide underlying problems with positive outcomes. Performance indicators are a worry for them. Winning while you build up momentum and good performances though is the mark of a top side. As long as the good performances come though. United last season showed poor performances and good results are unsustainable. Luckily for us, the opposite is also true.

Palace's numbers are ridiculous this season so far. No idea how they haven't scored. Replacing De Boer with Owl face is the most knee jerk thing I have seen.

[KissThisGuy]

Palace's numbers are ridiculous this season so far. No idea how they haven't scored. Replacing De Boer with Owl face is the most knee jerk thing I have seen.

By all accounts they absolutely battered Burnley in his final game. It was only their profligacy in front of goal that cost them the 3 points. He also showed that he was adaptable in their game at Anfield, that was a typical Palace away performance against the big teams. I have no idea how it would have turned out, but the man deserved the opportunity to prove himself for them.

[Triad]

Could it be because Palace produce very little.but the chances that they have produced are clear goalscoring opportunities.Because they produce very little ,the chances that they do produce are usually their player's first shot in the game,so perhaps making them harder to score.I think the sitter Benteke missed against us was his first shot in the game,same goes for RLC against City.

[BabuYagu]

By all accounts they absolutely battered Burnley in his final game. It was only their profligacy in front of goal that cost them the 3 points. He also showed that he was adaptable in their game at Anfield, that was a typical Palace away performance against the big teams. I have no idea how it would have turned out, but the man deserved the opportunity to prove himself for them.

Yep. The thing with Frank de Boer is that you know he is a good coach. Modern thinking. Innovative. It's everything you would want to turn your side into something better as happened with Swansea and Bournemouth. But the start of those processes weren't easy. You can't expect to go from Alardyce to de Boer football with no transition period of shaky results. Think he missed the pace of Zaha too

[BabuYagu]

Could it be because Palace produce very little.but the chances that they have produced are clear goalscoring opportunities.Because they produce very little ,the chances that they do produce are usually their player's first shot in the game,so perhaps making them harder to score.I think the sitter Benteke missed against us was his first shot in the game,same goes for RLC against City.

Good point which I hadn't thought of.

But I had a look and they are 9th in that particular table sandwiched between Chelsea and Stoke averaging 12.7 shots per game.

[riismeister]

Yep. The thing with Frank de Boer is that you know he is a good coach. Modern thinking. Innovative. It's everything you would want to turn your side into something better as happened with Swansea and Bournemouth. But the start of those processes weren't easy. You can't expect to go from Alardyce to de Boer football with no transition period of shaky results. Think he missed the pace of Zaha too

They also failed to recruit to his needs imo. Brought in 4 players, 2 only on loan, and 3 of them aged 21 or under. Not really what you want to do at a time when the club's going through a massive shift in playing style and system.

[Rick Sanchez]

Yep. The thing with Frank de Boer is that you know he is a good coach.

Stop right there. Frank De Boer a good coach   

[BabuYagu]

They also failed to recruit to his needs imo. Brought in 4 players, 2 only on loan, and 3 of them aged 21 or under. Not really what you want to do at a time when the club's going through a massive shift in playing style and system.

Excellent point, hadn't even looked at their transfer business. I agree. Plus the one over 21 was Sakho who was brought in last minute, while they haggled over a few million and had no pre-season whatsoever. The fucking point in that?

[Chris~]

The palace thing is just something that happens. It's very unlikely to happen, but unlikely things happen, especially over a small sample size

[Eeyore]

Good point which I hadn't thought of.

But I had a look and they are 9th in that particular table sandwiched between Chelsea and Stoke averaging 12.7 shots per game.

I remember Ian Rush talking about the difference between playing for Liverpool and for Wales.

With Liverpool he was relaxed if he missed a chance then he knew he would get a chance to make amends. With Wales he was anxious he knew if he got a chance then he was unlikely to get another one. There was more pressure and if he missed the anxiety only got worse. He started thinking more became less instinctive and his finishing dipped.

It's a bit like the chances that go begging when it's 0-0 that fly in when you are a couple of goals ahead and cruising.

Every time Palace miss a chance the next one becomes harder. Players know they can have one of those nights well for me Palace are having one of those seasons. Which begs the question about whether XG and XA should include the state of the match because as Lukaku shows week in week out goals are simply easier to score when you the pressure drops.

[Chris~]

Game state matters more for how say defensive your side is. Chelsea for example spent loads of time in the lead last season and were happy to concede space and shots from certain positions and didn't need to push for goals. Compared to say United who spent a lot more time drawing/losing to teams they should beating and needed to be pushing/creating chances.

Players will miss chances at 0-0 or 5-0. There are no perfect finishers. Lukaku just shows that putting a great goalscorer in a much better team is going to very often lead to goals.

[BabuYagu]

I remember Ian Rush talking about the difference between playing for Liverpool and for Wales.

With Liverpool he was relaxed if he missed a chance then he knew he would get a chance to make amends. With Wales he was anxious he knew if he got a chance then he was unlikely to get another one. There was more pressure and if he missed the anxiety only got worse. He started thinking more became less instinctive and his finishing dipped.

It's a bit like the chances that go begging when it's 0-0 that fly in when you are a couple of goals ahead and cruising.

Every time Palace miss a chance the next one becomes harder. Players know they can have one of those nights well for me Palace are having one of those seasons. Which begs the question about whether XG and XA should include the state of the match because as Lukaku shows week in week out goals are simply easier to score when you the pressure drops.

Good point. I agree 100% that it should. Look how easy it was for Man City to put us to the sword when we were down to 10 men. Every chance was easy. Pressure - both physical from opponents, and mental from the crowd/yourself/gamestate I reckon has the biggest impact of all on finishing ability. I also reckon it's why there are some players who just bottle big games. They know they will get 1 chance. When it comes, the snatch at it.

Btw - It always absolutely bewildered me how David Healy could be one of international footballs most clinical strikers, outshining Torres, Villa, Zlatan and others in the same qualifying group while surrounded by the likes of Clingan supplying him. Yet couldn't do anything at all in club football. It made me think of the opposite of this - where being the hero matters. That you handle that pressure in the best of ways. This is how you find big game players - I think Kuyt was one for us. Gerrard too.

[JackWard33]

Caley posted today we're conceding more expected goals from set pieces than ANY side in the division

Given how few set pieces we will face compared to the relegation fodder that's quite some achievement!!!

[McrRed]

Caley posted today we're conceding more expected goals from set pieces than ANY side in the division

Given how few set pieces we will face compared to the relegation fodder that's quite some achievement!!!

I know you have to be careful with outliers Vs trends but I'm less concerned with xG than actual goals conceded.
Though to be fair to your point, the actual isn't that good (though Watford and City account for a huge chunk of actual).

[JCB]

Based on this rather interesting article in statsbomb https://statsbomb.com/2017/07/trend-ing-topics-how-basic-data-vis-can-show-incredibly-important-football-info/ I decided to see if I can replicate the xG chart for Liverpool especially with what's been going on recently.

I also wanted to map the actual vs the xG as a trend and see how well (or badly) we are performing based on what we should be expecting...



I'll leave it to the others to discuss the implications and insights of the data in depth...

You can find the Tableau chart in the link below. You can use it to chart any of the teams in the top 5 leagues as well as see trends based on seasons and home/away matches as well as seeing each graph in isolation. The charts use a moving average to smooth out the fluctuations in the underlying dataset and to determine the trend, feel free to change the 'Moving Average Range' to calibrate the chart based on the data.

It's my first attempt at this so apologies if it displays badly on other computers and feel free to pm me with feedback if you discover any problems or have any ideas on how it can be improved.

https://public.tableau.com/profile/jcbviz#!/vizhome/xGTrendAnalysis/xGTrendAnalysisDashboard?publish=yes

EDIT: Updated graph after Funky Gibbons pointer

[Funky_Gibbons]

Nice work JCB.

Probably me being thick, but why are the scales different on the LHS and RHS of each graph?

Also, our exG against Leicester, how many of them came from that corner?

[JCB]

Nice work JCB.

Probably me being thick, but why are the scales different on the LHS and RHS of each graph?

Also, our exG against Leicester, how many of them came from that corner?

Nope, I'm the thick one...looks like Tableau did that automatically, which means it's skewing the dataset , and i didn't even realise. I'll need to investigate.

EDIT: Corrected

[Funky_Gibbons]

Nope, I'm the thick one...looks like Tableau did that automatically, which means it's skewing the dataset , and i didn't even realise. I'll need to investigate.

EDIT: Corrected

Thank you. 

So if I'm reading this right according to exG we should have scored ~2.5 and conceded ~1.5?

I remember around the 23rd minute we had about three shots in the area from a corner, Lovren, Wijnaldum and Mane I think. How many exG do they account for in our ~2.5?

[BabuYagu]

Thank you. 

So if I'm reading this right according to exG we should have scored ~2.5 and conceded ~1.5?

I remember around the 23rd minute we had about three shots in the area from a corner, Lovren, Wijnaldum and Mane I think. How many exG do they account for in our ~2.5?

For xg calculations you have to work on a accumulative basis based on every outcome. There was a demonstration of this on Medium.com

https://medium.com/@dannypage/expected-goals-just-don-t-add-up-they-also-multiply-1dfd9b52c7d0

A third instance of expected goals not “adding up” is when you’re evaluating a single attack with multiple shots. Whether that’s a few shots after a zone entry in hockey, or the keeper rebounds the ball and it results in a second shot, adding up expected goals will not give you a clear result. For example:
Imagine a striker takes a solid shot at the keeper, evaluated at 0.25 xG. It’s deflected by the keeper.
That midfielder takes a volley, worth 0.4 xG. It bounces off the crossbar.
The chasing wingback moves into the area, collects the rebound and scores a close header, worth 0.6 xG.
How should we evaluate this attack? If we added these shots together, it would equal 1.25 xG, which is a meaningless number. This would seem to mean that a goal was certain, but we know that’s not true. Likely, but definitely not 100% (or 125%!) certain. If you remember Independent Probabilities from stats class, this will be familiar:

A = 0.25 | A' = 0.75 // P(A) means 25% chance of it happening
B = 0.40 | B' = 0.60 // The ' in B' means not converting the shot.
C = 0.60 | C' = 0.40 // P(Goal) is overall chance of a goal

Addition // Bad
P(Goal) = P(A) + P(B) + P(C) = 0.25 + 0.4 + 0.6 = 1.25
Multiplication // Good
P(Goal) = P(A) + P(A' | B) + P( C | A'| B')
P(Goal) = 0.25 + 0.75*0.40 + 0.60*0.60*0.75
P(Goal) = 0.25 + 0.30      + 0.27
P(Goal) = 0.82

Therefore, we can see that with multiplication, we can appropriately determine the chance of a set of shots producing one goal. In this case, the 3-shot attack would have produced a goal 82% of the time.
If you were to include 1.25 xG in your scoreline instead of 0.82 xG, your expected goals model may overrate teams that created rebounds or unfairly punish teams that give up rebounds. Rebounds are valuable, but must also be evaluated properly, so any expected goals model must keep this in mind. Each model must determine when a shot can be considered independent of any previous attempts on goal. For example: all attacks in a specific zone entry, waiting for a 15 second cool down period after a shot, or waiting for the defense to take control.

[JCB]

Nice work JCB.

Probably me being thick, but why are the scales different on the LHS and RHS of each graph?

Also, our exG against Leicester, how many of them came from that corner?

Looks like my previous concerns became moot with the change Funky but in answer to your question based on the data I have they scored an xG of 3.03 for that game. That corner (the goal and the attempt just before it) contributed to a 1/3 of that total.

[Funky_Gibbons]

Ok, that makes sense, thanks for the answers.

[LiamG]

spotted this on twitter

https://twitter.com/experimental361/status/914554832106778626

[robgomm]

Caley posted today we're conceding more expected goals from set pieces than ANY side in the division

Given how few set pieces we will face compared to the relegation fodder that's quite some achievement!!!

It's a legit weakness and easily the most frustrating thing about us at the moment. STOP IT LIVERPOOL.

[JackWard33]

spotted this on twitter

https://twitter.com/experimental361/status/914554832106778626

We are better than we're perceived at the moment.
Chelsea are worse - especially their attack which has been unsustainably over expected goals
Also probably not enough matches for these numbers to be set (uniteds attack is almost certainly distorted by the weakness of their schedule)

Also city are going to win the league by 10 points ....

[Chris~]

Not playing well and hitting a shit finishing streak is always going to cause frustration/anger from fans though.

[Achilles Heel]

It's my first attempt at this so apologies if it displays badly on other computers and feel free to pm me with feedback if you discover any problems or have any ideas on how it can be improved.

https://public.tableau.com/profile/jcbviz#!/vizhome/xGTrendAnalysis/xGTrendAnalysisDashboard?publish=yes

EDIT: Updated graph after Funky Gibbons pointer

Some positive feedback for you. I've had a play on your graphs this  morning and I think that is brilliant! Very nice work indeed.

I've been recording xG/xGA/xGD on an individual game basis for the Big 6 since the beginning of last season. I'd been using the  infogol app (but recently switched to understat which I only discovered from a conversation on here)  Coincidentally I'd just started building in 10 and 20 game moving averages with a view to charting on excel. My excel skills,  let alone tableau,  are not the best so I really appreciate the link.

[JCB]

Some positive feedback for you. I've had a play on your graphs this  morning and I think that is brilliant! Very nice work indeed.

I've been recording xG/xGA/xGD on an individual game basis for the Big 6 since the beginning of last season. I'd been using the  infogol app (but recently switched to understat which I only discovered from a conversation on here)  Coincidentally I'd just started building in 10 and 20 game moving averages with a view to charting on excel. My excel skills,  let alone tableau,  are not the best so I really appreciate the link.

Thanks for the compliment Achiles, I couldn't find this type of trend data out there for us, so just like you it was something that I had to build just to see it.

My eye keeps on getting drawn to the change in the 3rd graph (xg Against vs Goals Against) around April. Goals Against had remained pretty much above the xG value for 2 years and now we're experiencing a flip in results. I'm sure there's an interesting story there somewhere.

Edit -  Actually now that I think about it, it's probably due to how we clamped down in defence in the run in to the end of last season. The M.A as it takes into account previous games is still including that form in the data but we can slowly see that it's trending back up again (unfortunately) due to our bad defending this season. It concerns me that the xG Against has gone up around the same period relative to what it used to be.

[killer-heels]

What was the situation in terms of Xg and does that bare any relation to how we expect to perform in the future. For example, i think I read that last season City and Utd had pretty high Xg’s but that was not translating to results and wins. I recall numerous City and Utd games where they created but couldnt score and get a winner.

Obviously long term that may have benefitted them this year but it didnt really happen this year. I know some stats people claim that we should improve, but what are the chances that we dont this season?

[redmark]

What was the situation in terms of Xg and does that bare any relation to how we expect to perform in the future. For example, i think I read that last season City and Utd had pretty high Xg’s but that was not translating to results and wins. I recall numerous City and Utd games where they created but couldnt score and get a winner.

Obviously long term that may have benefitted them this year but it didnt really happen this year. I know some stats people claim that we should improve, but what are the chances that we dont this season?

There was an article (I think by Ted Knutson) which discussed some of the drawbacks of xG/xGA and included some (fairly common sense) objections by a coach. xG/xGA (or the precise formula being discussed in that article, at least) rely more or less only on the location, not the wider context (e.g. primarily position of defenders).

So, you see from our xGA/actual GA in that Arpil/May period, our actual goals against dropped more than our xGA. Similarly, teams dominating games against opponents setup defensively may underperform their xG because it is assuming an average defence, not one packing the box with pesky defenders getting in the way of shots.

[killer-heels]

There was an article (I think by Ted Knutson) which discussed some of the drawbacks of xG/xGA and included some (fairly common sense) objections by a coach. xG/xGA (or the precise formula being discussed in that article, at least) rely more or less only on the location, not the wider context (e.g. primarily position of defenders).

So, you see from our xGA/actual GA in that Arpil/May period, our actual goals against dropped more than our xGA. Similarly, teams dominating games against opponents setup defensively may underperform their xG because it is assuming an average defence, not one packing the box with pesky defenders getting in the way of shots.

Ok. The problem i have is that I am clinging on to the belief that the stats guru’s have that our form will inprove.

I dont think Klopp and his side have shown anything that suggests they can go on a long winning run because we are just so poor at the back but maybe up the wins.

[Phil M]

For the year 2017 to date:  - Premier League

Played 26 Won 12 Drew 9 Lost 5 (Swansea H - Hull A - Leic A - Palace H - City A)

Home record: P13 Won 7 Drew 4 Lost 2  Pts: 25/ 39
Away record: P13  Won 5 Drew 5 Lost 3 Pts: 20 / 39
Total Pts: 45 / 78

Scored:      45
Conceded:  33
Overall GD: + 12

At Anfield:

Scored:      25
Conceded:  12

Away:

Scored:      20
Conceded:  21

So it's quite clear we've been conceding far too many away from home. Although 9 of the 21 this season have been in our last 4 league games. (The City 5-0 has a big effect of course)

Clean Sheets:

8 from possible 26

Home: 5 from possible 13      Away: 3 from possible 13

[killer-heels]

8 clean sheets from a possible 26.

We are so bad at defending.

[JCB]

What was the situation in terms of Xg and does that bare any relation to how we expect to perform in the future. For example, i think I read that last season City and Utd had pretty high Xg’s but that was not translating to results and wins. I recall numerous City and Utd games where they created but couldnt score and get a winner.

Obviously long term that may have benefitted them this year but it didnt really happen this year. I know some stats people claim that we should improve, but what are the chances that we dont this season?

Just in case you're interested: xG for the two teams you mentioned relative to Liverpool for last season (and this one)

[Block 203]

For the year 2017 to date:  - Premier League

Played 26 Won 12 Drew 9 Lost 5 (Swansea H - Hull A - Leic A - Palace H - City A)

Home record: P13 Won 7 Drew 4 Lost 2  Pts: 25/ 39
Away record: P13  Won 5 Drew 5 Lost 3 Pts: 20 / 39
Total Pts: 45 / 78

Scored:      45
Conceded:  33
Overall GD: + 12

At Anfield:

Scored:      25
Conceded:  12

Away:

Scored:      20
Conceded:  21

So it's quite clear we've been conceding far too many away from home. Although 9 of the 21 this season have been in our last 4 league games. (The City 5-0 has a big effect of course)

Clean Sheets:

8 from possible 26

Home: 5 from possible 13      Away: 3 from possible 13

What this says to me is that what we are seeing this season points wise is about par for us, as so far we are going along at around 1.7 points per game on average, which is the same when taken over a longer period, from the start of the year. A bit concerning that, do we have the players to up that average? Or is what we've seen so far gonna be how the season goes?
I'm not sure we are capable of putting a big number of wins together.

[Groundskeeper Willie]

What this says to me is that what we are seeing this season points wise is about par for us, as so far we are going along at around 1.7 points per game on average, which is the same when taken over a longer period, from the start of the year. A bit concerning that, do we have the players to up that average? Or is what we've seen so far gonna be how the season goes?
I'm not sure we are capable of putting a big number of wins together.

First 13 games last season yielded 9W 3D 1L. Last 12 games yielded 8W 3D 1L.

So that right there is proof we can string good winning runs together.

[Block 203]

First 13 games last season yilded 9W 3D 1L. Last 12 games yilded 8W 3D 1L.

So that right there is proof we can string good winning runs together.

What it says to me is we do a lot better when we've only got one game per week, which we haven't this season

[Lustig]

The big chance offensive and defensive stat is the one that it looks as it matters the most as it gives you a great insight into your overall level of play.

Offensively we're joint third with 19 Big Chances created - City lead with 30.

Defensively we're joint third worst with 15 Big Chances conceded. Palace are leading with 25. Tottenham have conceded only 5, Chelsea 6, City 7, United, WBA and Burnley 8.