02 Apr 2018 18:48:57
Hi Eds

Was wondering for your opinion on the Expected goals (xG) statistic. How well do you think it reflects the quality of a team?

With Liverpools Expected goals conceded being lower than United's for example, would you agree with the conclusion that our defence is actually stronger, and the difference in goalkeeping between the clubs has actually been the issue?

{Ed001's Note - no idea, I don't take any notice of stats like that as they really do not truly reflect anything. There are not enough games a season for them to be of use. You need a much higher sample size.}

1.) 02 Apr 2018
02 Apr 2018 19:58:05
I work for Opta so I can try to help explain it for you.

Expected Goals (or xG) analyses the quality of a chance based on over ten years worth of similar shots. It takes into account factors such as the distance of the shot, angle of the shot, type of shot (header, volley, left/ right foot), how the chance was created (was it a through ball, rebound, cross etc. ) It is based on how likely the “average” player is to score that chance.

So for example, a shot with an xG of 0.25 means that historically, shots with similar distance, angle, means of creation etc. were scored 25% of the time by most Premier League players. A better goal scorer, such as Harry Kane or Sergio Aguero, might score these chances 50%, 60%, or even 70% of the time, but that is irrelevant. It is based on the average player. How likely any random player would be to score that chance, not the best players.

Liverpool’s xG conceded is indeed the second lowest in the Prem, behind only Man City. But that doesn’t tell the full story. Liverpool have also conceded the 2nd fewest amount of shots. So we’re bound to have conceded a lower xG. Like Ed001 says, the sample size is very important there.

However, one thing I will add, is that when you take the average xG of the shots faced by each Prem team this season, Liverpool’s is the highest. In other words, we might be conceding very few shots, but when we do, they have a higher xG - or based on the definition, they are closer to goal and easier types of shot.

So that actually backs up what we often see. Defensive mistakes often gift opportunities to our opposition, making the chances easier, even though we’re not giving away many of them. So we are actually more likely to concede than Man Utd based on that as they aren’t giving away as good quality chances.

Sorry if that’s really complicated! Does that explain it at all Ed001? Does it perhaps give a little more context than just the numbers themselves, for example? Of course the stat isn’t perfect - none are, particularly when used without context! They can be misinterpreted or misrepresented to make a point. But hopefully that explains it a little better and how it can be used to evidence what we can already see.

2.) 02 Apr 2018
02 Apr 2018 20:08:20
Good read that. Reid.
Thanks.

{Ed001's Note - very good Reid, sorry someone else edited that post and never thought to hold it for me to read. That is my point, they are not of use in and of themselves, it is understanding the context around them that makes them useful. Sadly very few people ever understand context.}

3.) 02 Apr 2018
02 Apr 2018 20:25:47
Glad you guys understood that post! Tried to keep it as simple as I could!

My job at Opta is to editorialise our data - or in other words, make it relevant and give context as best as possible. I see in my job just how many stupid stats are thrown out there with no meaning whatsoever. Expected goals is a much newer stat, so needs to be bedded in gradually and explained more than other stats are. When given context, it can be used well to back up a point. It's just hard sometimes sifting through the rubbish to find good context. I'm happy to help anybody who wants to understand xG or any other stat better! So glad you guys liked that post 😊.

4.) 02 Apr 2018
02 Apr 2018 20:47:51
Great to have you on the site Reid, what with the Ed’s and yourself there’s some real expertise on here. Thanks for the explanation. Now, can you explain why we haven’t won the league for 20 odd years. 😂.

5.) 02 Apr 2018
02 Apr 2018 20:52:51
Really interesting read, thanks for that Reid.

6.) 02 Apr 2018
02 Apr 2018 21:44:02
Very interesting. Thanks Reid.

7.) 02 Apr 2018
02 Apr 2018 22:46:54
Cheers Reid. Good read that.

I'm not an expert on statistics but we've conceded the 2nd fewest shots but let in the 6th fewest goals. Says a lot about how poor we are as a defensive unit in giving up easy chances/ opportunities.

Gotta improve over the summer.

8.) 02 Apr 2018
02 Apr 2018 23:32:35
Thank you Reid, great post mate.

9.) 03 Apr 2018
03 Apr 2018 01:52:44
If we could put a stop to the mistakes we could have a very serious defence. One of the best, contrary to people's belief in liverpool defensively.

10.) 03 Apr 2018
03 Apr 2018 01:56:53
How intriguing. Thanks for that.
My maths teacher used to say there are three types of lies in this world.
Lies,
Damn lies,
And statistics.

11.) 03 Apr 2018
03 Apr 2018 05:23:34
Although it would be a much smaller sample size would anyone know how those goals conceded tally varied by which keeper we had in? Last 15 games for each maybe ( as I said it would be a small sample size)

12.) 03 Apr 2018
03 Apr 2018 10:06:11
Swishcleaner - in Loris Karius' 13 Premier League games this season, we have conceded 11 goals (0.85 per game) . In Simon Mignolet's 19 such games, we have conceded 24 goals (1.26 per game) .

13.) 03 Apr 2018
03 Apr 2018 10:21:19
That was a really clear and helpful post Reid, thank you.

I have a follow-up question: Your analysis helps to explain why we can still look defensively fragile at times, so it gives statistical support to something most of us probably feel (that a lot of the goals we concede are down to individual errors) . Do you have any good, preferably Liverpool-related, examples of where this kind of analysis contradicts a general impression most people have?

Brilliant post Reid, would love to hear more.

14.) 03 Apr 2018
03 Apr 2018 11:47:46
Hi Lemonhead,

It's hard to specifically think of an example where the statistics largely disprove a general impression about something - it all depends on context and how those stats are used.

It's easier to give an example of a commonly misrepresented statistic that does not prove what people try to make out it does.

The best one that springs to mind is the "possession lost" stat. I'm sure you'll have all seen it banded about on SSN and the like. The way I see it utilised most is to criticise Alexis Sanchez - since his Man Utd debut, he has "lost possession" on the second most total number of occasions of all Prem players (behind only Jose Holebas) . But that is so misleading it's untrue.

If we take the whole of 2017-18 into account, the three players who have "lost possession" the most are Kevin De Bruyne, Christian Eriksen and Alexis Sanchez. Now I'm sure that NONE of us would say that KDB is having a poor season. So why on earth is that stat used to criticise Sanchez?

So, as part of my job, I try to give some context there. The fact is, these attacking midfield players are more likely to lose possession more. It's the way they play - trying to play that incisive final ball. It's less likely to come off than John Stones making a 5-yard pass to Kyle Walker in the defensive third, so they are invariably going to lose possession more - it's the nature of their role. Indeed, they have the ball more than most players do, and have it in more attacking areas - so of course when you have it more, you will lose it more pro rata.

Opta cannot avoid people using that possession lost stat. Unfortunately, it exists, it's out there. So myself, and others in my department, have to try give context around that to better explain it.

So we use the possession lost stat in relation to number of touches a player has, to see how regularly a player gives the ball away when they have it.

This is where it becomes more Liverpool related as you ask. When we output possession lost this way, Alexis Sanchez now ranks similarly to Mohamed Salah - they both lose the ball just under a third of the time they receive it (again, they're attacking players, so this is inevitable. ) But nobody could ever say that Salah is playing poorly. There are around 50 outfield players who lose possession more often in relation to their touches than both Sanchez and Salah (minimum 250 touches) .

Again, this is a good example of where context around a stat paints a different picture to what the media perhaps want to. Sanchez is being vilified for a stat where he ranks similarly to Kevin De Bruyne when taken at it's face value, and with Mo Salah when taken alongside touches. The two frontrunners for the player of the season. So why is that used to say Sanchez is playing poorly? The fact is, it shouldn't, and it's a terrible misinterpretation.

Hope that's a good enough example! If you can think of any theory you would like backed up, or disproved even, by stats, let me know and I can always have a look!

15.) 03 Apr 2018
03 Apr 2018 12:35:28
That is bloody brilliant raid.
Eds is always say don’t look at the stats. And that just confirmed it.
You should do an article about it so everyone can read it not just peeps in this feed 👍🏻.

16.) 03 Apr 2018
03 Apr 2018 12:55:31
Great information again Reid, and thanks for the reply. As you say, that ‘giving away possession’ statistic is nonsense in isolation.

The previous example (about quality of chances) leads to a further conclusion about the play itself (i. e. that the goals conceded are more frequently down to individual errors in Liverpool’s case than with a lot of other teams), and that in turn might help with analysing the team. E. g. having established that we commit more errors, is that to do with the individual players, or is it because of (e. g. ) a high press from the midfield leaves the defence exposed if we concede possession?

Impressionistically, I might think that both things are true (the high press without a designated DM is a calculated risk, and the defenders are a bit error-prone), but there are certainly posters on here like Ed001 or MK Scouser who would be able to analyse that example and set me straight without the statistical analysis. What I don’t know is whether the kind of analysis you provide would be able to contribute to that debate in a way that helps to establish an answer. It’s just an example, not sure if it’s a good one.

Anyway great to have you on the site Reid. I love Liverpool but know nothing really about tactics etc, so it’s always good to see someone with a different angle on things, especially when they’re backing up what they’re saying with solid evidence. Keep going mate!

17.) 03 Apr 2018
03 Apr 2018 15:28:24
Thanks for the comments guys - really is nice to read!

Since I do work for Opta, I can't go with the 'don't look at the stats' mantra because then I'd be out of a job!

But my role is to make sure those stats are relevant and used in good context. The wealth of stats we have at Opta is unbelievable and can be so overwhelming, so it's important to have your own opinion on something before you look for the stats on it. Your own personal understanding is so important.

I think the key lesson is not to just take those stats that the likes of Sky throw at you at face value (which they do obtain from us, but will often ignore the context we give to them) . Question what it really means - is it really relevant?

I'd be more than happy to do an article on this at some point when I get the time. Tbh, I'd absolutely love to discuss it a bit with Ed001, because of course I know he hates stats that are given out of context or with little to no relevance (as do I! ), so it would be really interesting to have a conversation about it like we've touched on in this thread.

How do I go about posting an article Eds? I'd love to do it once I get the chance if it seems like people are interested in it!

{Ed033's Note - Send it as a post using a posting form.