Jump to content

A statistical analysis of our start to the season


Recommended Posts

8 minutes ago, OwlinOldham said:

As an example, if the stats showed were were good at creating chances, but terrible at finishing these, we could then at least be on the look out for a player with a great abilty to finish chances above all other parts of his game.


We just had one of those, supposedly… it didn’t work out too well!

Link to comment
Share on other sites

Guest LondonOwl313
55 minutes ago, Sultan_Pepper said:

This is my main issue with the whole expected goals thing.

 

There was a very famous but if research back in the late seventies that stated that a large percentage of goals were conceded within 2 touches of the opposition last  the ball.

 

One interpretation at the time was that you should get the ball up the pitch with one big kick after your defence get it and long ball football was born and it set England back as a footballing nation by about 30 years.

 

Years later the same piece of research was looked at again but thought was given to areas of the pitch and the high press style of play the klopp and co play came to fruition.

 

In between those two times there was a conference where lots of managers like Rafa benitez were shown similar research based on player possessions since getting the ball rather than touches and the focus shifted for a few years with Brazil of all teams adopting a counterattacking style.

 

Expected goals is all the rage but I think it's looking in the wrong area, or at least it's being interpreted the wrong way. With expected goals George Graham's Arsenal side were rubbish, but they won loads. Mourinho's Chelsea were sub par but won the league.

 

Different styles of play date very badly under expected goals, particularly teams that cross the ball a lot.

 

With expected goals Chris Waddle could swing in a wonderful cross but if Hirst or Bright is inches away from getting their head on it it doesn't even register in the calculations but Darren Potter shooting over the roof of the stand from 40 yards does.

To be honest with that example I don’t see how expected goals wouldn’t still ‘work’. The Darren Potter shot is going to have an expected goals of about 0.02xG so if he does that 5 times a match it would add 0.1xG. The Waddle cross.. yeah if the cross is too high and the striker just misses it it’s a zero.. and the more crosses put in the greater the chance of one being good. For the ones that do connect and the striker gets a clean header from 6 yards out the xG might be 0.6, so it still works even if only counting the clean crosses. 

Link to comment
Share on other sites

13 hours ago, Morepork said:

The power of stats.

 

 

great watch 👏- and absolutely get this looking at a season it points to us last season - if our stats after 15 games are like now then you would be worried ... but first 10 games are so unpredictable ... also look at the quality of officials in L1/L2 which I am sure must completely screw up everything too!

 

  • Like 1
Link to comment
Share on other sites

Guest LondonOwl313
3 minutes ago, S25OWL said:

great watch 👏- and absolutely get this looking at a season it points to us last season - if our stats after 15 games are like now then you would be worried ... but first 10 games are so unpredictable ... also look at the quality of officials in L1/L2 which I am sure must completely screw up everything too!

 

I don’t think it’s anything to worry about, we’re doing fine.

 

Our expected goals for is between 1.1 and 1.2 depending on the source.. the 0.62 has to be wrong as we’ve had a few tap ins. Defensively excluding the pens it’s a very low number.. so it kind of reflects where the team is at. Defensively very strong, attacking wise it’s a work in progress and those numbers will increase as we improve.

 

https://understat.com/league/EPL

 

Most PL teams near the top seem to average around 2 xG per game so if we’re going to be right at the top of our league that’s what you’d expect 

Link to comment
Share on other sites

19 hours ago, CircleSeven said:

I can see how it makes sense to look at these stats. It’s about looking at deficiencies in your game that are sometimes masked by results so that you can improve your chances of winning over the long term. If you have more good chances over the long term you will get more goals. We beat Rotherham 2-0, which was great. But a similar performance against a different team could result in a draw or a loss.
 

That doesn’t mean the win and the points against Rotherham weren’t brilliant, just that the coaching staff might want to work on things that give us more chances of scoring more frequently so we are less reliant on say a penalty save or a goalkeeping howler. 
 

As long as the deficiencies in our game carry on being masked by our results, we will be OK lol

  • Like 1
Link to comment
Share on other sites

12 hours ago, Morepork said:

Posters suggesting bookmakers don’t use stats….also rounduns.

I'm sure they do, but the same stats are available for the punter.

 

Just a hunch, but if 50 people here predicted the results of say 10 games then the 10 most popular predictions of each game used it would be just as accurate as any statistical analysts for the same matches.

Basically what the bookies do to form a book.

 

What statisticaly is more or less constant is.

50% of home teams will win

25% will be away wins

And 25% will be draws.

 

Robotic stop start American sport is not a good comparison imho.

 

I'm sure measuring work rate etc could could be useful for coaches though.

 

But what is more useful for a teams success...a lazy but very skilful striker or a striker who never stops running but couldn't hit a barn door from 5yds.

 

The best guide of all is the league table coupled with a manager who knows the game.

Aimho of course🤔

 

 

 

 

  • Like 1
Link to comment
Share on other sites

Moneyball won't work in league 1. Fans, especially at away games can have a huge effect to the outcome. Also the mental attitude of players is tangible but can't be measured in the same way as these figures but are arguably more important.

Its the whole thing coming together we need to worry about and not a stat that, as we have shown so far, means absolutely nothing to our main aim of picking up more points than anyone else in the league. 

  • Like 1
Link to comment
Share on other sites

1 hour ago, LondonOwl313 said:

To be honest with that example I don’t see how expected goals wouldn’t still ‘work’. The Darren Potter shot is going to have an expected goals of about 0.02xG so if he does that 5 times a match it would add 0.1xG. The Waddle cross.. yeah if the cross is too high and the striker just misses it it’s a zero.. and the more crosses put in the greater the chance of one being good. For the ones that do connect and the striker gets a clean header from 6 yards out the xG might be 0.6, so it still works even if only counting the clean crosses. 

 

Don't disagree but 20 balls across like the one for Gazza in Euro 96 in a match would register as no likelihood to score whereas any football fan watching a match where your team fizzes 20 balls across the six yard box would be watching a really good attacking performance.

 

I'm sure the full equations etc have more detail than I'm aware of but I'm just not convinced that the conclusions drawn from it are always correct.

 

Link to comment
Share on other sites

2 minutes ago, Sultan_Pepper said:

 

Don't disagree but 20 balls across like the one for Gazza in Euro 96 in a match would register as no likelihood to score whereas any football fan watching a match where your team fizzes 20 balls across the six yard box would be watching a really good attacking performance.

 

I'm sure the full equations etc have more detail than I'm aware of but I'm just not convinced that the conclusions drawn from it are always correct.

 

You’re right, xG only counts the likelihood of scoring shots so those scenarios would make it daft. I think the volume of those kinds of situations over the course of the game would rarely be so high to completely invalidate the model though. It works the other way round too, Gregory’s goal against Rotherham probably had a high xG due to how close he was to the goal / the fact he was one on one with the keeper. All of that hides the fact that 99% of the time the chance doesn’t happen because the keeper holds on to the ball. 

 

The point is that it gives a better idea of the way a match went than the score on its own, or things like shots etc but it shouldn’t be used on its own. 

  • Like 1
Link to comment
Share on other sites

2 hours ago, beswetherick said:


We just had one of those, supposedly… it didn’t work out too well!

 

I get what your saying, i just meant that it would be interesting to see how we were actually performing throughout that time in regards to things like chances created etc. would the stats have actually shown we didnt need this type at all, or that we werent playing anywhere near well enough to get the best form such a player? 

 

I just think it would be good to know and find out if it was actually just the player being poo, or was it due to the way that the team played and things like low chances creation that also effected the abilty for the player to excel. I expect we will never get the answer, but it could maybe shine a light on things not being as balck and white as said player just being solely to blame. 

 

 

 

Link to comment
Share on other sites

On 23/08/2021 at 15:25, tcuc3e said:

It's great to be top of the league, unbeaten and without conceding a goal.  Ultimately those are the stats that matter and if they continue then HMS PSL will be full steam ahead.  However the odds are that it won't be quite that straightforward over the next 40+ games.

 

For a slightly more detailed look at the performances so far I took a look at the xG stats.  For those that don't know xG stands for expected goals scored per game. It's calculated by taking into account shot positions, shot accuracy (on/off target), shot frequency (number of shots), attack dangerousness and overall attack pressure (possession amount and depth of possession).

 

It's generally understood to provide a more detailed assessment of how a team is playing than the win/loss column, even if that is all that matters at the end of the day.  Top teams use it to assess their performances (and it's used a lot in fantasy football).

 

So far Wednesday are:

 

18th in xG with 0.62 per game.

 

I think many have pointed out that despite picking up points we've not actually been clicking on all cylinders yet and this would back that up. Certainly areas to improve and as the squad gels hopefully better performances will come.

 

A note about Saturday.  That game will be between the worst side in xG at Home, Morecambe, with 0.37 per game against the us, the second worst side in Away xG at 0.32.   It might not be a classic for the neutrals! 

 

 

 

 


Good post, a rare analytical thread in Owlstalk…all for that.

 

Judging by how DM has set us up starting the season I cant imagine we’ll climb too high in the xG.
 

The good news for us, at this level we should have significantly better quality at the back and in the box than most teams. Sure we might be creating less, but I’d imagine our conversion % of shots on target/goals is quite high? We are going to be very efficient & tactical in how we play. 
 

Keen to see what these look like after 10+ games when teams have settled on strongest line ups etc. I remember diving down the statistical rabbit hole with Monk & Pulis…. bottom on just about every advanced metric and then rock bottom when Pulis left. Depressing. But looking forward to something completely different under DM when we have more data. 

Edited by Lionel Fessi
  • Like 1
Link to comment
Share on other sites

22 minutes ago, areNOTwhatTHEYseem said:

If you adjust the table to account for xG in each game last season, we finish...one place higher.

 

:duntmatter:

 

2021-05-08-101-1.png.65372e944a0bc371ebffd82617b3ee87.png

 

There are a few anomalies in that table like Swansea, Bristol, Blackburn and Rotherham - but it is basically close enough. We were very bad for 3/4s, although I though Moore was unlucky towards the end in one or two games like Rotherham and QPR. To my untrained eye Norwich and Brentford were the two best teams in the league and the stats seemed to back it up (although Watford sneaked above Brentford in the table, the thing that matters most of all). 

  • Like 1
Link to comment
Share on other sites

14 minutes ago, Emerson Thome said:

There are a few anomalies in that table like Swansea, Bristol, Blackburn and Rotherham - but it is basically close enough. We were very bad for 3/4s, although I though Moore was unlucky towards the end in one or two games like Rotherham and QPR. To my untrained eye Norwich and Brentford were the two best teams in the league and the stats seemed to back it up (although Watford sneaked above Brentford in the table, the thing that matters most of all). 

 

It does suggest Bristol City will be lucky to escape a relegation fight again unless they've significantly altered their approach.

 

Of course, clubs have access to all this stuff, so they've probably done just that!

 

:duntmatter:

Link to comment
Share on other sites

On 23/08/2021 at 16:13, 0742 said:

 

 

I did this course a while back, found it very interesting and well worth the fee to do all of the courses in the programme too!

 

https://www.coursera.org/specializations/sports-analytics?

 

 

@0742 Being a bit of a data geek, I've signed up and enjoyed working though this - thanks for the details.  Its quite a undertaking, as there is a lot of content to get through, but very interesting so far.  I like the main presenter but the Chinese girl is hard to follow, so just ignored her and worked through the notebooks myself.

 

Note, with CourseRa, you don't need to pay for these courses, you can audit for free and get all the content other than the solutions to practicals/quizzes (and no certificate).

  • Love 1
Link to comment
Share on other sites

20 hours ago, AsioOtus said:

 

@0742 Being a bit of a data geek, I've signed up and enjoyed working though this - thanks for the details.  Its quite a undertaking, as there is a lot of content to get through, but very interesting so far.  I like the main presenter but the Chinese girl is hard to follow, so just ignored her and worked through the notebooks myself.

 

Note, with CourseRa, you don't need to pay for these courses, you can audit for free and get all the content other than the solutions to practicals/quizzes (and no certificate).

 

Glad to hear you enjoyed it mate! I found the same with the Chinese girl presenter too.

  • Like 1
Link to comment
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
×
×
  • Create New...