BuzzJack
Entertainment Discussion

Welcome Guest ( Log In | Register | Help )

21 Pages V  « < 19 20 21  
Reply to this topicStart new topic
> Premier League 2019/2020 Part 2
Track this topic - Email this topic - Print this topic - Download this topic - Subscribe to this forum
Houdini
post Aug 2 2020, 11:25 AM
Post #401
BuzzJack Platinum Member
******
Group: Members
Posts: 5,726
Member No.: 21,725
Joined: 4-April 15
   No Gallery Pics
 


I would be surprised if Bournemouth make an immediate return to the Premier League now. They will probably be in The Championship for a number of years, a part of me even suspects that they could go down further to League One if there is a mass exodus at the club.
Go to the top of the page
 
+Quote Post
Suedehead2
post Aug 2 2020, 01:42 PM
Post #402
BuzzJack Legend
*******
Group: Veteran
Posts: 30,231
Member No.: 3,272
Joined: 13-April 07
   No Gallery Pics
 


Somebody has calculated a Premier League table based on the number of pounds per player spent by each club on wages. Sheffield Utd finish well clear at the top with an average of under £13,500. The other Champions League places go to Burnley (£35,500), Wolves (£37,250) and Brighton (£44,000). Arsenal (£85,500), Man City (£86,250) and Man Utd (£93,000) finish in the bottom three. Liverpool (£56,000) are in mid-table.
Go to the top of the page
 
+Quote Post
Dobbo
post Aug 3 2020, 08:57 AM
Post #403
BuzzJack Legend
*******
Group: Members
Posts: 22,508
Member No.: 20,053
Joined: 4-November 13
   No Gallery Pics
 


Howe signing Ibe & Solanke for a combined £34M should be a crime against humanity.
Go to the top of the page
 
+Quote Post
Rooney
post Aug 3 2020, 09:11 AM
Post #404
WINTER IS COMING
*******
Group: Global Mod
Posts: 42,118
Member No.: 88
Joined: 7-March 06
 


QUOTE(Dobbo @ Aug 3 2020, 09:57 AM) *
Howe signing Ibe & Solanke for a combined £34M should be a crime against humanity.


Ibe was criminal, but Solanke was the type of player he should have been going for. It's just a shame it took 18 months for him to score a goal. If he scores 20 goals in the Championship and gets them promoted does that mean he was a good signing? Who knows.
Go to the top of the page
 
+Quote Post
Suedehead2
post Monday, 12:48 PM
Post #405
BuzzJack Legend
*******
Group: Veteran
Posts: 30,231
Member No.: 3,272
Joined: 13-April 07
   No Gallery Pics
 


Somebody has recalculated the Premier League table based on expected goals. Quite how they have calculated it, I don't know but somehow they have produced a table with fractions of points. I would have simply rounded the expected goals to the nearer whole number and used that as the result. If anyone wants to do that, feel free laugh.gif

Anyway, it shows Man City well clear at the top with Liverpool only just getting second ahead of Chelsea and Man Utd. Newcastle finish bottom with Norwich and Aston Villa also relegated.

https://understat.com/league/EPL
Go to the top of the page
 
+Quote Post
RabbitFurCoat
post Monday, 01:42 PM
Post #406
You brought the sun to guide you through
*******
Group: Members
Posts: 23,694
Member No.: 118
Joined: 8-March 06
   No Gallery Pics
 


I suppose you could round up the final table to a whole number but generally it wouldn’t be as expected points is calculated based on probability of what would happen from a match based on its expected goals through 1000s of simulations. If there’s a game where the xG is 2-1 that would result in a home win most of the time but not always, as you’d see by looking at a set of matches on any given weekend. That’s mainly because football is very low scoring and therefore small amounts of variance has a big influence. It’s not uncommon to watch your team and think you’ve been unlucky/lucky based on the result and the chances that were created.

If you simulate a 2-1 on xG 100 times you may end up with something along the lines of 65 home wins, 25 draws and 10 away wins. Therefore the expected points would be (65*3) + (25*1) + (10*0) /100 for the home team and the reverse for the away team.

On an xP table Team A is awarded 2.20 points and Team B 0.55 points.

Another way of looking at it is if Team A played 100 games where they were twice as dominant as their opponent they’d likely get 220 points and their opponents win 55. I’ve no idea how accurate that would be, just chose easy numbers that I could calculate in my head!

Can you tell I’ve recently read The Expected Goals Philosophy? laugh.gif
Go to the top of the page
 
+Quote Post
Rooney
post Monday, 01:59 PM
Post #407
WINTER IS COMING
*******
Group: Global Mod
Posts: 42,118
Member No.: 88
Joined: 7-March 06
 


QUOTE(Suedehead2 @ Aug 10 2020, 01:48 PM) *
Somebody has recalculated the Premier League table based on expected goals. Quite how they have calculated it, I don't know but somehow they have produced a table with fractions of points. I would have simply rounded the expected goals to the nearer whole number and used that as the result. If anyone wants to do that, feel free laugh.gif

Anyway, it shows Man City well clear at the top with Liverpool only just getting second ahead of Chelsea and Man Utd. Newcastle finish bottom with Norwich and Aston Villa also relegated.

https://understat.com/league/EPL


Expected Goals is the baby of football analytics, Ted Knutson did a piece on it and I've seen a few Liverpool based blogs doing some work on it. I think the algorithm and analytics at the pioneering of football are moving away from xG as a metric, or at least it only tells a small piece of the tale. What it does not take in to consideration is the caliber of the goalkeeper or the quality of the person finishing the chances. It only touched on it very briefly but essentially the reasoning why xG is quite out of control (make no mistake the table goes WELL against xG as a performance metric) is that when Liverpool score first they control games and conserve energy, whereas Man City tend to do the opposite.

I can't even begin to understand football analytics at the depth the top teams do, but clearly something is wrong with xG in it's purest form to be so out.
Go to the top of the page
 
+Quote Post
RabbitFurCoat
post Monday, 02:20 PM
Post #408
You brought the sun to guide you through
*******
Group: Members
Posts: 23,694
Member No.: 118
Joined: 8-March 06
   No Gallery Pics
 


I think it works well in isolation of the measure of a specific shot based on location, strong foot, type of assist, keeper location etc. but there are always ways where it gets innovated as more people become more understanding and technology advances too.

One of the problems of it in looking at specific matches is game state, as you say some teams may sit off a bit more when taking the lead and may play in more defensive way that allows the opponent half a dozen attempts, but if their 1 goal came later in the game those half dozen attempts are far less likely to have been allowed so have been ‘punished’ by taking the lead early. Man City have certainly given out a lot more thrashing over the season so I don’t find it surprising that they’d be ahead in an xP table

I don’t think there’s that much of an issue with regards to the players taking the shots, there are very few who outperform their xG on any kind of consistent or long term basis.
Go to the top of the page
 
+Quote Post
Dobbo
post Monday, 03:00 PM
Post #409
BuzzJack Legend
*******
Group: Members
Posts: 22,508
Member No.: 20,053
Joined: 4-November 13
   No Gallery Pics
 


Never been a fan of xG, xA or xP. Not a good indicator imo, look at Leicester winning the league with an average possession of like 42% or something. Can't be fluke cos they done it over an entire season so...
Go to the top of the page
 
+Quote Post
Suedehead2
post Monday, 03:01 PM
Post #410
BuzzJack Legend
*******
Group: Veteran
Posts: 30,231
Member No.: 3,272
Joined: 13-April 07
   No Gallery Pics
 


There was a very good explanation of how expected goals work in the Royal Institution Christmas Lectures last year. From what Mark says, it seems they then applied a second probability model to calculate the points so thanks to him!
Go to the top of the page
 
+Quote Post
Suedehead2
post Monday, 03:06 PM
Post #411
BuzzJack Legend
*******
Group: Veteran
Posts: 30,231
Member No.: 3,272
Joined: 13-April 07
   No Gallery Pics
 


QUOTE(Rooney @ Aug 10 2020, 02:59 PM) *
Expected Goals is the baby of football analytics, Ted Knutson did a piece on it and I've seen a few Liverpool based blogs doing some work on it. I think the algorithm and analytics at the pioneering of football are moving away from xG as a metric, or at least it only tells a small piece of the tale. What it does not take in to consideration is the caliber of the goalkeeper or the quality of the person finishing the chances. It only touched on it very briefly but essentially the reasoning why xG is quite out of control (make no mistake the table goes WELL against xG as a performance metric) is that when Liverpool score first they control games and conserve energy, whereas Man City tend to do the opposite.

I can't even begin to understand football analytics at the depth the top teams do, but clearly something is wrong with xG in it's purest form to be so out.

Yes, the piece I read (which referenced Ted Knutson) suggested that the calculations took no account of the identity of the goalkeeper. Therefore, a team with an outstanding goalkeeper would, more often than not, concede fewer goals in a match than xG would suggest. It also mentioned the fact that some teams are more inclined to keep going when 3-0 up than others. When Arsenal were at their peak, they developed a habit against weaker teams of going 3-0 up before half-time and taking it easy for the rest of the match.
Go to the top of the page
 
+Quote Post
RabbitFurCoat
post Monday, 04:50 PM
Post #412
You brought the sun to guide you through
*******
Group: Members
Posts: 23,694
Member No.: 118
Joined: 8-March 06
   No Gallery Pics
 


QUOTE(Dobbo @ Aug 10 2020, 04:00 PM) *
Never been a fan of xG, xA or xP. Not a good indicator imo, look at Leicester winning the league with an average possession of like 42% or something. Can't be fluke cos they done it over an entire season so...


But xG is nothing to do with possession?

Expected goals is basically an improvement of shots/shots on target, which have always been used and shown and fans have always used to show they dominated a game, or moaned about when they had 10 more shots than the opposition but lost. Rather than measuring how many shots are taken, xG measures the quality of those shots, and quality of shot is surely more important than the quantity of them.

xG models have databases of 100,000s of shots from all levels at the game that can then accurately calculate the chance of a shot going in from that location based on thousands of others that have been taken from the same place. The simplest one to calculate is a penalty - it will be given 0.76xG based on the fact that historically 76% of penalties have been scored.

A limitation of xG is that it doesn't take into account who's taking the shot, so Man City's xG will be the same whether it is John Stones or Sergio Aguero taking the shot, but even then it's surprising just how accurate it is and how close the best finishers actual goals are to their expected goals and how it regresses when a player goes through a purple patch. In the last 6 seasons, Cristiano Ronaldo, Robert Lewandowski and Sergio Aguero (3 players considered world class, particularly at finishing) have a combined xG (in the league) of 470.69. They've scored 476 goals. It's remarkably accurate, and the reason they score so many goals isn't because they're lethal, it's because they can get into positions to enable quality players to create them multiple excellent chances.


This post has been edited by RabbitFurCoat: Monday, 04:51 PM
Go to the top of the page
 
+Quote Post
Suedehead2
post Monday, 06:33 PM
Post #413
BuzzJack Legend
*******
Group: Veteran
Posts: 30,231
Member No.: 3,272
Joined: 13-April 07
   No Gallery Pics
 


The BBC doesn't show xG in FA Cup matches. Is that because the difference in level of the teams could make the calculation too inaccurate? For example, I would expect a top Premier League side to have a higher than 76% chance of converting a penalty against a team from League Two.
Go to the top of the page
 
+Quote Post
Dobbo
post Tuesday, 10:44 AM
Post #414
BuzzJack Legend
*******
Group: Members
Posts: 22,508
Member No.: 20,053
Joined: 4-November 13
   No Gallery Pics
 


QUOTE(RabbitFurCoat @ Aug 10 2020, 05:50 PM) *
But xG is nothing to do with possession?

Expected goals is basically an improvement of shots/shots on target, which have always been used and shown and fans have always used to show they dominated a game, or moaned about when they had 10 more shots than the opposition but lost. Rather than measuring how many shots are taken, xG measures the quality of those shots, and quality of shot is surely more important than the quantity of them.

xG models have databases of 100,000s of shots from all levels at the game that can then accurately calculate the chance of a shot going in from that location based on thousands of others that have been taken from the same place. The simplest one to calculate is a penalty - it will be given 0.76xG based on the fact that historically 76% of penalties have been scored.

A limitation of xG is that it doesn't take into account who's taking the shot, so Man City's xG will be the same whether it is John Stones or Sergio Aguero taking the shot, but even then it's surprising just how accurate it is and how close the best finishers actual goals are to their expected goals and how it regresses when a player goes through a purple patch. In the last 6 seasons, Cristiano Ronaldo, Robert Lewandowski and Sergio Aguero (3 players considered world class, particularly at finishing) have a combined xG (in the league) of 470.69. They've scored 476 goals. It's remarkably accurate, and the reason they score so many goals isn't because they're lethal, it's because they can get into positions to enable quality players to create them multiple excellent chances.


Interesting read, I suppose I hadn't looked into it to quite an extent. The pens one makes sense yep and that is surprisingly accurate re those top 3 goals tally.
Go to the top of the page
 
+Quote Post


21 Pages V  « < 19 20 21
Reply to this topicStart new topic

2 User(s) are reading this topic (2 Guests and 0 Anonymous Users)
0 Members:


 

Lo-Fi Version Time is now: 14th August 2020 - 02:12 AM