Help
Maybe we will all have to wait and see how many clubs go to the wall before any
leagues regional or not are formatted?
Not a nice thought but very real
Accy chairman already worried about closure.
National League
#402
hewittfan 
- Reserve Team Player
-
- Group: Members
- Posts: 1,454
- Joined: 09-March 09
- Gender:Male
- Location:Newbold
Posted 24 April 2020 - 01:19 PM
clarevoyant., on 24 April 2020 - 09:16 AM, said:
Very true, on the plus side with the takeover unlikely to happen, maybe (what am I saying hear) just maybe, DA becomes all benevolent in appeasement for passed 'errors' and we could find ourselves in a far better position than the majority of clubs in our division. Thus putting us on the road to becoming a saleable commodity.
Yes I know I'm in need of some medicinal gargles.
Yes I know I'm in need of some medicinal gargles.
Just don't get them from Donald Trump.
#403
JonB
- Legend
-
- Group: Members
- Posts: 31,755
- Joined: 22-February 06
Posted 24 April 2020 - 04:04 PM
Other countries starting to make decisions, Germany looking to start the Bundesliga back up behind closed doors in a couple of weeks whilst Holland have ended the seaosn with Ajax not awarded the title despite being top plus no promotion or relegation.
#404
CFC91
- Legend
-
- Group: Members
- Posts: 11,060
- Joined: 23-May 11
- Gender:Male
- Location:Grassmoor
Posted 24 April 2020 - 05:52 PM
Spain and potentially Italy won’t be a million miles behind. Maybe end of May early June. Our own PL are still looking at 6th June to resume. Feels very ambitious but that’s what are told to work too.
I’m sure it’ll likely be a week or two later however
I’m sure it’ll likely be a week or two later however
#406
moondog
- Legend
-
- Group: Moderators
- Posts: 27,037
- Joined: 09-June 05
- Gender:Male
- Location:Chesterfield
Posted 25 April 2020 - 08:39 AM
A statistical solution (for the Premier League) which is close to Dim View's thoughts (sorry unable to reproduce the graph)
The maths that shows leagues shouldn’t be decided on points per game
https://theathletic....-shared-article
By Tom Worville
UEFA’s stance on completing this paused season sees the sporting body favour the conclusion of competitions where possible, and where not, to have clubs selected for their various continental cups in 2020-21 based on what they’re deeming “sporting merit”.
While there is no set definition of how to numerically calculate sporting merit, it’s assumed that points per game will be the go-to. The reason? It’s easy to calculate and is as close to how tables currently stand as you’re going to get. If all teams had played the same number of games, you’d use the league table as is, but in most leagues they won’t have, hence the suggestion to use points per game.
So, if the Premier League for 2019-20 gets decided on points per game, what would that look like?
No surprise that Liverpool would win the title. The points-per-game difference between themselves and Manchester City (0.79) is greater than that from Norwich City at the foot of the table and seventh-placed Wolverhampton Wanderers (0.76). There’s also very little movement elsewhere, the only changes are Sheffield United leapfrogging Wolves and Tottenham Hotspur being told to “mind the gap” by Arsenal Twitter for the first time in four years.
Using this approach, there’s no change for the teams currently in the relegation places. Sorry Bournemouth fans, your team have been relegated due to goal difference per game — sadly this season hasn’t seen you muster enough “sporting merit”.
Yet, points per game is not what sporting merit should be derived from. At all.
For a start, it doesn’t take into account the strength of the run-in each team has left. The matrix below shows the games all the clubs have outstanding, with both axis sorted by current league position. This gives a picture of how hard everybody’s remaining games are — something that is so important to take into consideration if a league needs to be concluded statistically. Clubs should really use that as the first argument against using points per game.
To understand this graphic, look at each team’s row from left to right. The more deep-red squares that team has near the left-hand side (ie, against teams near the top of the table), the harder their run-in is. The more light-red squares they have nearer the right-hand side, the easier their run-in is.
There’s plenty to unpack here.
For a start, focus on the three relegation-threatened teams. Norwich are comfortably bottom of the table in terms of their points per game, but their run-in is markedly easier on paper than those of Aston Villa and Bournemouth. Note how they still have to play the four teams directly above the relegation zone, and only two of the top four. Bournemouth and Villa, on the other hand, have plenty of teams in the top half left to play.
It’s not out of the question that Watford get dragged back into the relegation fight either as their run-in sees them having to play Manchester City, Leicester City and Chelsea. Although their shock result against Liverpool at the end of February might give them hope they have enough to get something from those games and stay up.
While not the most thrilling battle this season, also take a look at how Arsenal’s and Spurs’ fixtures differ. While their arch-rivals’s are more spread out — facing all of the top three and the current bottom two — Tottenham face more mid-table sides, with Newcastle United, Everton and Crystal Palace yet to play Mourinho’s men.
This is a very basic look at why points per game is a poor measure, however. The league table can lie, for a start, and current position is not a perfect measure of team strength. It also doesn’t take into account the ordering of when these fixtures take place, both factors an actual robust model would take into account.
One such model is the one created by Opta.
It calculates how good each team is from an attacking and defensive perspective from their previous results. To measure attack and defence, two numbers are derived based on goals scored and goals conceded in historical matches. The attacking coefficient shows how good a team is at scoring goals, while the defensive coefficient signals a team’s strength at preventing goals. Using these coefficients together, alongside those of a given opposition and other required context (such as who is playing at home), Opta can simulate the outcome of a single match.
Let’s look at how this works with an example.
Say that the home team’s attacking coefficient is 2.1 (that is, in an average game, we’d expect them to score 2.1 goals) and their defensive coefficient is 1.1 (that is, they concede 1.1 goals on average). The away team’s respective coefficients are 0.3 (they can barely score goals) and 1.8 (they ship plenty of them). From this information, we can “play” the match to see what the scoreline is.
The playing of the game isn’t quite as fun as a virtual football match might sound and is not too dissimilar to drawing coloured balls out of a bag, like you might have learned in an introductory stats lesson at school. The colour of the balls in the bag relate to the number of goals scored, and the proportion of each ball is dictated by the attacking coefficient combined with the opponent’s defensive coefficient.
In our example, the home team have a good attack, and the visitors an awful defence, so the balls in the bag for the home team are likely to be of higher value, meaning they’ll likely pick a ball that represents a high number of goals. When the away team draws from the bag, they’ll likely pick a ball worth zero goals on average.
Anyway, let’s say that from this bag we draw two goals for the home team, and one for the away team. Our result, therefore, is 2-1 — and the home team has won. Do that plenty of times, over lots of games, and you can simulate full seasons and be able to calculate the number of times a given team finishes a league in each position.
The matrix below shows exactly the results of the process above, but Opta didn’t have to draw any balls out of a bag for it.
Taking into account the strengths of the teams from historical data and simulating the season 10,000 times, it shows the proportion of times each team finishes in various places in the table. Norwich, Villa and Bournemouth all get relegated more often than not in the 10,000 “seasons”, Liverpool win the title and Spurs leap up to sixth. Note that these predictions are from early April — taking into account the table positions of all teams before the COVID-19 shutdown.
Compared to points per game, this model takes into account schedule strength, and uses attacking and defensive coefficients that are far more accurate means of measuring sporting merit (and team strength, which is ultimately what we should care about here).
Statistical smarts and a sound model are one thing, but the process of getting results and deciding the final table isn’t really grounded in reality: we don’t live in a world where things happen “on average”.
Take Leicester’s 5,000-1 title winning season four years ago. That 5,000-1 figure was calculated by the bookies, whose approach isn’t far off what was done above, but will be informed with more information to tweak these attacking and defensive coefficients based on players available, form, suspensions and the like. In probability terms, they gave Leicester a 0.02 per cent chance of winning the 2015-16 title. So if this was a roulette wheel with 10,000 pockets, Leicester occupied just two of them.
If we simulated that season instead of playing it, we would have seen Leicester, over 10,000 roulette wheel spins, win the league twice. On average, though, the title would likely have gone to Manchester City (who were the favourites at the time) and who would have had the most pockets on the wheel. The more you spin the wheel, the more you tend to arrive at the more “likely” outcome.
Life, however, is a game of chance, and at times, so is football.
For that reason alone, the best way to resolve a stagnant league is by spinning the roulette wheel, not counting up how many pockets a given team has.
The better way to resolve this situation would be to find a model that is agreed to be the best and pick one of the 10,000 simulations as the result.
Whether this is done or not is unlikely, but presented above are plenty of reasons why points per game is a poor measure of resolving a league season, and hopefully UEFA and others soon realise the same.
The maths that shows leagues shouldn’t be decided on points per game
https://theathletic....-shared-article
By Tom Worville
UEFA’s stance on completing this paused season sees the sporting body favour the conclusion of competitions where possible, and where not, to have clubs selected for their various continental cups in 2020-21 based on what they’re deeming “sporting merit”.
While there is no set definition of how to numerically calculate sporting merit, it’s assumed that points per game will be the go-to. The reason? It’s easy to calculate and is as close to how tables currently stand as you’re going to get. If all teams had played the same number of games, you’d use the league table as is, but in most leagues they won’t have, hence the suggestion to use points per game.
So, if the Premier League for 2019-20 gets decided on points per game, what would that look like?
No surprise that Liverpool would win the title. The points-per-game difference between themselves and Manchester City (0.79) is greater than that from Norwich City at the foot of the table and seventh-placed Wolverhampton Wanderers (0.76). There’s also very little movement elsewhere, the only changes are Sheffield United leapfrogging Wolves and Tottenham Hotspur being told to “mind the gap” by Arsenal Twitter for the first time in four years.
Using this approach, there’s no change for the teams currently in the relegation places. Sorry Bournemouth fans, your team have been relegated due to goal difference per game — sadly this season hasn’t seen you muster enough “sporting merit”.
Yet, points per game is not what sporting merit should be derived from. At all.
For a start, it doesn’t take into account the strength of the run-in each team has left. The matrix below shows the games all the clubs have outstanding, with both axis sorted by current league position. This gives a picture of how hard everybody’s remaining games are — something that is so important to take into consideration if a league needs to be concluded statistically. Clubs should really use that as the first argument against using points per game.
To understand this graphic, look at each team’s row from left to right. The more deep-red squares that team has near the left-hand side (ie, against teams near the top of the table), the harder their run-in is. The more light-red squares they have nearer the right-hand side, the easier their run-in is.
There’s plenty to unpack here.
For a start, focus on the three relegation-threatened teams. Norwich are comfortably bottom of the table in terms of their points per game, but their run-in is markedly easier on paper than those of Aston Villa and Bournemouth. Note how they still have to play the four teams directly above the relegation zone, and only two of the top four. Bournemouth and Villa, on the other hand, have plenty of teams in the top half left to play.
It’s not out of the question that Watford get dragged back into the relegation fight either as their run-in sees them having to play Manchester City, Leicester City and Chelsea. Although their shock result against Liverpool at the end of February might give them hope they have enough to get something from those games and stay up.
While not the most thrilling battle this season, also take a look at how Arsenal’s and Spurs’ fixtures differ. While their arch-rivals’s are more spread out — facing all of the top three and the current bottom two — Tottenham face more mid-table sides, with Newcastle United, Everton and Crystal Palace yet to play Mourinho’s men.
This is a very basic look at why points per game is a poor measure, however. The league table can lie, for a start, and current position is not a perfect measure of team strength. It also doesn’t take into account the ordering of when these fixtures take place, both factors an actual robust model would take into account.
One such model is the one created by Opta.
It calculates how good each team is from an attacking and defensive perspective from their previous results. To measure attack and defence, two numbers are derived based on goals scored and goals conceded in historical matches. The attacking coefficient shows how good a team is at scoring goals, while the defensive coefficient signals a team’s strength at preventing goals. Using these coefficients together, alongside those of a given opposition and other required context (such as who is playing at home), Opta can simulate the outcome of a single match.
Let’s look at how this works with an example.
Say that the home team’s attacking coefficient is 2.1 (that is, in an average game, we’d expect them to score 2.1 goals) and their defensive coefficient is 1.1 (that is, they concede 1.1 goals on average). The away team’s respective coefficients are 0.3 (they can barely score goals) and 1.8 (they ship plenty of them). From this information, we can “play” the match to see what the scoreline is.
The playing of the game isn’t quite as fun as a virtual football match might sound and is not too dissimilar to drawing coloured balls out of a bag, like you might have learned in an introductory stats lesson at school. The colour of the balls in the bag relate to the number of goals scored, and the proportion of each ball is dictated by the attacking coefficient combined with the opponent’s defensive coefficient.
In our example, the home team have a good attack, and the visitors an awful defence, so the balls in the bag for the home team are likely to be of higher value, meaning they’ll likely pick a ball that represents a high number of goals. When the away team draws from the bag, they’ll likely pick a ball worth zero goals on average.
Anyway, let’s say that from this bag we draw two goals for the home team, and one for the away team. Our result, therefore, is 2-1 — and the home team has won. Do that plenty of times, over lots of games, and you can simulate full seasons and be able to calculate the number of times a given team finishes a league in each position.
The matrix below shows exactly the results of the process above, but Opta didn’t have to draw any balls out of a bag for it.
Taking into account the strengths of the teams from historical data and simulating the season 10,000 times, it shows the proportion of times each team finishes in various places in the table. Norwich, Villa and Bournemouth all get relegated more often than not in the 10,000 “seasons”, Liverpool win the title and Spurs leap up to sixth. Note that these predictions are from early April — taking into account the table positions of all teams before the COVID-19 shutdown.
Compared to points per game, this model takes into account schedule strength, and uses attacking and defensive coefficients that are far more accurate means of measuring sporting merit (and team strength, which is ultimately what we should care about here).
Statistical smarts and a sound model are one thing, but the process of getting results and deciding the final table isn’t really grounded in reality: we don’t live in a world where things happen “on average”.
Take Leicester’s 5,000-1 title winning season four years ago. That 5,000-1 figure was calculated by the bookies, whose approach isn’t far off what was done above, but will be informed with more information to tweak these attacking and defensive coefficients based on players available, form, suspensions and the like. In probability terms, they gave Leicester a 0.02 per cent chance of winning the 2015-16 title. So if this was a roulette wheel with 10,000 pockets, Leicester occupied just two of them.
If we simulated that season instead of playing it, we would have seen Leicester, over 10,000 roulette wheel spins, win the league twice. On average, though, the title would likely have gone to Manchester City (who were the favourites at the time) and who would have had the most pockets on the wheel. The more you spin the wheel, the more you tend to arrive at the more “likely” outcome.
Life, however, is a game of chance, and at times, so is football.
For that reason alone, the best way to resolve a stagnant league is by spinning the roulette wheel, not counting up how many pockets a given team has.
The better way to resolve this situation would be to find a model that is agreed to be the best and pick one of the 10,000 simulations as the result.
Whether this is done or not is unlikely, but presented above are plenty of reasons why points per game is a poor measure of resolving a league season, and hopefully UEFA and others soon realise the same.
#407
jack bauer
- Key Player
-
- Group: Members
- Posts: 8,913
- Joined: 31-March 10
Posted 25 April 2020 - 10:13 AM
apologies not read the whole thread, for me i'd make a fairly straight forward scenario, no relegation from league but promote barrow to restore the 92. that leaves our league with 23, chorley to be relegated as they are the only side in our league who were nailed on for relegation then promote one team from NLN and NLS, the last bit is the trickiest as no one in those leagues was running away with it.
#408
DerbySpireite59
- DerbySpireite57
-
- Group: Members
- Posts: 2,251
- Joined: 22-April 15
- Gender:Male
- Location:Derby
Posted 25 April 2020 - 10:42 AM
jack bauer, on 25 April 2020 - 10:13 AM, said:
apologies not read the whole thread, for me i'd make a fairly straight forward scenario, no relegation from league but promote barrow to restore the 92. that leaves our league with 23, chorley to be relegated as they are the only side in our league who were nailed on for relegation then promote one team from NLN and NLS, the last bit is the trickiest as no one in those leagues was running away with it.
Is that accounting for recent form (last 3 home and last 3 away games) and respective run-ins? Who did the bookies last make favourite? They're not usually wrong at this stage of a football season.
"No one is 100% good, nor 100% evil"
#409
metallilad
- Legend
-
- Group: Members
- Posts: 11,684
- Joined: 01-May 11
- Gender:Male
- Location:Stonegravels
-
Interests:CFC.
Music (the louder,faster the better).
Posted 25 April 2020 - 11:07 AM
jack bauer, on 25 April 2020 - 10:13 AM, said:
apologies not read the whole thread, for me i'd make a fairly straight forward scenario, no relegation from league but promote barrow to restore the 92. that leaves our league with 23, chorley to be relegated as they are the only side in our league who were nailed on for relegation then promote one team from NLN and NLS, the last bit is the trickiest as no one in those leagues was running away with it.
TBF. Barrow weren't running away with it either but having said that. Your idea seems to be a "make sense" option.
Life goes on. Whatever happens.
#410
clarevoyant.
- Legend
-
- Group: Members
- Posts: 13,677
- Joined: 15-July 09
- Location:If You see me Sat night point me towards Calow
Posted 25 April 2020 - 11:31 AM
moondog, on 25 April 2020 - 08:39 AM, said:
A statistical solution (for the Premier League) which is close to Dim View's thoughts (sorry unable to reproduce the graph)
The maths that shows leagues shouldn’t be decided on points per game
https://theathletic....-shared-article
By Tom Worville
UEFA’s stance on completing this paused season sees the sporting body favour the conclusion of competitions where possible, and where not, to have clubs selected for their various continental cups in 2020-21 based on what they’re deeming “sporting merit”.
While there is no set definition of how to numerically calculate sporting merit, it’s assumed that points per game will be the go-to. The reason? It’s easy to calculate and is as close to how tables currently stand as you’re going to get. If all teams had played the same number of games, you’d use the league table as is, but in most leagues they won’t have, hence the suggestion to use points per game.
So, if the Premier League for 2019-20 gets decided on points per game, what would that look like?
No surprise that Liverpool would win the title. The points-per-game difference between themselves and Manchester City (0.79) is greater than that from Norwich City at the foot of the table and seventh-placed Wolverhampton Wanderers (0.76). There’s also very little movement elsewhere, the only changes are Sheffield United leapfrogging Wolves and Tottenham Hotspur being told to “mind the gap” by Arsenal Twitter for the first time in four years.
Using this approach, there’s no change for the teams currently in the relegation places. Sorry Bournemouth fans, your team have been relegated due to goal difference per game — sadly this season hasn’t seen you muster enough “sporting merit”.
Yet, points per game is not what sporting merit should be derived from. At all.
For a start, it doesn’t take into account the strength of the run-in each team has left. The matrix below shows the games all the clubs have outstanding, with both axis sorted by current league position. This gives a picture of how hard everybody’s remaining games are — something that is so important to take into consideration if a league needs to be concluded statistically. Clubs should really use that as the first argument against using points per game.
To understand this graphic, look at each team’s row from left to right. The more deep-red squares that team has near the left-hand side (ie, against teams near the top of the table), the harder their run-in is. The more light-red squares they have nearer the right-hand side, the easier their run-in is.
There’s plenty to unpack here.
For a start, focus on the three relegation-threatened teams. Norwich are comfortably bottom of the table in terms of their points per game, but their run-in is markedly easier on paper than those of Aston Villa and Bournemouth. Note how they still have to play the four teams directly above the relegation zone, and only two of the top four. Bournemouth and Villa, on the other hand, have plenty of teams in the top half left to play.
It’s not out of the question that Watford get dragged back into the relegation fight either as their run-in sees them having to play Manchester City, Leicester City and Chelsea. Although their shock result against Liverpool at the end of February might give them hope they have enough to get something from those games and stay up.
While not the most thrilling battle this season, also take a look at how Arsenal’s and Spurs’ fixtures differ. While their arch-rivals’s are more spread out — facing all of the top three and the current bottom two — Tottenham face more mid-table sides, with Newcastle United, Everton and Crystal Palace yet to play Mourinho’s men.
This is a very basic look at why points per game is a poor measure, however. The league table can lie, for a start, and current position is not a perfect measure of team strength. It also doesn’t take into account the ordering of when these fixtures take place, both factors an actual robust model would take into account.
One such model is the one created by Opta.
It calculates how good each team is from an attacking and defensive perspective from their previous results. To measure attack and defence, two numbers are derived based on goals scored and goals conceded in historical matches. The attacking coefficient shows how good a team is at scoring goals, while the defensive coefficient signals a team’s strength at preventing goals. Using these coefficients together, alongside those of a given opposition and other required context (such as who is playing at home), Opta can simulate the outcome of a single match.
Let’s look at how this works with an example.
Say that the home team’s attacking coefficient is 2.1 (that is, in an average game, we’d expect them to score 2.1 goals) and their defensive coefficient is 1.1 (that is, they concede 1.1 goals on average). The away team’s respective coefficients are 0.3 (they can barely score goals) and 1.8 (they ship plenty of them). From this information, we can “play” the match to see what the scoreline is.
The playing of the game isn’t quite as fun as a virtual football match might sound and is not too dissimilar to drawing coloured balls out of a bag, like you might have learned in an introductory stats lesson at school. The colour of the balls in the bag relate to the number of goals scored, and the proportion of each ball is dictated by the attacking coefficient combined with the opponent’s defensive coefficient.
In our example, the home team have a good attack, and the visitors an awful defence, so the balls in the bag for the home team are likely to be of higher value, meaning they’ll likely pick a ball that represents a high number of goals. When the away team draws from the bag, they’ll likely pick a ball worth zero goals on average.
Anyway, let’s say that from this bag we draw two goals for the home team, and one for the away team. Our result, therefore, is 2-1 — and the home team has won. Do that plenty of times, over lots of games, and you can simulate full seasons and be able to calculate the number of times a given team finishes a league in each position.
The matrix below shows exactly the results of the process above, but Opta didn’t have to draw any balls out of a bag for it.
Taking into account the strengths of the teams from historical data and simulating the season 10,000 times, it shows the proportion of times each team finishes in various places in the table. Norwich, Villa and Bournemouth all get relegated more often than not in the 10,000 “seasons”, Liverpool win the title and Spurs leap up to sixth. Note that these predictions are from early April — taking into account the table positions of all teams before the COVID-19 shutdown.
Compared to points per game, this model takes into account schedule strength, and uses attacking and defensive coefficients that are far more accurate means of measuring sporting merit (and team strength, which is ultimately what we should care about here).
Statistical smarts and a sound model are one thing, but the process of getting results and deciding the final table isn’t really grounded in reality: we don’t live in a world where things happen “on average”.
Take Leicester’s 5,000-1 title winning season four years ago. That 5,000-1 figure was calculated by the bookies, whose approach isn’t far off what was done above, but will be informed with more information to tweak these attacking and defensive coefficients based on players available, form, suspensions and the like. In probability terms, they gave Leicester a 0.02 per cent chance of winning the 2015-16 title. So if this was a roulette wheel with 10,000 pockets, Leicester occupied just two of them.
If we simulated that season instead of playing it, we would have seen Leicester, over 10,000 roulette wheel spins, win the league twice. On average, though, the title would likely have gone to Manchester City (who were the favourites at the time) and who would have had the most pockets on the wheel. The more you spin the wheel, the more you tend to arrive at the more “likely” outcome.
Life, however, is a game of chance, and at times, so is football.
For that reason alone, the best way to resolve a stagnant league is by spinning the roulette wheel, not counting up how many pockets a given team has.
The better way to resolve this situation would be to find a model that is agreed to be the best and pick one of the 10,000 simulations as the result.
Whether this is done or not is unlikely, but presented above are plenty of reasons why points per game is a poor measure of resolving a league season, and hopefully UEFA and others soon realise the same.
The maths that shows leagues shouldn’t be decided on points per game
https://theathletic....-shared-article
By Tom Worville
UEFA’s stance on completing this paused season sees the sporting body favour the conclusion of competitions where possible, and where not, to have clubs selected for their various continental cups in 2020-21 based on what they’re deeming “sporting merit”.
While there is no set definition of how to numerically calculate sporting merit, it’s assumed that points per game will be the go-to. The reason? It’s easy to calculate and is as close to how tables currently stand as you’re going to get. If all teams had played the same number of games, you’d use the league table as is, but in most leagues they won’t have, hence the suggestion to use points per game.
So, if the Premier League for 2019-20 gets decided on points per game, what would that look like?
No surprise that Liverpool would win the title. The points-per-game difference between themselves and Manchester City (0.79) is greater than that from Norwich City at the foot of the table and seventh-placed Wolverhampton Wanderers (0.76). There’s also very little movement elsewhere, the only changes are Sheffield United leapfrogging Wolves and Tottenham Hotspur being told to “mind the gap” by Arsenal Twitter for the first time in four years.
Using this approach, there’s no change for the teams currently in the relegation places. Sorry Bournemouth fans, your team have been relegated due to goal difference per game — sadly this season hasn’t seen you muster enough “sporting merit”.
Yet, points per game is not what sporting merit should be derived from. At all.
For a start, it doesn’t take into account the strength of the run-in each team has left. The matrix below shows the games all the clubs have outstanding, with both axis sorted by current league position. This gives a picture of how hard everybody’s remaining games are — something that is so important to take into consideration if a league needs to be concluded statistically. Clubs should really use that as the first argument against using points per game.
To understand this graphic, look at each team’s row from left to right. The more deep-red squares that team has near the left-hand side (ie, against teams near the top of the table), the harder their run-in is. The more light-red squares they have nearer the right-hand side, the easier their run-in is.
There’s plenty to unpack here.
For a start, focus on the three relegation-threatened teams. Norwich are comfortably bottom of the table in terms of their points per game, but their run-in is markedly easier on paper than those of Aston Villa and Bournemouth. Note how they still have to play the four teams directly above the relegation zone, and only two of the top four. Bournemouth and Villa, on the other hand, have plenty of teams in the top half left to play.
It’s not out of the question that Watford get dragged back into the relegation fight either as their run-in sees them having to play Manchester City, Leicester City and Chelsea. Although their shock result against Liverpool at the end of February might give them hope they have enough to get something from those games and stay up.
While not the most thrilling battle this season, also take a look at how Arsenal’s and Spurs’ fixtures differ. While their arch-rivals’s are more spread out — facing all of the top three and the current bottom two — Tottenham face more mid-table sides, with Newcastle United, Everton and Crystal Palace yet to play Mourinho’s men.
This is a very basic look at why points per game is a poor measure, however. The league table can lie, for a start, and current position is not a perfect measure of team strength. It also doesn’t take into account the ordering of when these fixtures take place, both factors an actual robust model would take into account.
One such model is the one created by Opta.
It calculates how good each team is from an attacking and defensive perspective from their previous results. To measure attack and defence, two numbers are derived based on goals scored and goals conceded in historical matches. The attacking coefficient shows how good a team is at scoring goals, while the defensive coefficient signals a team’s strength at preventing goals. Using these coefficients together, alongside those of a given opposition and other required context (such as who is playing at home), Opta can simulate the outcome of a single match.
Let’s look at how this works with an example.
Say that the home team’s attacking coefficient is 2.1 (that is, in an average game, we’d expect them to score 2.1 goals) and their defensive coefficient is 1.1 (that is, they concede 1.1 goals on average). The away team’s respective coefficients are 0.3 (they can barely score goals) and 1.8 (they ship plenty of them). From this information, we can “play” the match to see what the scoreline is.
The playing of the game isn’t quite as fun as a virtual football match might sound and is not too dissimilar to drawing coloured balls out of a bag, like you might have learned in an introductory stats lesson at school. The colour of the balls in the bag relate to the number of goals scored, and the proportion of each ball is dictated by the attacking coefficient combined with the opponent’s defensive coefficient.
In our example, the home team have a good attack, and the visitors an awful defence, so the balls in the bag for the home team are likely to be of higher value, meaning they’ll likely pick a ball that represents a high number of goals. When the away team draws from the bag, they’ll likely pick a ball worth zero goals on average.
Anyway, let’s say that from this bag we draw two goals for the home team, and one for the away team. Our result, therefore, is 2-1 — and the home team has won. Do that plenty of times, over lots of games, and you can simulate full seasons and be able to calculate the number of times a given team finishes a league in each position.
The matrix below shows exactly the results of the process above, but Opta didn’t have to draw any balls out of a bag for it.
Taking into account the strengths of the teams from historical data and simulating the season 10,000 times, it shows the proportion of times each team finishes in various places in the table. Norwich, Villa and Bournemouth all get relegated more often than not in the 10,000 “seasons”, Liverpool win the title and Spurs leap up to sixth. Note that these predictions are from early April — taking into account the table positions of all teams before the COVID-19 shutdown.
Compared to points per game, this model takes into account schedule strength, and uses attacking and defensive coefficients that are far more accurate means of measuring sporting merit (and team strength, which is ultimately what we should care about here).
Statistical smarts and a sound model are one thing, but the process of getting results and deciding the final table isn’t really grounded in reality: we don’t live in a world where things happen “on average”.
Take Leicester’s 5,000-1 title winning season four years ago. That 5,000-1 figure was calculated by the bookies, whose approach isn’t far off what was done above, but will be informed with more information to tweak these attacking and defensive coefficients based on players available, form, suspensions and the like. In probability terms, they gave Leicester a 0.02 per cent chance of winning the 2015-16 title. So if this was a roulette wheel with 10,000 pockets, Leicester occupied just two of them.
If we simulated that season instead of playing it, we would have seen Leicester, over 10,000 roulette wheel spins, win the league twice. On average, though, the title would likely have gone to Manchester City (who were the favourites at the time) and who would have had the most pockets on the wheel. The more you spin the wheel, the more you tend to arrive at the more “likely” outcome.
Life, however, is a game of chance, and at times, so is football.
For that reason alone, the best way to resolve a stagnant league is by spinning the roulette wheel, not counting up how many pockets a given team has.
The better way to resolve this situation would be to find a model that is agreed to be the best and pick one of the 10,000 simulations as the result.
Whether this is done or not is unlikely, but presented above are plenty of reasons why points per game is a poor measure of resolving a league season, and hopefully UEFA and others soon realise the same.
The mechanics of determining 'sporting merit" are alot more complex but a lot less devious then the 1997 version.
Apologies.
Derbyshire is Derbyshire
Yorkshire is Yorkshire
Never the twain shall meet.
Again
Yorkshire is Yorkshire
Never the twain shall meet.
Again
#411
jack bauer
- Key Player
-
- Group: Members
- Posts: 8,913
- Joined: 31-March 10
Posted 25 April 2020 - 12:06 PM
metallilad, on 25 April 2020 - 11:07 AM, said:
TBF. Barrow weren't running away with it either but having said that. Your idea seems to be a "make sense" option.
that's kind of where i was going really, read the bbc article and whilst i think everything has to be done to be fair and thorough, every scenario they look at for our league still puts barrow at the top.
#412
The Earl of Chesterfield
- Legend
-
- Group: Members
- Posts: 27,156
- Joined: 24-February 08
- Gender:Male
- Location:With the Rainbow People
Posted 25 April 2020 - 02:02 PM
If there's gotta be a promotion, it's gotta be Barrow...
Never underestimate the stupidity of people
#413
DerbySpireite59
- DerbySpireite57
-
- Group: Members
- Posts: 2,251
- Joined: 22-April 15
- Gender:Male
- Location:Derby
Posted 25 April 2020 - 08:19 PM
The Earl of Chesterfield, on 25 April 2020 - 02:02 PM, said:
If there's gotta be a promotion, it's gotta be Barrow...
Whilst I'd like evo and hird to be promoted, my money would have been on harrogate
"No one is 100% good, nor 100% evil"
#414
mr. smith
- Legend
-
- Group: Members
- Posts: 11,844
- Joined: 06-June 05
Posted 25 April 2020 - 08:47 PM
DerbySpireite57, on 25 April 2020 - 08:19 PM, said:
Whilst I'd like evo and hird to be promoted, my money would have been on harrogate
if theres 1 going up id say barrow just cos Harrogate is the easier trip next season.
theres a quarter of season left & to me you cant just say team x will go up or be relegated or end in play offs etc.
if you cant finish season its void.
#415
tomjoad
- First Team Player
-
- Group: Members
- Posts: 3,046
- Joined: 07-June 05
- Gender:Male
- Location:Chesterfield
- Interests:Bruce Springsteen Southside Counting Crows Abbot ale and LMS no 8s!
Posted 25 April 2020 - 10:40 PM
mr. smith, on 25 April 2020 - 08:47 PM, said:
if there's 1 going up id say barrow just cos Harrogate is the easier trip next season.
theres a quarter of season left & to me you cant just say team x will go up or be relegated or end in play offs etc.
if you cant finish season its void.
theres a quarter of season left & to me you cant just say team x will go up or be relegated or end in play offs etc.
if you cant finish season its void.
The problem for me is how can you promote anyone after 37 games? Nothing is guaranteed, injuries, questionable refereeing decisions teams bottling it, the list is endless.
"we were top all season" counts for nothing. Its not about Barrow or Evatt or individuals but again how can you reward a team for something that MAY have happened?
Carlisle winning in he last minute of the last game to stay up springs to mind, Newcastle among others squandering huge leads. Also Barrow in pole position were only 4 points in front of Harrogate in second, and had them to play.
Void it and be done.
#416
dim view
- Legend
-
- Group: Members
- Posts: 23,015
- Joined: 09-June 05
- Gender:Male
Posted 26 April 2020 - 08:00 AM
tomjoad, on 25 April 2020 - 10:40 PM, said:
The problem for me is how can you promote anyone after 37 games? Nothing is guaranteed, injuries, questionable refereeing decisions teams bottling it, the list is endless.
"we were top all season" counts for nothing. Its not about Barrow or Evatt or individuals but again how can you reward a team for something that MAY have happened?
Carlisle winning in he last minute of the last game to stay up springs to mind, Newcastle among others squandering huge leads. Also Barrow in pole position were only 4 points in front of Harrogate in second, and had them to play.
Void it and be done.
"we were top all season" counts for nothing. Its not about Barrow or Evatt or individuals but again how can you reward a team for something that MAY have happened?
Carlisle winning in he last minute of the last game to stay up springs to mind, Newcastle among others squandering huge leads. Also Barrow in pole position were only 4 points in front of Harrogate in second, and had them to play.
Void it and be done.
I'm struggling to understand why the simple 'convert the current competition into a 2 season, play each other 4 times' idea hasn't at least been suggested. The only complication would be sorting out the qualifiers for European comps, which is a Premier League problem and which might be cancelled anyway.
Get it on, bang the gong , get it on
#417
Tha Knows...
- Legend
-
- Group: Members
- Posts: 21,184
- Joined: 29-June 13
- Gender:Male
Posted 26 April 2020 - 10:11 AM
dim view, on 26 April 2020 - 08:00 AM, said:
I'm struggling to understand why the simple 'convert the current competition into a 2 season, play each other 4 times' idea hasn't at least been suggested. The only complication would be sorting out the qualifiers for European comps, which is a Premier League problem and which might be cancelled anyway.
You are way too visionary for the authorities.
#418
dim view
- Legend
-
- Group: Members
- Posts: 23,015
- Joined: 09-June 05
- Gender:Male
Posted 26 April 2020 - 10:18 AM
Spire-Power, on 26 April 2020 - 10:11 AM, said:
You are way too visionary for the authorities.
A point I often make to the missus, her being the authority.
Get it on, bang the gong , get it on
#419
Westbars Spireite
- Legend
-
- Group: Members
- Posts: 66,590
- Joined: 18-September 06
- Gender:Male
- Location:Chesterfield, Derbyshire
- Interests:Chesterfield FC, cricket, beer
Posted 26 April 2020 - 11:30 AM
dim view, on 26 April 2020 - 08:00 AM, said:
I'm struggling to understand why the simple 'convert the current competition into a 2 season, play each other 4 times' idea hasn't at least been suggested. The only complication would be sorting out the qualifiers for European comps, which is a Premier League problem and which might be cancelled anyway.
It had crossed my mind. Seems a relatively simple solution to me.
#420
Ernie Ernie Ernie
- Legend
-
- Group: Members
- Posts: 30,561
- Joined: 06-June 05
- Gender:Male
Posted 26 April 2020 - 08:15 PM
dim view, on 26 April 2020 - 08:00 AM, said:
I'm struggling to understand why the simple 'convert the current competition into a 2 season, play each other 4 times' idea hasn't at least been suggested. The only complication would be sorting out the qualifiers for European comps, which is a Premier League problem and which might be cancelled anyway.
Be crap for us
