Why aren't losses factored into sorting teams in the standings?

[morehockeystats]

Good. So we have 17-x-7 vs. 17-y-6 where x & y don't need to be known.

It would therefore follow that we don't even need to have x & y in the equation, as your formula indicates that regardless of the value of x and y, the point total ends up being the same. Therefore, x & y are completely disregarded when calculating the points total.

However, out of 4920 results for the 30 teams over an 82 game schedule (82 x 30 x 2 = 4920), some 40% of those results are an L or fit right into your equation where the x and y are. So I'd like to ask
1) how do you feel about ignoring or not using some 40% of NHL game results when compiling the standings?
2) would a system for sorting teams which factored in 100% of the results tell us more than a system which factors in only about 60% of the results?

Zeno of Elea, please leave your aporias to yourself.

After a moment’s silence, he added, ‘I suppose the hardest thing is to convince anybody that 0+0+0=0...'

G.K. Chesterton, "The Blast Of The Book", 1933.
​

Although it's highly entertaining to see how you got 4920 results in 1230 games.

 

[Canadiens1958]

Remediation

Good. So we have 17-x-7 vs. 17-y-6 where x & y don't need to be known.

It would therefore follow that we don't even need to have x & y in the equation, as your formula indicates that regardless of the value of x and y, the point total ends up being the same. Therefore, x & y are completely disregarded when calculating the points total.

However, out of 4920 results for the 30 teams over an 82 game schedule (82 x 30 x 2 = 4920), some 40% of those results are an L or fit right into your equation where the x and y are. So I'd like to ask
1) how do you feel about ignoring or not using some 40% of NHL game results when compiling the standings?
2) would a system for sorting teams which factored in 100% of the results tell us more than a system which factors in only about 60% of the results?

You have an 82 game schedule for 30 teams or 82 x 15 = 1230 NHL games in a regular season or 1230 total results. There may be 4920 possible outcomes but they are not to be confused with results.

http://www.hockey-reference.com/leagues/NHL_2017.html

2016-17 NHL standings linked. Team summary section shows an average team with 41W 31L 10OL. So the results balance. 41W = 31 + 10 (two types of loses RS and OT/SO). Each result produces a win and a loss - they go hand in hand cannot have one without the other.

 

[Nenikoj]

You are just trying to find the golden BB that will give your idea legs. NOT interested.

Going back to my question.

If your task is to sort the teams (Detroit and Toronto) in the standings by the methods used in the NHL standings and all you had was Detroit 17-x-7 and Toronto 17-y-6 - where x and y are unknown, would you be able to sort the teams? If so, how would you do it? If not, what missing information would you need?

Can you identify any situations in which you would need to know the values of x and y to determine how to sort the teams using points?

 

[Nenikoj]

Good. So we have 17-x-7 vs. 17-y-6 where x & y don't need to be known.

It would therefore follow that we don't even need to have x & y in the equation, as your formula indicates that regardless of the value of x and y, the point total ends up being the same. Therefore, x & y are completely disregarded when calculating the points total.

However, out of 4920 results for the 30 teams over an 82 game schedule (82 x 30 x 2 = 4920), some 40% of those results are an L or fit right into your equation where the x and y are. So I'd like to ask
1) how do you feel about ignoring or not using some 40% of NHL game results when compiling the standings?
2) would a system for sorting teams which factored in 100% of the results tell us more than a system which factors in only about 60% of the results?

Zeno of Elea, please leave your aporias to yourself.

After a moment’s silence, he added, ‘I suppose the hardest thing is to convince anybody that 0+0+0=0...'

G.K. Chesterton, "The Blast Of The Book", 1933.
​

Although it's highly entertaining to see how you got 4920 results in 1230 games.

My bad. I multiplied it by 2 twice!
82 teams
30 games each team
82 x 30 = 2460

Approximately 40% of the games result in one of the two teams getting zero points. So approximately 1000 out of the 2460 results are not factored into the standings. So it leads me back to my questions:

1) how do you feel about ignoring or not using some 40% of NHL game results when compiling the standings?
2) would a system for sorting teams which factored in 100% of the results tell us more than a system which factors in only about 60% of the results?

 

[Nenikoj]

You have an 82 game schedule for 30 teams or 82 x 15 = 1230 NHL games in a regular season or 1230 total results. There may be 4920 possible outcomes but they are not to be confused with results.

http://www.hockey-reference.com/leagues/NHL_2017.html

2016-17 NHL standings linked. Team summary section shows an average team with 41W 31L 10OL. So the results balance. 41W = 31 + 10 (two types of loses RS and OT/SO). Each result produces a win and a loss - they go hand in hand cannot have one without the other.

As I explained to morehockeystats:

Approximately 40% of the games result in one of the two teams getting zero points. So approximately 1000 out of the 2460 results are not factored into the standings. So it leads me back to my questions:

1) how do you feel about ignoring or not using some 40% of NHL game results when compiling the standings?
2) would a system for sorting teams which factored in 100% of the results tell us more than a system which factors in only about 60% of the results?

 

[morehockeystats]

As I explained to morehockeystats:

Approximately 40% of the games result in one of the two teams getting zero points. So approximately 1000 out of the 2460 results are not factored into the standings. So it leads me back to my questions:

1) how do you feel about ignoring or not using some 40% of NHL game results when compiling the standings?
2) would a system for sorting teams which factored in 100% of the results tell us more than a system which factors in only about 60% of the results?

As I already stated, you're engaging in aporias. The "result" is not for each team. The result is for both teams, and it is reflected in the standings.

 

[Canadiens1958]

Aporias

As I explained to morehockeystats:

Approximately 40% of the games result in one of the two teams getting zero points. So approximately 1000 out of the 2460 results are not factored into the standings. So it leads me back to my questions:

1) how do you feel about ignoring or not using some 40% of NHL game results when compiling the standings?
2) would a system for sorting teams which factored in 100% of the results tell us more than a system which factors in only about 60% of the results?

First the definition of aporias "an irresolvable internal contradiction or logical disjunction in a text, argument, or theory."

To your "numbers" yes but that leaves at least 1460. Given that an NHL regular season generates only 1230 results, I have no problem with your perceived overage of results. As long as the important 1230 results are included, the NHL will survive.

 

[Nenikoj]

As I already stated, you're engaging in aporias. The "result" is not for each team. The result is for both teams, and it is reflected in the standings.

If the results for both teams for all games are reflected in the standings, then you would need to know the values for x and y in the Detroit 17-x-7 vs. Toronto 17-y-6 example I cited earlier. But you said you don't need to know those values. So which is it - do you need to know the values of x & y or do you not need to know the values of x and y to do a standings sort?

 

[Nenikoj]

First the definition of aporias "an irresolvable internal contradiction or logical disjunction in a text, argument, or theory."

To your "numbers" yes but that leaves at least 1460. Given that an NHL regular season generates only 1230 results, I have no problem with your perceived overage of results. As long as the important 1230 results are included, the NHL will survive.

Whether it is aporias or not is irrelevant. I asked a couple of very fair questions which you haven't answered. I am very curious what your feeling about these two points is.

1) how do you feel about ignoring or not using some 40% of NHL game results when compiling the standings?
2) would a system for sorting teams which factored in 100% of the results tell us more than a system which factors in only about 60% of the results?

 

[Canadiens1958]

Reality

Whether it is aporias or not is irrelevant. I asked a couple of very fair questions which you haven't answered. I am very curious what your feeling about these two points is.

1) how do you feel about ignoring or not using some 40% of NHL game results when compiling the standings?
2) would a system for sorting teams which factored in 100% of the results tell us more than a system which factors in only about 60% of the results?

I answer reality based questions only. The scenario that you describe and ask for an answer does not exist and cannot exist. This has been explained to you by posters a few times.

 

[morehockeystats]

If the results for both teams for all games are reflected in the standings, then you would need to know the values for x and y in the Detroit 17-x-7 vs. Toronto 17-y-6 example I cited earlier. But you said you don't need to know those values. So which is it - do you need to know the values of x & y or do you not need to know the values of x and y to do a standings sort?

Have you stopped drinking brandy in the mornings or have you not stopped drinking brandy in the mornings? (C)

 

[Nenikoj]

I answer reality based questions only. The scenario that you describe and ask for an answer does not exist and cannot exist. This has been explained to you by posters a few times.

The reality is some 40% of team's results in NHL games are - with respect to how teams are sorted in the standings - completely ignored. Thus my questions:

1) how do you feel about ignoring or not using some 40% of NHL game results when compiling the standings?
2) would a system for sorting teams which factored in 100% of the results tell us more than a system which factors in only about 60% of the results?

 

[Nenikoj]

Have you stopped drinking brandy in the mornings or have you not stopped drinking brandy in the mornings? (C)

I will rephrase:

1) In the 17-x-7 vs. 17-y-6 example, how do you feel about ignoring or not using the x and y values results when compiling the standings?
2) would a system for sorting teams which factored in the values of x & y in the aforementioned example tell us more than a system which doesn't factor in the values of x & y?

 

[morehockeystats]

I will rephrase:

1) In the 17-x-7 vs. 17-y-6 example, how do you feel about ignoring or not using the x and y values results when compiling the standings?
2) would a system for sorting teams which factored in the values of x & y in the aforementioned example tell us more than a system which doesn't factor in the values of x & y?

1) good
2) tell us more about what? About brandy in the mornings?

 

[Doctor No]

I will rephrase:

1) In the 17-x-7 vs. 17-y-6 example, how do you feel about ignoring or not using the x and y values results when compiling the standings?
2) would a system for sorting teams which factored in the values of x & y in the aforementioned example tell us more than a system which doesn't factor in the values of x & y?

Let me try this again. What is your goal for sorting the standings prior to season's end?

 

[Nenikoj]

Let me try this again. What is your goal for sorting the standings prior to season's end?

My apologies if I overlooked your question earlier.

The primary goal of sorting the teams in the standings is to give the readers a sense of where each team stands relative to other teams based on their performance in all the games they've played to date. So a 23-24-1 team should be listed behind a 22-21-1 team, even though the 23-24-1 team has more points, as the 22-21-1 team has a higher PPCT and a higher GA500. However, the NHL lists the 23-24-1 team ahead of the 22-21-1 team, as they are really looking at 23-x-1 vs. 22-y-1 where x and y are immaterial or don't need to be known.

What I'm asking is what is the advantage of ignoring the L column when sorting teams?

 

[Nenikoj]

1) good
2) tell us more about what? About brandy in the mornings?

I think you're in a serious case of cognitive dissonance on this matter. I'll move on to discuss it with those who wish to apply some intellectual honesty on the matter.

 

[morehockeystats]

I think you're in a serious case of cognitive dissonance on this matter. I'll move on to discuss it with those who wish to apply some intellectual honesty on the matter.

If anyone shows an intellectual dishonesty here, it's you.

There are two possible purposes of the standings:
1) To show who is currently qualified for playoffs or medals
2) To show who's been more effective so far

MLB and NBA use 2). NHL, MLS, Euro Soccer, Euro Basket, chess, IOC, Nordic skiing, biathlon (with staggered start) and others use 1).

If you want to use 2) you should also provide:
* The Buchholz coefficient for the teams
* The remaining strength of schedule for the teams.
E.g. the "effectiveness" standings of the Blues this season after the All-Star game was very deceptive because they still had five games against the Avalanche and three or four against the Coyotes, far more than their main competitors.

I hope this creates some order in your head.

 

[Nenikoj]

If anyone shows an intellectual dishonesty here, it's you.

There are two possible purposes of the standings:
1) To show who is currently qualified for playoffs or medals
2) To show who's been more effective so far

MLB and NBA use 2). NHL, MLS, Euro Soccer, Euro Basket, chess, IOC, Nordic skiing, biathlon (with staggered start) and others use 1).

If you want to use 2) you should also provide:
* The Buchholz coefficient for the teams
* The remaining strength of schedule for the teams.
E.g. the "effectiveness" standings of the Blues this season after the All-Star game was very deceptive because they still had five games against the Avalanche and three or four against the Coyotes, far more than their main competitors.

I hope this creates some order in your head.

To help me better understand you, can you please answer the following:

1) Let's say team A is 23-24-1 and team B is 22-21-1. Would you agree that team B has the better W-L-OTL record of the two? If so, why? If not, why?
2) would a system for sorting teams which factored in the values in the L column tell us more about the strength of a team's record than a system in which the values in the L column were not factored in? If yes, why? If no, why?

 

[morehockeystats]

To help me better understand you, can you please answer the following:

1) Let's say team A is 23-24-1 and team B is 22-21-1. Would you agree that team B has the better W-L-OTL record of the two? If so, why? If not, why?

What is the meaning of a "W-L-OTL" record? What would it indicate?

2) would a system for sorting teams which factored in the values in the L column tell us more about the strength of a team's record than a system in which the values in the L column were not factored in? If yes, why? If no, why?

Yes, only if accompanied by the Buchholz coefficient and the remaining schedule strength.

You're doing exactly the same mistake as with the GWG thread, you are humping the wrong mule. NHL chose to produce the standings that show other things that you want them to show. These aren't worse or better standings, they serve a purpose different from what you want them to. So either produce your own (and define your own GWG) standings, or petition the NHL board of governors to change what they do now.

 

[Nenikoj]

What is the meaning of a "W-L-OTL" record? What would it indicate?

Yes, only if accompanied by the Buchholz coefficient and the remaining schedule strength.

You're doing exactly the same mistake as with the GWG thread, you are humping the wrong mule. NHL chose to produce the standings that show other things that you want them to show. These aren't worse or better standings, they serve a purpose different from what you want them to. So either produce your own (and define your own GWG) standings, or petition the NHL board of governors to change what they do now.

Since you keep wanting to avoid the issue, I'll reword again:

Let's say your job is to determine under which coach your team performed better. All you have is W-L-OTL numbers, but it can be assumed that all games were regular season NHL games. Under coach A, your team went 20-25-5 and under coach B, your team went 22-0-0. And scenario #2 with a larger sample size - under coach C, your team went 30-60-1 and under coach D, your team went 30-0-0. How would you go about determining under which coach your team did better? And in each scenario, under which coach would you say your team did better?

 

[morehockeystats]

Since you keep wanting to avoid the issue, I'll reword again:

Let's say your job is to determine under which coach your team performed better. All you have is W-L-OTL numbers, but it can be assumed that all games were regular season NHL games. Under coach A, your team went 20-25-5 and under coach B, your team went 22-0-0. And scenario #2 with a larger sample size - under coach C, your team went 30-60-1 and under coach D, your team went 30-0-0. How would you go about determining under which coach your team did better? And in each scenario, under which coach would you say your team did better?

For the task of determining which coach did a better job, I'd rather use a normalized Sonneborne-Berger coefficient, like I do on my website.

But once again, if you decide, for yourself, that coach performance should be evaluated by points percentage, 30-60-1 is worse than 30-0-0.

And if someone else decides that coach performance should be evaluated by placement of a team, then it doesn't matter what W-L-T record is, what matters is whether the 30-60-1 yielded a better placement than 30-0-0.

Moreover, if you ever looked at the AHL standings, you can see an outstanding application of the points percentage - because the standings are determined by it, rather than by 2*W + OTL, since some of the Pacific division plays less games than the rest of the league.

Still, the NHL, like most of the competitions around the world, considers that the primary purpose of standings is to show who is currently qualified for the playoffs/medals, and for that purpose the number of losses DOES NOT MATTER.

 

[Nenikoj]

For the task of determining which coach did a better job, I'd rather use a normalized Sonneborne-Berger coefficient, like I do on my website.

But once again, if you decide, for yourself, that coach performance should be evaluated by points percentage, 30-60-1 is worse than 30-0-0.

And if someone else decides that coach performance should be evaluated by placement of a team, then it doesn't matter what W-L-T record is, what matters is whether the 30-60-1 yielded a better placement than 30-0-0.

Please explain why 30-60-1 should yield a better placement than 30-0-0.

If 30-60-1 does yield a better placement than 30-0-0, please explain why one should use the "better placement" method to determine which coach did better.

Moreover, if you ever looked at the AHL standings, you can see an outstanding application of the points percentage - because the standings are determined by it, rather than by 2*W + OTL, since some of the Pacific division plays less games than the rest of the league.

How do you think it should be determined if points percentage or points should be used in situations where there are discrepancies in games played?

Still, the NHL, like most of the competitions around the world, considers that the primary purpose of standings is to show who is currently qualified for the playoffs/medals, and for that purpose the number of losses DOES NOT MATTER.

How would you go about determining which team is more "qualified" for the playoffs between 23-24-1 and 22-21-1?

Do you have the URL for your website. I'd like to plug in 30-60-1 & 30-0-0 or 23-24-1 & 22-21-1 and see which it says is the better performance.

 

[Canadiens1958]

Logical Coherence

Standings require logical coherence at all levels and across all levels within a specific league. Likewise statistical definitions.

Necessary for comparables that arise as the league evolves.

Fixating on one feature gets away from this requirement.

 

[morehockeystats]

Please explain why 30-60-1 should yield a better placement than 30-0-0.

If 30-60-1 does yield a better placement than 30-0-0, please explain why one should use the "better placement" method to determine which coach did better.

How do you think it should be determined if points percentage or points should be used in situations where there are discrepancies in games played?

Because sometimes a qualification is what matters.
Imagine a tournament: 2 ECHL teams, 10 amateur teams, 10 NHL teams. The ECHL teams do not play against each other, but play 61 games each against the NHL and 30 games against the amateurs. One team completed the tournament already and managed to squeeze 1 OTL against the NHL. The other only played the amateurs and beat them all, but still has 61 games against the NHLers where it's expected to go 0-61-0. The 30-60-1 team has an advantage over the 30-0-0 one.

The points percentage is only really insightful when combined with the Buchholz and remaining strength coefficients.

How would you go about determining which team is more "qualified" for the playoffs between 23-24-1 and 22-21-1?

The one that is higher placed is usually closer to qualification.

Do you have the URL for your website. I'd like to plug in 30-60-1 & 30-0-0 or 23-24-1 & 22-21-1 and see which it says is the better performance.

I'd think you'd figure out by now. You can't "plug in" numbers in it. But you can understand what the Buchholz and the Sonneborn-Berger coefficients are:
http://morehockeystats.com/teams/buchberg
http://morehockeystats.blogspot.com/2017/03/on-buchholz-and-sonneborn-berger.html
http://morehockeystats.blogspot.com/2017/03/on-buchholz-and-sonneborn-berger_13.html