Vs#2 Thread

[Czech Your Math]

There's a at least a couple of factors which complicate this type of metric. First, the changes in the talent pool over time. Second, the expanding size of the league, which gives more opportunity to more players.

I don't think it will ever be easy to fairly compare seasons from many decades ago to more recent decades. However, I still believe adjusted points is one of the fairest metrics for more modern seasons. Seasons before the 50s and during the turbulent post-O6 expansion period are particularly difficult to compare.

Along those lines, I wanted to direct anyone interested to a "work in progress" study of the top 1N (N = number of teams) players and how their average adjusted points compared to a predictive model using regression:

Using Regression to Study Factors Which Influence Scoring for Top Tier Players 1968-Present

 

[Dreakmur]

If you're talking about the adjusted points from Hockey-Reference, then those are garbage.

 

[Sturminator]

If you're talking about the adjusted points from Hockey-Reference, then those are garbage.

Adjusted points work ok if you don't go back any further than the mid-70's. They're useless before that point, though.

 

[BM67]

Going back to what I posted here: http://hfboards.mandatory.com/showpost.php?p=46173207&postcount=33

Some examples to compare:

#1-#T

Season|Stewart|Iginla|Alfredsson|Peter Stastny|Bobby Hull|Lach
1|134.87|56.77|58.37|102.78|63.09|45.37
2|70.71|40.95|80.61|122.34| 60.48|1.98
3|103.98|61.77|57.58|117.73|106.82|87.88
4|113.73|80.99|39.97|109.12|68.86|96
5|97.17|82.15|75.85|89.78|109.09|122.14
6|93.3|128.17|80.99|108.43|81.58|91.26
7|70.39|81.14|94.79|78.99|109.89|47
8|88.04|97.76|94.47|103.51|95.09|107.02
9|94.61|73.17|107.14|77.37|122.01|51.17
10|86.67|103.3|112.48|68.99|109.1|72.92
11|73.64|116.35|95.6|58.58|100.33|68.01
12|86.16|105.7|105.66|62.03|117.8|99.49
13|84.21|82.73|87.89|34.15|83.83|61.65
14|83.06|111.01|85.13|16.26|103.86|41.1
15|30.85|87.21|40.02|3.76|99.39|-
16|-|-|76.79|-|16.98|-
Career|1311.39|1309.16|1293.24|1153.81|1448.2|992.98

Bobby Hull has the best career value, but his best season is beat by everyone except Alfredsson, although basically tied with Stastny and Lach. This fits with Hull as an incredible goal scorer, but just very good with regards to points.

Stewart comes in 2nd in career, but 1st in best season. This fits well with his one for the ages rookie season, and his solid and steady but not incredible production afterwards.

Iginla should pass Stewart into 2nd career wise this year, and while his Art Ross season truly stands out, he does have more very good seasons than Stewart.

Alfredsson doesn't have the peak of the others, nor is he as steadily very good as Stewart or Hull, and that seems about right.

Lach for a guy that was the career record holder for points for a while, doesn't look so great.

#2-#T

Season|Stewart|Iginla|Alfredsson|Peter Stastny|Bobby Hull|Lach
1| 143.2 | 57.66 | 59.66 | 105.67 | 64.74 | 48.09
2| 72.69 | 41.45 | 81.87 | 127.87 | 62.5 | 2.01
3| 110.86 | 63.08 | 58.29 | 123.02 | 108.29 | 89.78
4| 117.03 | 81.7 | 40.82 | 114.15 | 71.25 | 97.83
5| 100.4 | 83.29 | 76.51 | 93.85 | 111.11 | 127.8
6| 95.64 | 129.43 | 82.12 | 113.59 | 83.78 | 94.76
7| 71.74 | 81.95 | 95.73 | 82.62 | 112.69 | 48.23
8| 90.57 | 98.65 | 95.4 | 106.53 | 98.34 | 108.54
9| 97.79 | 74.11 | 108.11 | 80.65 | 127.63 | 53.31
10| 89.66 | 104.44 | 113.93 | 70.19 | 116.62 | 75.71
11| 75.19 | 117.68 | 96.67 | 60.36 | 101.85 | 72.35
12| 87.81 | 106.96 | 106.87 | 62.96 | 122.09 | 106.21
13| 86.89 | 83.73 | 88.93 | 34.7 | 87.85 | 67.43
14| 84.68 | 112.34 | 86.15 | 16.47 | 109.28 | 44.01
15| 32.1 | 88.48 | 40.5 | 3.81 | 102.72 | -
16| - | - | 77.92 | - | 17.29 | -
Career| 1356.24 | 1324.95 | 1309.48 | 1196.44 | 1498.02 | 1036.06

 

[Czech Your Math]

Going back to what I posted here: http://hfboards.mandatory.com/showpost.php?p=46173207&postcount=33

I like that system, because:

A) it's a large enough sample to not be influenced by a couple of players having unusually great years

B) it's still a small enough sample to contain only top players

C) it's proportional to league size, and therefore to opportunity

D) it doesn't discard any player-seasons as outliers

It's sort of an "adjusted points for superstars". The toughest issue to assess is change in talent pool and thus league quality. One could effectively double the talent pool overnight (let's say perfectly identical clones are suddenly introduced), yet there's really no way to capture that effect. That's why I sometimes compare players to "Canadian players only", just to get an idea of the magnitude of that effect.

 

[seventieslord]

I like that system too, as it has a built in accounting for the generally accepted fact that "5 th in 1937 is not the same as 5th in 1987".

As far as not removing outliers goes, as long as the results in the end "work" (I.e. a distribution of players across history who hit certain benchmark scores every year that makes sense and rises slowly over time) then it's gone. What BM67 proposed hours far down enough that it's rare one would need to care that a lemieux Or gretzky at the top is an outlier because the difference between using them and not, would be a benchmark score difference of maybe three points.

 

[TheDevilMadeMe]

I think seventies is making a very basic rational mistake here. The point of the Vs2 system is to remove outliers and then compare performance against the highest non-outlier. In this case, that is very obviously Joe Sakic. Completely removing Mario and Jagr and then treating Sakic as the #1 and removing him as well is just goofy. The whole point is to compare to the first non-outlier. Sakic is that guy.

Agreed. Though personally, I prefer using Vs5 for post-expansion players only. No arbitrary determination of outliers is ever needed if you use Vs5.

 

[seventieslord]

Agreed. Though personally, I prefer using Vs5 for post-expansion players only. No arbitrary determination of outliers is ever needed if you use Vs5.

This is a cherry picked example, but isn't 1989 still right out if whack then? It is littered with outliers and their linemates.

 

[TheDevilMadeMe]

This is a cherry picked example, but isn't 1989 still right out if whack then? It is littered with outliers and their linemates.

In 1988, 5th place (3 way tie Peter Stastny/Luc Robitaille/Mark Messier) had 111 points
In 1989, 5th place Rob Brown had 115 points
In 1990, 5th place Brett Hull had 113 points.

I'm fine with using Rob Brown for 1989, even though he played with 199 point Lemieux.

You would user Joe Mullen (7th place with 110 points), the top scorer who did not play with Gretzky or Lemieux, other than Yzerman's outlier 155?

 

[seventieslord]

Yes.

 

[Rob Scuderi]

Let's talk Defensemen scores in the early 70s for no other reason than I'm doing them for Stapleton.

Bobby Orr in 1970 is my first outlier (surprise) and that carries through 1975. I'm using the first non-outlier as my #2, but I don't know what to do with Brad Park in 1972 and 1974 (doesn't affect Stapleton, but should be discussed). It looks like a clear outlier to me, but I'd rather bring it to the group.

Here's the placements:

1972
Orr - 117
Park - 73
Tremblay - 57
Guevremont - 53
Lapointe - 49

1974
Orr - 122
Park - 82
Redmond - 59
Vadnais - 59
Potvin - 54

 

[BraveCanadian]

In 1988, 5th place (3 way tie Peter Stastny/Luc Robitaille/Mark Messier) had 111 points
In 1989, 5th place Rob Brown had 115 points
In 1990, 5th place Brett Hull had 113 points.

I'm fine with using Rob Brown for 1989, even though he played with 199 point Lemieux.

You would user Joe Mullen (7th place with 110 points), the top scorer who did not play with Gretzky or Lemieux, other than Yzerman's outlier 155?

Yes.

Yikes.

 

[seventieslord]

Let's talk Defensemen scores in the early 70s for no other reason than I'm doing them for Stapleton.

Bobby Orr in 1970 is my first outlier (surprise) and that carries through 1975. I'm using the first non-outlier as my #2, but I don't know what to do with Brad Park in 1972 and 1974 (doesn't affect Stapleton, but should be discussed). It looks like a clear outlier to me, but I'd rather bring it to the group.

Here's the placements:

1972
Orr - 117
Park - 73
Tremblay - 57
Guevremont - 53
Lapointe - 49

1974
Orr - 122
Park - 82
Redmond - 59
Vadnais - 59
Potvin - 54

Smaller sample size, so much more room for multiple outliers. I like to look at where the "pack" begins.

Yikes.

You wouldn't be saying yikes at using Mullen (a non outlier who came across his points honestly) if Brown hadn't conveniently missed 12 games and 20 points, getting him to 135. Then it would be obvious that using him as the standard for comparison in a percentage system is a horrible idea. The fact that he "only" got 115 points makes it easier to dismiss his outlier status, regardless of the fact that he sat in the back of the bus.

The point of removing gretzky and lemieux is because no other era had players who dominated this much offensively season after season. If you keep them in, you hold the savards and hawerchuks to an impossible standard. So obviously you must remove them. The 1980s without those two, are fairly comparable to most other decades so it levels the playing field. But if we do that then their residual effects must also be accounted for.

That said, from my end, all this discussion is geared towards having results that "work" and can be used as a shorthand. I think what most of us would like to see is a system where the number of players reaching certain benchmarks gradually increases over time. I did some quick formulas using the results in my sheet and how each season looks compared to how it "should" look on the time.curve, and it appears that from 1980-1999 the most unforgiving seasons, in order are:

1999 - this was a crap year and results should lag a bit, but half as many 60 scores as 1984 makes little sense
1995 - might make sense if the tie for 1st is treated as one player making the standard 65 instead
1993 - uses oates as the standard (#2 non lemieux) but that is a bit harsh especially since this was the season of offense and if anything results should look extreme this year, not subdued.
1982, 1981 - Nilsson (12 points ahead of 4th) and Stastny ( in 3rd with 31% more than 9th) should not be the standard. Weird year for top end scorers. Savages should be the standard if we're being logical here

The easiest are:

1987: not much can be done here, even if we used 2nd place Kurri that's only 3 points more than my current standard
1989: there do appear to be "too many" hitting 60, 70, 80 compared to surrounding years, so that is a good reason to consider using a higher standard. But at the same time it is worth noting that this is a unique year. Percentage wise, 30th is closer to 10th than any other season I could find.
1984: very 1987-like. Using coffey instead would only help marginally.
1992: I should use Hull and his 109, not Messier and his 107 here. This year would look perfect.
1994. Very 1989- like with the punching at the 2nd tier. Ran numbers with fedorov as the standard instead of oates, but then the pendulum swings farther the other way. This year should be a bit heavier, lots of great players in their primes.

 

[BraveCanadian]

You wouldn't be saying yikes at using Mullen (a non outlier who came across his points honestly) if Brown hadn't conveniently missed 12 games and 20 points, getting him to 135.

But he didn't!

Then it would be obvious that using him as the standard for comparison in a percentage system is a horrible idea. The fact that he "only" got 115 points makes it easier to dismiss his outlier status, regardless of the fact that he sat in the back of the bus.

He was NOT an outlier!

The point of removing gretzky and lemieux is because no other era had players who dominated this much offensively season after season. If you keep them in, you hold the savards and hawerchuks to an impossible standard. So obviously you must remove them. The 1980s without those two, are fairly comparable to most other decades so it levels the playing field. But if we do that then their residual effects must also be accounted for.

Do you take out all of Jari Kurri's seasons when he played with Gretzky?

All of Paul Coffey's?

Kevin Stevens?

Where does it end?

Who decides?

That said, from my end, all this discussion is geared towards having results that "work" and can be used as a shorthand. I think what most of us would like to see is a system where the number of players reaching certain benchmarks gradually increases over time. I did some quick formulas using the results in my sheet and how each season looks compared to how it "should" look on the time.curve, and it appears that from 1980-1999 the most unforgiving seasons, in order are:

I actually agree that as their are more teams, and outside other influences like war etc, their should be more players acknowledged.

However, my definition of a system is not one that you go picking through each year to make it look like you think it should.

Whatever is arrived upon should function well enough without any influence from the user and be as consistent as possible.

 

[seventieslord]

But he didn't!



He was NOT an outlier!



Do you take out all of Jari Kurri's seasons when he played with Gretzky?

All of Paul Coffey's?

Kevin Stevens?

Where does it end?

Who decides?



I actually agree that as their are more teams, and outside other influences like war etc, their should be more players acknowledged.

However, my definition of a system is not one that you go picking through each year to make it look like you think it should.

Whatever is arrived upon should function well enough without any influence from the user and be as consistent as possible.

1.*** Yes, Stevens, Kurri and Coffey are outliers when they finished top-3 in scoring as a result of being on the ice with a generational talent for the majority of the season. That should be obvious.
2.*** There is absolutely no way for there to not be “influence from the userâ€. Otherwise we would just blindly use #2 every year as the standard and not think of each season as an individual closed system. We all know that there are a lot of seasons where that would make absolutely no sense. The individual nuances and tweaks that I, and others, have added, get us closer to a useable shorthand system, not further away. It’s obvious that tweaks are required. The only question is which tweaks get us closer.

 

[BraveCanadian]

1.*** Yes, Stevens, Kurri and Coffey are outliers when they finished top-3 in scoring as a result of being on the ice with a generational talent for the majority of the season. That should be obvious.
2.*** There is absolutely no way for there to not be “influence from the userâ€. Otherwise we would just blindly use #2 every year as the standard and not think of each season as an individual closed system. We all know that there are a lot of seasons where that would make absolutely no sense. The individual nuances and tweaks that I, and others, have added, get us closer to a useable shorthand system, not further away. It’s obvious that tweaks are required. The only question is which tweaks get us closer.

1. *** No they aren't. Their totals aren't far enough away from the pack to make them outliers.

2. *** Yes there is... the whole point of making a "system" is to remove the influence of the user so that everyone knows what they are looking at and it is applied fairly across the board.

 

[Sturminator]

1. *** No they aren't. Their totals aren't far enough away from the pack to make them outliers.

He doesn't seem to want to understand this point.

2. *** Yes there is... the whole point of making a "system" is to remove the influence of the user so that everyone knows what they are looking at and it is applied fairly across the board.

1970-71 and 1988-89 will probably always need as asterix. There was one O6 season when the Production Line did stupid things that may need an asterix, as well. Otherwise, a simple 10% cutoff formula for determining if we should use the #2 or #3 score works in basically all cases. This is hardly an intractable problem.

 

[Dreakmur]

He doesn't seem to want to understand this point.

There's a big difference between not understanding and not agreeing.

 

[seventieslord]

He doesn't seem to want to understand this point.



1970-71 and 1988-89 will probably always need as asterix. There was one O6 season when the Production Line did stupid things that may need an asterix, as well. Otherwise, a simple 10% cutoff formula for determining if we should use the #2 or #3 score works in basically all cases. This is hardly an intractable problem.

Ok, prove it then.

And why use asterisks when we are smart enough to solve the problem?

 

[TheDevilMadeMe]

Ok, prove it then.

And why use asterisks when we are smart enough to solve the problem?

Why does he have to prove that when you are using a system called Vs2, you should be comparing players to the second best scorer whenever possible?

 

[seventieslord]

1. *** No they aren't. Their totals aren't far enough away from the pack to make them outliers.

*
Now that I have more time…
*
1.****** I’m not saying players like that are “outliers†in the sense that their point totals are far away from the pack, it’s that they’re in those high spots as a result of a generational talent.
2.****** I know you think you’re being clever “spoofing†me with the 3 stars - let me just add that I can’t control this. My phone is throwing it in somehow in the formatting from an outlook email to an android email to a web page. It's not there when I hit send...

He doesn't seem to want to understand this point.

Why be such a big bad meanie pants deliberately? It doesn’t help anything. You are obviously smart enough to know that I understand and just don’t agree.

Why does he have to prove that when you are using a system called Vs2, you should be comparing players to the second best scorer whenever possible?

*
1.****** I don’t call it “vs. 2â€, I tend to call it “the percentage systemâ€, as unaesthetic as that may be.
2.****** Good point. In a season when Kurri finishes #2 in scoring thanks to Gretzky, who is actually the “second best scorer� I’d almost guarantee you that it was not Kurri. And that’s all I’ve ever been trying to say.

 

[Bear of Bad News]

Bringing this one to the surface, now that it's been moved to "By The Numbers".

 

[BM67]

Updating this and adding a few players from the Top Centers Project.

Going back to what I posted here: http://hfboards.mandatory.com/showpost.php?p=46173207&postcount=33

Some examples to compare:

#1-#T

Season|Stewart|Iginla|Alfredsson|Peter Stastny|Bobby Hull|Lach|Delvecchio|Francis|Ullman
1|134.87|56.77|58.37|102.78|63.09|45.37|0|59.85|24.55
2|70.71|40.95|80.61|122.34| 60.48|1.98|56.64|85.45|67.68
3|103.98|61.77|57.58|117.73|106.82|87.88|88.72|76.11|68.46
4|113.73|80.99|39.97|109.12|68.86|96|47.67|72.72|70.16
5|97.17|82.15|75.85|89.78|109.09|122.14|70.07|68.43|76.49
6|93.3|128.17|80.99|108.43|81.58|91.26|69.55|95.4|86.07
7|70.39|81.14|94.79|78.99|109.89|47|53.36|69.94|83.12
8|88.04|97.76|94.47|103.51|95.09|107.02|79.19|70.09|73.68
9|94.61|73.17|107.14|77.37|122.01|51.17|65.32|95.45|64.42
10|86.67|103.3|112.48|68.99|109.1|72.92|61.98|84.94|111.16
11|73.64|116.35|95.6|58.58|100.33|68.01|76.23|54.02|90.57
12|86.16|105.7|105.66|62.03|117.8|99.49|89.61|85.38|95.46
13|84.21|82.73|87.89|34.15|83.83|61.65|84.21|94.49|96.32
14|83.06|111.01|85.13|16.26|103.86|41.1|66.94|110.92|84.77
15|30.85|87.21|40.02|3.76|99.39|-|89.73|113.88|75.08
16|-|68.65|76.79|-|16.98|-|86.79|102.18|91.96
17|-|-|54.09|-|-|-|75|111.32|78.02
18|-|-|-|-|-|-|93.65|62.98|57.82
19|-|-|-|-|-|-|91.38|93.84|76.62
20|-|-|-|-|-|-|85.09|75.21|34.71
21|-|-|-|-|-|-|59.5|102.8|-
22|-|-|-|-|-|-|69.47|69.03|-
23|-|-|-|-|-|-|74.64|53.57|-
24|-|-|-|-|-|-|5.55|-|-
Career|1311.39|1377.81|1347.33|1153.81|1448.2|992.98|1640.31|1908.03|1507.1

#2-#T

Season|Stewart|Iginla|Alfredsson|Peter Stastny|Bobby Hull|Lach|Delvecchio|Francis|Ullman
1|143.2|57.66|59.66|105.67|64.74|48.09|0|62.56|25.57
2|72.69|41.45|81.87|127.87|62.5|2.01|60.46|89.29|69.89
3|110.86|63.08|58.29|123.02|108.29|89.78|97.04|79.62|70.25
4|117.03|81.7|40.82|114.15|71.25|97.83|51.06|76.02|72.5
5|100.4|83.29|76.51|93.85|111.11|127.8|71.43|71.69|77.54
6|95.64|129.43|82.12|113.59|83.78|94.76|72.44|99.79|89.06
7|71.74|81.95|95.73|82.62|112.69|48.23|55.11|71.98|84.66
8|90.57|98.65|95.4|106.53|98.34|108.54|81.27|73.06|75.68
9|97.79|74.11|108.11|80.65|127.63|53.31|67.5|97.12|66.06
10|89.66|104.44|113.93|70.19|116.62|75.71|62.83|87.53|114.96
11|75.19|117.68|96.67|60.36|101.85|72.35|78.88|54.83|94.74
12|87.81|106.96|106.87|62.96|122.09|106.21|91.27|86.76|102.04
13|86.89|83.73|88.93|34.7|87.85|67.43|86.49|95.72|97.77
14|84.68|112.34|86.15|16.47|109.28|44.01|68.65|112.34|87.86
15|32.1|88.48|40.5|3.81|102.72|-|92.8|116.39|78.67
16|-|69.24|77.92|-|17.29|-|90.79|103.78|96.76
17|-|-|54.55|-|-|-|80.17|112.69|80.63
18|-|-|-|-|-|-|95.06|64.32|59.27
19|-|-|-|-|-|-|94.71|94.67|79.86
20|-|-|-|-|-|-|89.17|76.26|35.42
21|-|-|-|-|-|-|62.61|103.82|-
22|-|-|-|-|-|-|71.79|69.72|-
23|-|-|-|-|-|-|76.51|54.05|-
24|-|-|-|-|-|-|5.79|-|-
Career|1356.24|1394.19|1364.03|1196.44|1498.02|1036.06|1703.8|1953.98|1559.17

 

[BM67]

Alex Delvecchio by various methods.

Vs#2 is % score of #2 scorer
Vs#3 is % score of #3 scorer
#1-#T is % score of the average of the scorers from #1 to the number of teams in the NHL (6 from 51 to 67, then 12-16)
VsX is % score of a benchmark set Here
H-RAP is adjusted points from Hockey-Reference

Season|Vs#2|Vs#3|#1-#T|VsX|H-RAP
1|0|0|0|0|0
2|53.62|56.92|56.64|53.62|45
3|83.99|96.72|88.72|96.72|83
4|43.28|46.77|47.67|47.54|40
5|64.86|65.75|70.07|64.86|65
6|64.56|71.83|69.55|71.83|68
7|48.24|48.81|53.36|53.25|50
8|73.75|75.64|79.19|83.1|70
9|59.34|61.36|65.32|65.06|61
10|58.75|63.51|61.98|58.75|52
11|68.89|73.81|76.23|68.89|66
12|82.14|89.61|89.61|82.14|74
13|79.01|84.21|84.21|79.01|70
14|60.92|67.95|66.94|67.95|61
15|80.72|88.16|89.73|80.72|76
16|88.46|88.46|86.79|88.46|73
17|68.75|78.57|75|78.57|59
18|83.33|85.37|93.65|83.33|77
19|77.57|80.58|91.38|77.57|81
20|68.69|79.07|85.09|79.07|70
21|39.57|47.41|59.5|55|50
22|55.56|59.63|69.47|55.56|66
23|68.27|70.3|74.64|68.27|65
24|4.1|4.76|5.55|4.72|5
Career|1475.48|1585.22|1640.31|1564|1427

Same but ranked highest to lowest score

Method|1|2|3|4|5|6|7|8|9|10|11|12|13|14|15|16|17|18|19|20|21|22|23|24|Total
Vs#2|88.46|83.33|83.10|82.14|80.72|79.01|77.57|73.75|68.89|68.75|68.69|68.27|64.86|64.56|60.92|59.34|58.75|55.56|53.62|48.24|43.28|39.57|4.10|0|1475.48
Vs#3|96.72|89.61|88.46|88.16|85.37|84.21|80.58|79.07|78.57|75.64|73.81|71.83|70.3|67.95|65.75|63.51|61.36|59.63|56.92|48.81|47.41|46.77|4.76|0|1585.22
#1-#T|93.65|91.38|89.73|89.61|88.72|86.79|85.09|84.21|79.19|76.23|75|74.64|70.07|69.55|69.47|66.94|65.32|61.98|59.5|56.64|53.36|47.67|5.55|0|1640.31
VsX|96.72|88.46|83.33|83.1|82.14|80.72|79.07|79.01|78.57|77.57|71.83|68.89|68.27|67.95|65.06|64.86|58.75|55.56|55|53.62|53.25|47.54|4.72|0|1564
H-RAP|83|81|77|76|74|73|70|70|70|68|66|66|65|65|61|61|59|52|50|50|45|40|5|0|1427