Vs#2 Thread

[seventieslord]

Looking at this chart - I'm going to say it now. VS5 or VS10 is a superior method to VS2. The data at 5 and 10 follows a much more gentle curve and doesn't appear as prone to random fluxuations as the second place scorer.

Actually, looking at it again, there are some years when even the 5th place scorer fluxuates. VS10 is a nice smooth curve through the entire time period - where the only peaks are years like 1993, when there should be one.

Problem with this is, pre-expansion, you are going to see some unholy scores for the top3 scorers a few times. Particularly from 49-55, the scoring lists were awful beyond the top few (this is the inverse of the problem you perceived last time when mackell had crap scores in two years he was a top-10 scorer)

 

[TheDevilMadeMe]

Problem with this is, pre-expansion, you are going to see some unholy scores for the top3 scorers a few times. Particularly from 49-55, the scoring lists were awful beyond the top few (this is the inverse of the problem you perceived last time when mackell had crap scores in two years he was a top-10 scorer)

I wasn't thinking of pre-expansion, because I think vs2 is of limited use for the Original 6 era. vs5 and vs10 would be worse though.

The wild fluxuation of the #2 scorer make its use questionable, however.

 

[BM67]

Are wild fluctuations in the 70s really a problem with the system?

The league had expansion every couple of years. They had players moving back and forth with the WHA. The schedule wasn't balanced, and they realigned from 2 divisions to 4. Throw in the folding of the Seals/Barons and the first wave of European players for added spice.

Fluctuation is to be expected.

The #10-15-20-25 scorer all jumped by 11 points from 1971-72 to 1972-73. At the same time the top 6 scoring spots all had a decline in points.

The 72-73 NHL is a very different place than the 71-72 NHL. Adding in the Flames and Islanders as well as the WHA completely changes the landscape, and we haven't even looked at possible rule changes.

Point totals for selected spots in the top 25 68-80

Year |Teams|GP| #1 | #2 | #3 | #4 | #5 | #10 | #15 | #20 | #25
1968 |12|74| 87 | 84 | 82 | 78 | 77 | 69 | 64 | 57 | 54
1969 |12|76| 126 | 107 | 103 | 97 | 90 | 78 | 70 | 65 | 62
1970 |12|76| 120 | 99 | 86 | 78 | 77 | 70 | 67 | 63 | 59
1971 |14|78| 152 | 139 | 116 | 105 | 96 | 76 | 72 | 66 | 63
1972 |14|78| 133 | 117 | 109 | 106 | 97 | 81 | 74 | 70 | 66
1973 |16|78| 130 | 104 | 101 | 100 | 95 | 92 | 85 | 81 | 77
1974 |16|78| 145 | 122 | 105 | 89 | 87 | 82 | 77 | 75 | 72
1975 |18|80| 135 | 127 | 121 | 119 | 117 | 95 | 86 | 79 | 77
1976 |18|80| 125 | 119 | 113 | 112 | 111 | 99 | 91 | 83 | 82
1977 |18|80| 136 | 122 | 105 | 97 | 95 | 90 | 81 | 78 | 74
1978 |18|80| 132 | 123 | 117 | 97 | 94 | 87 | 81 | 77 | 73
1979 |17|80| 134 | 130 | 129 | 126 | 108 | 91 | 85 | 80 | 77
1980 |21|80| 137 | 137 | 125 | 106 | 105 | 94 | 92 | 89 | 80

 

[Hockey Outsider]

My take on the "Vs. #2" is here: http://hfboards.mandatory.com/showthread.php?t=589911&highlight=

I used the average of 2nd through 4th place. My rationale is that if you base a ranking on a single position (whether it's 2nd or 5th or 10th), there will always be outliers. By taking an average of a few positions, the numbers will be less likely to be skewed by a particularly good/bad performance. It also mitigates the need for you to arbitrarily identify and exclude "outlier" seasons.

If you use a lower position (say average of 5th to 10th), the numbers become seriously skewed pre-expansion (in many years, there may not have been a lot of elite talent beyond the top few spots).

All of the results (current as of 2008 or whenever I last updated this) are in the link. I think they look intuitively reasonable.

 

[TheDevilMadeMe]

My take on the "Vs. #2" is here: http://hfboards.mandatory.com/showthread.php?t=589911&highlight=

I used the average of 2nd through 4th place. My rationale is that if you base a ranking on a single position (whether it's 2nd or 5th or 10th), there will always be outliers. By taking an average of a few positions, the numbers will be less likely to be skewed by a particularly good/bad performance. It also mitigates the need for you to arbitrarily identify and exclude "outlier" seasons.

If you use a lower position (say average of 5th to 10th), the numbers become seriously skewed pre-expansion (in many years, there may not have been a lot of elite talent beyond the top few spots).

All of the results (current as of 2008 or whenever I last updated this) are in the link. I think they look intuitively reasonable.

Good work. This system still doesn't work for me in 1989 though - when the top 4 were all outliers. It still also compares everyone to a bunch of Bruins in a couple of seasons in the 1970s too.

 

[seventieslord]

I don't see a problem with removing those players and them comparing to a 2-4 or 2-5 group.

 

[TheDevilMadeMe]

I don't see a problem with removing those players and them comparing to a 2-4 or 2-5 group.

Arbitrarily removing players because they scored too much in a certain year makes the method less scientific. I think there's merit to creating a system with basic rules as to when to remove a player, like I tried above.

Maybe you're right though; maybe we should just give up trying to find a formula.

I do think I'm going to start using VS5 when comparing post-expansion players to each other.

 

[BM67]

Using the number of teams to set the mark. We start at either #1 or #2 through to T, the number of teams, and use the average as the "vs#2" number.

The table shows how many points that would be, and where that would sit in the scoring race.

Year | Teams | #1-#T | Rank | #2-#T | Rank
1918 | 3 | 44.67 | 3 | 43 | 3
1919 | 3 | 29.67 | 2 | 28 | 2
1920 | 4 | 43 | 3 | 41 | 3
1921 | 4 | 39.75 | 3 | 38.67 | 4
1922 | 4 | 39.5 | 2 | 37.33 | 4
1923 | 4 | 32.5 | 3 | 31 | 3
1924 | 4 | 22.75 | 3 | 22.33 | 3
1925 | 6 | 38.83 | 5 | 37.4 | 5
1926 | 7 | 31.14 | 3 | 29.33 | 5
1927 | 10 | 29.7 | 6 | 28.89 | 6
1928 | 10 | 32.7 | 6 | 30.67 | 6
1929 | 10 | 25.5 | 7 | 24.78 | 7
1930 | 10 | 56.6 | 6 | 54.78 | 7
1931 | 10 | 41.8 | 6 | 40.78 | 7
1932 | 8 | 46.88 | 6 | 46 | 6
1933 | 9 | 40.89 | 6 | 39.75 | 6
1934 | 9 | 41.22 | 4 | 39.88 | 4
1935 | 9 | 45 | 4 | 43.5 | 6
1936 | 8 | 39.38 | 4 | 38.57 | 4
1937 | 8 | 40.62 | 5 | 39.86 | 6
1938 | 8 | 42.75 | 4 | 41.43 | 5
1939 | 7 | 42.14 | 3 | 41.33 | 4
1940 | 7 | 42.14 | 4 | 40.5 | 4
1941 | 7 | 46.29 | 2 | 43.67 | 7
1942 | 7 | 50.57 | 5 | 49.67 | 5
1943 | 6 | 66 | 4 | 64.6 | 4
1944 | 6 | 75 | 3 | 73.6 | 4
1945 | 6 | 65.5 | 4 | 62.6 | 5
1946 | 6 | 51.5 | 3 | 49.6 | 5
1947 | 6 | 63.83 | 3 | 62.2 | 4
1948 | 6 | 57 | 3 | 56.2 | 4
1949 | 6 | 56.67 | 3 | 54.4 | 3
1950 | 6 | 65.83 | 4 | 63.4 | 5
1951 | 6 | 66.17 | 2 | 62.2 | 3
1952 | 6 | 65.33 | 3 | 61.2 | 4
1953 | 6 | 66.5 | 3 | 60.8 | 4
1954 | 6 | 60.83 | 4 | 56.8 | 4
1955 | 6 | 68.5 | 4 | 67.2 | 4
1956 | 6 | 73.33 | 3 | 70.4 | 4
1957 | 6 | 76.83 | 5 | 74.4 | 5
1958 | 6 | 74.5 | 5 | 72.6 | 5
1959 | 6 | 82.67 | 4 | 80 | 4
1960 | 6 | 75.83 | 3 | 74.8 | 3
1961 | 6 | 81.33 | 4 | 78.6 | 4
1962 | 6 | 77 | 3 | 75.6 | 5
1963 | 6 | 76 | 3 | 74 | 4
1964 | 6 | 79.17 | 3 | 77.2 | 4
1965 | 6 | 74.67 | 4 | 72.2 | 4
1966 | 6 | 79.5 | 2 | 76 | 5
1967 | 6 | 73.33 | 3 | 68.6 | 4
1968 | 12 | 74.75 | 7 | 73.64 | 7
1969 | 12 | 90.83 | 5 | 87.64 | 6
1970 | 12 | 79.92 | 4 | 76.27 | 6
1971 | 14 | 92.43 | 6 | 87.85 | 6
1972 | 14 | 93.57 | 7 | 90.54 | 8
1973 | 16 | 95.12 | 5 | 92.8 | 10
1974 | 16 | 90.06 | 4 | 86.4 | 6
1975 | 18 | 100.83 | 7 | 98.82 | 8
1976 | 18 | 101.28 | 9 | 99.88 | 10
1977 | 18 | 93 | 8 | 90.47 | 8
1978 | 18 | 92.83 | 6 | 90.53 | 7
1979 | 17 | 100.88 | 8 | 98.81 | 8
1980 | 21 | 100.14 | 8 | 98.3 | 9
1981 | 21 | 106.05 | 7 | 103.15 | 10
1982 | 21 | 113.62 | 9 | 108.7 | 9
1983 | 21 | 105.33 | 8 | 100.8 | 12
1984 | 21 | 109.05 | 9 | 104.25 | 11
1985 | 21 | 111.38 | 7 | 106.55 | 7
1986 | 21 | 112.52 | 8 | 107.4 | 9
1987 | 21 | 97.48 | 8 | 93.2 | 11
1988 | 21 | 107.24 | 8 | 104.2 | 11
1989 | 21 | 109.86 | 8 | 105.4 | 8
1990 | 21 | 105.81 | 8 | 104 | 9
1991 | 21 | 102.43 | 10 | 99.4 | 11
1992 | 22 | 99.95 | 10 | 98.48 | 11
1993 | 24 | 117.12 | 11 | 115.26 | 11
1994 | 26 | 98.42 | 11 | 97.16 | 11
1995 | 26 | 53.19 | 10 | 52.52 | 12
1996 | 26 | 104.5 | 11 | 102.24 | 12
1997 | 26 | 88.08 | 10 | 86.72 | 11
1998 | 26 | 78.15 | 11 | 77.2 | 13
1999 | 27 | 82.56 | 13 | 80.85 | 14
2000 | 28 | 77.79 | 13 | 77.11 | 13
2001 | 30 | 86.43 | 13 | 85.24 | 13
2002 | 30 | 74.9 | 16 | 74.17 | 16
2003 | 30 | 82.57 | 13 | 81.76 | 13
2004 | 30 | 74.67 | 15 | 74 | 15
2006 | 30 | 91.57 | 12 | 90.41 | 13
2007 | 30 | 91 | 14 | 90 | 15
2008 | 30 | 84.23 | 11 | 83.28 | 12
2009 | 30 | 84.2 | 12 | 83.21 | 13
2010 | 30 | 83.4 | 14 | 82.41 | 15
2011 | 30 | 77.47 | 10 | 76.55 | 11
2012 | 30 | 76.83 | 15 | 75.72 | 18
2013 | 30 | 48.07 | 13 | 47.66 | 17

One mod I am thinking of is using T-1 for split league years, PCHA and WHA, and the "war years".

What do you think of this?

 

[seventieslord]

Using the number of teams to set the mark. We start at either #1 or #2 through to T, the number of teams, and use the average as the "vs#2" number.

The table shows how many points that would be, and where that would sit in the scoring race.

Year | Teams | #1-#T | Rank | #2-#T | Rank
1918 | 3 | 44.67 | 3 | 43 | 3
1919 | 3 | 29.67 | 2 | 28 | 2
1920 | 4 | 43 | 3 | 41 | 3
1921 | 4 | 39.75 | 3 | 38.67 | 4
1922 | 4 | 39.5 | 2 | 37.33 | 4
1923 | 4 | 32.5 | 3 | 31 | 3
1924 | 4 | 22.75 | 3 | 22.33 | 3
1925 | 6 | 38.83 | 5 | 37.4 | 5
1926 | 7 | 31.14 | 3 | 29.33 | 5
1927 | 10 | 29.7 | 6 | 28.89 | 6
1928 | 10 | 32.7 | 6 | 30.67 | 6
1929 | 10 | 25.5 | 7 | 24.78 | 7
1930 | 10 | 56.6 | 6 | 54.78 | 7
1931 | 10 | 41.8 | 6 | 40.78 | 7
1932 | 8 | 46.88 | 6 | 46 | 6
1933 | 9 | 40.89 | 6 | 39.75 | 6
1934 | 9 | 41.22 | 4 | 39.88 | 4
1935 | 9 | 45 | 4 | 43.5 | 6
1936 | 8 | 39.38 | 4 | 38.57 | 4
1937 | 8 | 40.62 | 5 | 39.86 | 6
1938 | 8 | 42.75 | 4 | 41.43 | 5
1939 | 7 | 42.14 | 3 | 41.33 | 4
1940 | 7 | 42.14 | 4 | 40.5 | 4
1941 | 7 | 46.29 | 2 | 43.67 | 7
1942 | 7 | 50.57 | 5 | 49.67 | 5
1943 | 6 | 66 | 4 | 64.6 | 4
1944 | 6 | 75 | 3 | 73.6 | 4
1945 | 6 | 65.5 | 4 | 62.6 | 5
1946 | 6 | 51.5 | 3 | 49.6 | 5
1947 | 6 | 63.83 | 3 | 62.2 | 4
1948 | 6 | 57 | 3 | 56.2 | 4
1949 | 6 | 56.67 | 3 | 54.4 | 3
1950 | 6 | 65.83 | 4 | 63.4 | 5
1951 | 6 | 66.17 | 2 | 62.2 | 3
1952 | 6 | 65.33 | 3 | 61.2 | 4
1953 | 6 | 66.5 | 3 | 60.8 | 4
1954 | 6 | 60.83 | 4 | 56.8 | 4
1955 | 6 | 68.5 | 4 | 67.2 | 4
1956 | 6 | 73.33 | 3 | 70.4 | 4
1957 | 6 | 76.83 | 5 | 74.4 | 5
1958 | 6 | 74.5 | 5 | 72.6 | 5
1959 | 6 | 82.67 | 4 | 80 | 4
1960 | 6 | 75.83 | 3 | 74.8 | 3
1961 | 6 | 81.33 | 4 | 78.6 | 4
1962 | 6 | 77 | 3 | 75.6 | 5
1963 | 6 | 76 | 3 | 74 | 4
1964 | 6 | 79.17 | 3 | 77.2 | 4
1965 | 6 | 74.67 | 4 | 72.2 | 4
1966 | 6 | 79.5 | 2 | 76 | 5
1967 | 6 | 73.33 | 3 | 68.6 | 4
1968 | 12 | 74.75 | 7 | 73.64 | 7
1969 | 12 | 90.83 | 5 | 87.64 | 6
1970 | 12 | 79.92 | 4 | 76.27 | 6
1971 | 14 | 92.43 | 6 | 87.85 | 6
1972 | 14 | 93.57 | 7 | 90.54 | 8
1973 | 16 | 95.12 | 5 | 92.8 | 10
1974 | 16 | 90.06 | 4 | 86.4 | 6
1975 | 18 | 100.83 | 7 | 98.82 | 8
1976 | 18 | 101.28 | 9 | 99.88 | 10
1977 | 18 | 93 | 8 | 90.47 | 8
1978 | 18 | 92.83 | 6 | 90.53 | 7
1979 | 17 | 100.88 | 8 | 98.81 | 8
1980 | 21 | 100.14 | 8 | 98.3 | 9
1981 | 21 | 106.05 | 7 | 103.15 | 10
1982 | 21 | 113.62 | 9 | 108.7 | 9
1983 | 21 | 105.33 | 8 | 100.8 | 12
1984 | 21 | 109.05 | 9 | 104.25 | 11
1985 | 21 | 111.38 | 7 | 106.55 | 7
1986 | 21 | 112.52 | 8 | 107.4 | 9
1987 | 21 | 97.48 | 8 | 93.2 | 11
1988 | 21 | 107.24 | 8 | 104.2 | 11
1989 | 21 | 109.86 | 8 | 105.4 | 8
1990 | 21 | 105.81 | 8 | 104 | 9
1991 | 21 | 102.43 | 10 | 99.4 | 11
1992 | 22 | 99.95 | 10 | 98.48 | 11
1993 | 24 | 117.12 | 11 | 115.26 | 11
1994 | 26 | 98.42 | 11 | 97.16 | 11
1995 | 26 | 53.19 | 10 | 52.52 | 12
1996 | 26 | 104.5 | 11 | 102.24 | 12
1997 | 26 | 88.08 | 10 | 86.72 | 11
1998 | 26 | 78.15 | 11 | 77.2 | 13
1999 | 27 | 82.56 | 13 | 80.85 | 14
2000 | 28 | 77.79 | 13 | 77.11 | 13
2001 | 30 | 86.43 | 13 | 85.24 | 13
2002 | 30 | 74.9 | 16 | 74.17 | 16
2003 | 30 | 82.57 | 13 | 81.76 | 13
2004 | 30 | 74.67 | 15 | 74 | 15
2006 | 30 | 91.57 | 12 | 90.41 | 13
2007 | 30 | 91 | 14 | 90 | 15
2008 | 30 | 84.23 | 11 | 83.28 | 12
2009 | 30 | 84.2 | 12 | 83.21 | 13
2010 | 30 | 83.4 | 14 | 82.41 | 15
2011 | 30 | 77.47 | 10 | 76.55 | 11

One mod I am thinking of is using T-1 for split league years, PCHA and WHA, and the "war years".

What do you think of this?

This is what I call not being a slave.

Sorry I didn't respond to this earlier.

I am a fan of a system that recognizes that the relative quality of a 3rd place finish 50 years ago might only be about as good as a 6th today. With comparables in the 10th-16 range the last 20 seasons, I suspect this may have taken it too far. I think that the size of the talent pool has expanded at a rate at least somewhat similar the same rate as the league, but not exactly, so doesn't this end up falling into the "10th place in the O6? Ha, that's like 50th in today's 30 team league!" trap? We all hate that argument.

On first glance it looks like seasons like 1970, 1971, 1974 and 1989 suffer from a lot of outliers and outlier beneficiaries at the top, as that "rank" number spikes downwards compared to its progression.

It is a good start. I think that the ranking number we use should go up throughout time; maybe just not this drastically. I know it's a case of trying to devise a system and sticking to it rather than fudging it. That's fine. It might even be an overall improvement. To say for sure, I'd need to see numbers such as how many players meet the 50, 60, 70, 80 benchmarks per season.

 

[TheDevilMadeMe]

Seventieslord,

I was shocked to find out that Eric Lindros' 6th place finish somehow scores 100 in a system that is supposed to compare players against the #2 scorer.

I was equally shocked that Luc Robitaille (who never finished above 5th) scores 100 in a season.

Even if you're going by the "remove players who look like outliers," I can't see any season where "2" is really #6 and 1989 is the only one where I can see justification in using the 5th place scorer.

I get removing Lemieux and Jagr in 1996 but why would you not then use Sakic as the standard?

 

[Rob Scuderi]

Seventieslord,

I was shocked to find out that Eric Lindros' 6th place finish somehow scores 100 in a system that is supposed to compare players against the #2 scorer.

I was equally shocked that Luc Robitaille (who never finished above 5th) scores 100 in a season.

Even if you're going by the "remove players who look like outliers," I can't see any season where "2" is really #6 and 1989 is the only one where I can see justification in using the 5th place scorer.

I get removing Lemieux and Jagr in 1996 but why would you not then use Sakic as the standard?

Looking at the specific seasons you mention, you've definitely got me stumped. Like I said, I'm fairly ignorant on this topic but Lindros's 100% in '96 seems strange.

Question though, after removing outliers do we still use the second best left? Or is the first non-outlier given the #2 designation?

Robitaille's 100% is in '92. Here's the top scorers.

1. Mario Lemieux*-PIT 131
2. Kevin Stevens-PIT 123
3. Wayne Gretzky*-LAK 121
4. Brett Hull*-STL 109
5. Luc Robitaille*-LAK 107
Mark Messier*-NYR 107
7. Jeremy Roenick-CHI 103
Steve Yzerman*-DET 103
9. Brian Leetch*-NYR 102
10. Adam Oates-TOT 99

If Hull becomes #1 with outliers then I'd say Robitaille earned that 100% for sure, but if Hull is also an outlier then I'm back to being confused. I'm sure this is 70s favorite topic by now but hopefully he can explain a bit better.

 

[BenchBrawl]

I hate the fact of completely taking out outliers , in a way the fact they are outliers in your era advantage you against second-tier players from no-outlier's eras because their superstars aren't destroying them enough.

 

[BM67]

Looking at the specific seasons you mention, you've definitely got me stumped. Like I said, I'm fairly ignorant on this topic but Lindros's 100% in '96 seems strange.

Question though, after removing outliers do we still use the second best left? Or is the first non-outlier given the #2 designation?

Robitaille's 100% is in '92. Here's the top scorers.

1. Mario Lemieux*-PIT 131
2. Kevin Stevens-PIT 123
3. Wayne Gretzky*-LAK 121
4. Brett Hull*-STL 109
5. Luc Robitaille*-LAK 107
Mark Messier*-NYR 107
7. Jeremy Roenick-CHI 103
Steve Yzerman*-DET 103
9. Brian Leetch*-NYR 102
10. Adam Oates-TOT 99

If Hull becomes #1 with outliers then I'd say Robitaille earned that 100% for sure, but if Hull is also an outlier then I'm back to being confused. I'm sure this is 70s favorite topic by now but hopefully he can explain a bit better.

70s treats Mario and Gretzky as outliers by name, so they are out automatically. Stevens played with Mario, so he's out. That leaves Hull as the #1. Then Robitaille is also removed as an outlier, and #2 would be Messier.

If you replaced Mario and Wayne with Trottier and Dionne, the 70s method would give a #2 of Stevens, as 131 is not an outlier score other than by who did it, and Stevens is only an outlier by who he played with.

 

[Sturminator]

70s treats Mario and Gretzky as outliers by name, so they are out automatically. Stevens played with Mario, so he's out. That leaves Hull as the #1. Then Robitaille is also removed as an outlier, and #2 would be Messier.

If you replaced Mario and Wayne with Trottier and Dionne, the 70s method would give a #2 of Stevens, as 131 is not an outlier score other than by who did it, and Stevens is only an outlier by who he played with.

Yes, I think Stevens is the #2 this season, and that's it. I see no sensible reason to adjust this season in a Vs2 analysis. Ignoring the actual performance of Gretzky/Lemieux in any given season and ignoring everyone with whom they played simply removes too much top offensive talent from the analysis, and ends up leaving the remaining players looking better than they really were.

 

[TheDevilMadeMe]

I have mixed feelings about removing Mario/Stevens from 1994 - they were a step up from the pack but not a big step. Either way, I see no reason to also remove Brett Hull. This controversy over outliers is why I think the VS5 method could be better - you never have to worry about outliers in VS5.

 

[TheDevilMadeMe]

Here is 1996:

1. Lemieux 161
2. Jagr 149
3. Sakic 120
4. Francis 119
5. Forsberg 116
6. Lindros 115
7. Selanne 108
7. Kariya 108

I would imagine 70s methodology is to remove Mario and his teammate Jagr. Sakic is removed as the #1 remaining. Francis and Forsberg are removed as teammates of guys in front of them. So Lindros becomes the standard

I strongly disagree with this methodology. I think if you're using VS2, you need to compare to the second place scorer except in extreme circumstances. Jagr in 1996 actually is extreme, so then why not just go down to the next name and compare to Sakic? It sure doesn't look like Sakic, Francis, and Forsberg are outliers to me.

 

[seventieslord]

Here is 1996:

1. Lemieux 161
2. Jagr 149
3. Sakic 120
4. Francis 119
5. Forsberg 116
6. Lindros 115
7. Selanne 108
7. Kariya 108

I would imagine 70s methodology is to remove Mario and his teammate Jagr. Sakic is removed as the #1 remaining. Francis and Forsberg are removed as teammates of guys in front of them. So Lindros becomes the standard

I strongly disagree with this methodology. I think if you're using VS2, you need to compare to the second place scorer except in extreme circumstances. Jagr in 1996 actually is extreme, so then why not just go down to the next name and compare to Sakic? It sure doesn't look like Sakic, Francis, and Forsberg are outliers to me.

BM and sturm answered your first inquiry.

Re: Forsberg, I can't believe I did that, but apparently I did. I should have left him in as the #2 after outliers.

I think the idea that a Mario Lemieux-fueled Kevin Stevens should be the player that you judge others' offensive totals by, is ridiculous.

you never have to worry about outliers in VS5.

sure you do. I can get behind a #5 system in theory (aside from my concerns raised in the Turgeon thread in the HOH section) but that doesn't make the outliers any less damaging to the system. I would personally still want to remove them. The goal is to not punish players for playing in a league with players who break the mold. If Gretzky and Lemieux did not exist, the most reasonable outcome would be Hull, Messier, Roenick, Yzerman, Leetch.

 

[TheDevilMadeMe]

I just don't understand why you don't use Joe Sakic as the standard in 1996. And if you count Mario Lemieux and Kevin Stevens as outliers in 1992, surely Brett Hull was not also one?

sure you do. I can get behind a #5 system in theory (aside from my concerns raised in the Turgeon thread in the HOH section) but that doesn't make the outliers any less damaging to the system. I would personally still want to remove them. The goal is to not punish players for playing in a league with players who break the mold. If Gretzky and Lemieux did not exist, the most reasonable outcome would be Hull, Messier, Roenick, Yzerman, Leetch.

Why? Name a single season in history where #5 is an outlier. 1989 is the only season I can think of where even #4 is an outlier

 

[BraveCanadian]

Here is 1996:

1. Lemieux 161
2. Jagr 149
3. Sakic 120
4. Francis 119
5. Forsberg 116
6. Lindros 115
7. Selanne 108
7. Kariya 108

I would imagine 70s methodology is to remove Mario and his teammate Jagr. Sakic is removed as the #1 remaining. Francis and Forsberg are removed as teammates of guys in front of them. So Lindros becomes the standard

I strongly disagree with this methodology. I think if you're using VS2, you need to compare to the second place scorer except in extreme circumstances. Jagr in 1996 actually is extreme, so then why not just go down to the next name and compare to Sakic? It sure doesn't look like Sakic, Francis, and Forsberg are outliers to me.

That would be because they aren't.

 

[vecens24]

I agree with TDMM that Sakic should be the standard that season. That makes more sense to me.

 

[seventieslord]

I just don't understand why you don't use Joe Sakic as the standard in 1996.

Because Sakic is the #1 non-outlier, not #2.

And if you count Mario Lemieux and Kevin Stevens as outliers in 1992, surely Brett Hull was not also one?

sorry, but why?

Why? Name a single season in history where #5 is an outlier. 1989 is the only season I can think of where even #4 is an outlier

If #5 is behind nothing but outliers, then that could very well be as impressive as #1 in a season without outliers. Therefore, selecting them as the standard instead of the "#2" would likely be disadvantaging that year, to the extent that the player you want me to use is ahead of the player I want to use.

the more I think about all this, and how we are trying to make this "fit" what we think it should look like (as in, number of players by year who are 60+, 70+, 80+, etc) we almost need to go to an "adjusted" rankings system. Maybe that would please everyone...

All we'd need to do is agree on what kinds of multipliers should be used from era to era. i.e. if you're 8th in 1930, that's like __ by 1960, __ by 1980, and ___ in today's league.

 

[BM67]

This illustrates how much moving the "decade" by one season can change things.

Player | GP | G | A | Pts | Yrs | G1 | G2 | G3 | G4 | A1 | A2 | A3 | A4 | P1 | P2 | P3 | P4
Number 1 2001-2011 | 820 | 532 | 757 | 1,103 | 10 | 1000% | 1106% | 1138% | 1204% | 1000% | 1123% | 1175% | 1225% | 1000% | 1042% | 1114% | 1157%
Number 2 2001-2011 | 820 | 482 | 673 | 1,060 | 10 | 911% | 1000% | 1029% | 1090% | 898% | 1000% | 1047% | 1092% | 960% | 1000% | 1069% | 1110%
Number 3 2001-2011 | 820 | 468 | 642 | 991 | 10 | 885% | 972% | 1000% | 1060% | 858% | 956% | 1000% | 1043% | 901% | 939% | 1000% | 1039%
Number 4 2001-2011 | 820 | 442 | 617 | 954 | 10 | 836% | 919% | 946% | 1000% | 824% | 919% | 960% | 1000% | 867% | 903% | 963% | 1000%
Number 1 2002-2012 | 820 | 533 | 755 | 1,091 | 10 | 1000% | 1117% | 1171% | 1223% | 1000% | 1126% | 1187% | 1233% | 1000% | 1052% | 1105% | 1159%
Number 2 2002-2012 | 820 | 478 | 669 | 1,039 | 10 | 903% | 1000% | 1047% | 1095% | 895% | 1000% | 1056% | 1096% | 951% | 1000% | 1051% | 1102%
Number 3 2002-2012 | 820 | 457 | 633 | 988 | 10 | 866% | 958% | 1000% | 1047% | 848% | 948% | 1000% | 1038% | 907% | 953% | 1000% | 1049%
Number 4 2002-2012 | 820 | 437 | 611 | 943 | 10 | 827% | 916% | 957% | 1000% | 818% | 915% | 965% | 1000% | 865% | 909% | 954% | 1000%

 

[BM67]

Here are the top 30 scorers for the post-lockout seasons.

Year|#1|#2|#3|#4|#5|#6|#7|#8|#9|#10|#11|#12|#13|#14|#15|#16|#17|#18|#19|#20|#21|#22|#23|#24|#25|#26|#27|#28|#29|#30|Avg 1-30
2006|125|123|106|103|103|102|100|98|97|93|92|91|90|90|89|89|87|87|86|85|85|85|84|81|80|80|79|79|79|79|91.57
2007|120|114|108|105|102|100|100|96|96|95|94|94|92|91|87|87|87|85|84|84|83|83|83|82|81|81|80|80|78|78|91
2008|112|106|98|97|96|92|92|92|89|87|84|83|83|83|82|82|82|81|80|78|78|78|77|76|75|74|74|72|72|72|84.23
2009|113|110|103|97|94|91|91|89|88|88|86|84|82|82|82|80|80|79|79|78|78|77|77|76|75|74|74|73|73|73|84.2
2010|112|109|109|101|95|94|91|89|88|86|85|85|84|83|82|82|81|79|77|76|76|75|71|71|71|70|70|70|70|70|83.4
2011|104|99|98|94|91|86|85|80|80|77|76|76|76|76|73|73|73|73|73|73|71|71|71|70|69|68|68|67|67|66|77.47
2012|109|97|93|84|83|82|81|81|81|78|78|77|77|77|76|76|76|74|73|71|70|69|69|69|69|67|67|67|67|67|76.83

With the exception of a tie for 2nd lowest at #3 between 2008 and 2011, the lowest two scores at all spots come in the last two seasons. Indeed the #30 from 2006 & 2007 would be in the top 10 in 2011 & 2012.

13 of the lowest totals come in 2011, and 13 come in 2012, with 4 ties. The biggest gap between the lowest and 2nd lowest comes at #4 where the 84 in 2012 is a full 10 points back of the 94 put up in 2011.

The points average from 1-30 has dropped every year since 2006.

 

[Sturminator]

I'm reviving this thread because I think we need another round of discussion on this subject. The Vs2 method is starting to become pretty comfortable to us in the ATD and its use among the GMs is becoming relatively widespread. We might want to discuss some kind of "official ATD standard" for Vs2 statistical analysis so as to avoid misunderstandings when discussing player statistics. I will continue the previous argument.

Here is 1996:

1. Lemieux 161
2. Jagr 149
3. Sakic 120
4. Francis 119
5. Forsberg 116
6. Lindros 115
7. Selanne 108
7. Kariya 108

I would imagine 70s methodology is to remove Mario and his teammate Jagr. Sakic is removed as the #1 remaining. Francis and Forsberg are removed as teammates of guys in front of them. So Lindros becomes the standard

I strongly disagree with this methodology. I think if you're using VS2, you need to compare to the second place scorer except in extreme circumstances. Jagr in 1996 actually is extreme, so then why not just go down to the next name and compare to Sakic? It sure doesn't look like Sakic, Francis, and Forsberg are outliers to me.

I just don't understand why you don't use Joe Sakic as the standard in 1996.

Because Sakic is the #1 non-outlier, not #2.

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.

 

[Dreakmur]

I'm reviving this thread because I think we need another round of discussion on this subject. The Vs2 method is starting to become pretty comfortable to us in the ATD and its use among the GMs is becoming relatively widespread. We might want to discuss some kind of "official ATD standard" for Vs2 statistical analysis so as to avoid misunderstandings when discussing player statistics. I will continue the previous argument.



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. The point of using #2 is based off the assumption that #1 is often an outlier.

I always use #2, or the first non-outlier.