View Single Post
Old
02-15-2012, 01:52 PM
  #69
BM67
Registered User
 
BM67's Avatar
 
Join Date: Mar 2002
Location: In "The System"
Country: Canada
Posts: 4,553
vCash: 500
Quote:
Originally Posted by seventieslord View Post
Not really, when the point is to attempt to assign players a “score” based on how impressive their totals are in a historical context. 1987 is definitely a weird year when you look at how every non-Gretzky fared. The issue is not just in the top-5 scorers.

The number of 90, 80, and 70-point scorers went from an average of 21-33-56 in 1985 and 1986, to 13-23-52 in 1987, to an average of 17-34-57 in 1988 and 1989.

Point is, I have no problem with how the 1987 scores turn out.



Aren’t we talking about the effect of removing Bossy in 1986? If I did, the comparable would be one point higher, meaning minimal differences in the scores.

If you’re referring to 1989, I disagree that the rest of the leaderboard is below average. It looks like any other season in that range to me. The effect of bumping up those players’ scores is desired, the huge undesired effect I was referring to would be the embarrassingly low scores they would have if based on Wayne’s 169 (or even that 140 average)



On the surface, it looks like you are probably right that using 139 in 1982 is not a good idea. At the time I didn’t see it as a “crazy” enough result to break the pattern and “arbitrarily” remove it (see 1989 Yzerman, 1996 Jagr, 2006 Jagr) and one would think that this season to fall more “in line” with other seasons, a lower comparable would likely have to be used. But it actually falls in line with the other seasons pretty nicely:

year 90+ 80+ 70+ 60+
1978 3 3 12 21
1979 4 6 10 23
1980 3 8 20 36
1981 4 8 18 30
1982 5 8 15 38
1983 4 12 24 56
1984 9 15 35 59
1985 6 12 26 39
1986 8 16 22 53
1987 10 19 43 71
1988 7 12 25 54
1989 9 22 42 69
1990 4 13 26 44
1991 9 14 26 49
1992 12 23 45 67
1993 6 11 22 44
1994 8 21 41 71

I know it can never be perfectly linear, and I wouldn’t want to try to “force” it to be, either, but I am satisfied with how this looks. It reflects that over this time, the number of players in the NHL capable of scoring at a level of x% of the #2 non-outlier has steadily increased. You can definitely pick out a couple odd spots, some are explainable (1980 was a spike, due to the absorption of the WHA), some aren’t (1987 is slightly but not obscenely out of line)

If I used 140 as the comparable in 1989, the numbers would look like this: 4, 6, 10, 29. Far too out of line with the seasons around it and, IMO, not reflective of how that season went.

A start would be taking the average of the #2-5 scorers that came across their scores “honestly” (Gretzky 168, Yzerman 155, Mullen 110, Kurri 102 = 134) Then we’d end up with 4, 7, 13, 34 – which I would still very much disagree with, but is more “right” than the above. Personally, I like that year the way I have it.
Lets look at 1987. The 71 players above 60% is tied for the highest in the 17 years on your table. 71st in scoring is 63 points. 63 points from 1985-1989 gets you: 70th, 85th, 71st, 71st, & 76th. 71st place is 63, 67, 63, 63 & 65 points. 60%+ equals: 76 pts, 73 pts, 63 pts, 72 pts, & 66 pts. That doesn't vary as much as your 60%+: 39, 53, 71, 54 & 69 players.

That's 63 pts: 70th-85th; 71st: 63-67 pts; 60%+: 66-76 pts; 60%+: 39-71 players.

Leaving the #2 at #2 and 108 points would only move that to 65 points and 64 players for 1987. (That 64 players would still be the 4th highest for the 17 years.) I don't see anything out of line there, so you fixed something that wasn't broken. Maybe your tweak is better, but I'm doubtful at this point.

On the other hand with only 39 players making the 60%+ mark in 1985, with 2 outliers already removed, I'd think that might need a closer look.

One will need to look at other points in the data, and compare them to the straight #2 numbers to get a better picture, but just with the numbers above it looks like you didn't "fix" at least two years. I also wonder how many years have you gone too far, as I believe you have in 1989.

In 1989 the top 4 are way ahead of average, and the rest of the pack are slightly below average. So a line like 4, 7, 13, 34 is much more in line with reality than 9, 22, 42, 69, which makes it look like the leaders were below average and the pack were above.

BM67 is offline