Rengorlex
Registered User
- Aug 25, 2021
- 4,775
- 8,633
Found this graph in Twitter. WJC performance just doesn't seem to matter at all.
Correlation between World Junior production and NHL success is non-existent
- Thread starter Rengorlex
- Start date
Hockey Outsider
Registered User
- Jan 16, 2005
- 9,155
- 14,477
My guess (and this is only a guess) is that, over a sufficiently large sample size, there would be a moderate correlation between points in the World Juniors, and production at the NHL level (controlling for age - see next paragraph). The biggest problem is the WJ tournaments are short. Therefore any type of forward-looking analysis is based on a tiny sample size (which, in part, would explain the wide error bars).
Two other issues. First, "points in the WJ" doesn't take a player's age into account. 4 points in 7 games would suggest two very different career trajectories depending on if it's, say, a 17 year old vs a 19 year old putting up those numbers.
The other issue is this seems to be comparing offensive production with GAR. I'm not completely sold on the validity of any "all in one" hockey stat, but this doesn't appear to be an apples-to-apples comparison, since raw offensive production (points) is being compared to GAR (which, if I understand correctly, takes into account defensive play, penalties drawn, etc). The input (points in WJ) is measuring one thing, and the output (WAR at NHL level) is measuring something different. Maybe the correlation would be higher if strictly offensive production in the NHL was measured.
Two other issues. First, "points in the WJ" doesn't take a player's age into account. 4 points in 7 games would suggest two very different career trajectories depending on if it's, say, a 17 year old vs a 19 year old putting up those numbers.
The other issue is this seems to be comparing offensive production with GAR. I'm not completely sold on the validity of any "all in one" hockey stat, but this doesn't appear to be an apples-to-apples comparison, since raw offensive production (points) is being compared to GAR (which, if I understand correctly, takes into account defensive play, penalties drawn, etc). The input (points in WJ) is measuring one thing, and the output (WAR at NHL level) is measuring something different. Maybe the correlation would be higher if strictly offensive production in the NHL was measured.
If we want to improve the predictions, could it make sense to limit the data to North American skaters? Both US and Canada have such a huge player pools in youth hockey that year-to-year variance is probably smaller in terms of role/usage. With Finns and Swedes, and even greater extent with other hockey nations, the strength of the program may vary from year to year, potentially skewing the numbers achieved by individual players (as an example see Kiviharju almost making the Team Finland this year as way underage defenseman). The counterpoint naturally is that we are limiting the sample size even further.
Jason MacIsaac
Registered User
It is such a result based event, I would be interested to see what if Mitch Brown's gamescore (without points) against the top 8 teams each year has a correlation. I believe playing well has its value just points is a terrible measurement in a short tournament that is PP driven.
McGuillicuddy
Registered User
- Sep 6, 2005
- 1,296
- 198
For the reasons you mention GAR is problematic. I wonder if something simpler like NHL games would be a better indicator. Longer careers can correlate with offensive production, defensive play, or anything in between. I think a successful career can be better captured by that measure better than any purely offensive or defensive metric.
