By The NumbersHockey Analytics... the Final Frontier. Explore strange new worlds, to seek out new algorithms, to boldly go where no one has gone before.
Estimating How the Available Hockey Population Pool Has Changed (Goalies/Top Scorers)
Estimating How the Available Hockey Population Pool Has Changed (Goalies/Top Scorers)
This is a rather crude model based on a few factors/assumptions:
A) Canadian population data by age group
B) The proportion of Canadian and non-Canadian players as measured by each metric (% total GP by goalies, finishes in top 2N of players in points for scorers)
C) Hockey population was assumed to be 100% Canadian before expansion (i.e. goalies born outside Canada who played before 1968 were considered immigrants and part of Canadian population)
D) British Columbia population has been excluded until 1977, and increased linearly until fully integrated (for the youngest age group) in 1986. Therefore, the "prime" and older age groups do not begin to or become fully integrated until later years.
E) The Canadian hockey population pool is assumed to remain in fixed proportion to the relevant age groups of the the general population. The non-Canadian hockey pop. pool is assumed to vary in proportion to the proportion of non-Canadian players as measured by each metric and to the Canadian pop. pool.
It's meant to give a general idea of the possible hockey population pool at goaltender at various times, not be an exact model. However, I did try to be as accurate as possible, given the limited data and limited time I used to perform this study. The number on the Y-axis is thousands of males of hockey age (weighted by age group, with 100% of males 25-34 considered of hockey age), and the effective/implied number for the non-Canadian countries as a whole. I tried to keep the number on the y-axis relevant, but the more important concept is the proportionality between Canadians & non-Canadian, and from one era to another.
Last edited by Czech Your Math: 08-21-2012 at 05:36 PM.
Reason: updated
The one part of the graph that doesn't quite pass the smell test is the sudden increase in the growth rate exactly at the 1967 expansion. The expansion shouldn't have had a direct effect on the talent pool, right?
The one part of the graph that doesn't quite pass the smell test is the sudden increase in the growth rate exactly at the 1967 expansion. The expansion shouldn't have had a direct effect on the talent pool, right?
The values on the Y-axis would be thousands of males of hockey age... except I didn't re-adjust them after weighting the numbers by age group... so it should probably be at least 50% higher, perhaps more, but the proportion is all that's really important in the grand scheme of things.
The number of teams does not affect the talent pool. Basically, the baby boom enabled expansion, and that is what you are seeing on the graph, the number of males of hockey age exploding around that time.
Now that I think about it, top line forwards may be very similar to this, except the non-Canadian portion would sharply increase in the early 90s to the level that goalies attained much later (depending on how one measured their impact) and then plateau. Grinders and possibly d-men may not have as much of an increase, although it would still be more/earlier than goalies, but may not reach the heights the goalies have in terms of % impact.
The one part of the graph that doesn't quite pass the smell test is the sudden increase in the growth rate exactly at the 1967 expansion. The expansion shouldn't have had a direct effect on the talent pool, right?
Two goalie system introduced four seasons before 1967 NHL expansion filtered down thru the hockey world effectively doubling the number of goalies per team.
To be more precise, I suspect you should exclude British Columbia, at least until a certain date. Not sure what that date is, or whether fully excluding is the best idea. But the province has a large population and few players in the earlier years, so it could skew the results.
Quote:
Originally Posted by Czech Your Math
This is the Canadian male population data (in thousands) from statcan.ca's site:
To be more precise, I suspect you should exclude British Columbia, at least until a certain date. Not sure what that date is, or whether fully excluding is the best idea. But the province has a large population and few players in the earlier years, so it could skew the results.
Thanks for the input. It appears BC had ~6-7% of the Canadian population in the early decades, so that shouldn't have too large an effect, although it's very possible BC's population was skewed toward younger males. Still, it's worth considering. When would you estimate BC began producing a significant number of hockey players, relative to its population?
Thanks for the input. It appears BC had ~6-7% of the Canadian population in the early decades, so that shouldn't have too large an effect, although it's very possible BC's population was skewed toward younger males. Still, it's worth considering. When would you estimate BC began producing a significant number of hockey players, relative to its population?
this won't definitively answer your question, but i drew out all-time provincial line ups a few years ago, and noticed that 'good' NHLrs from BC didn't start to show up often until the 70's, with most of the big names coming after, let me estimate because it's been a while, 1985. This may suggest that kids hockey got going, to a solid degree, in the late 50's, and maybe was on par with the rest, per capita, by the 70's?
this won't definitively answer your question, but i drew out all-time provincial line ups a few years ago, and noticed that 'good' NHLrs from BC didn't start to show up often until the 70's, with most of the big names coming after, let me estimate because it's been a while, 1985. This may suggest that kids hockey got going, to a solid degree, in the late 50's, and maybe was on par with the rest, per capita, by the 70's?
Thanks for your input. Due to the suggestions from both you and Iain, I have made some adjustments to the formulas. I assumed BC had the same demographic as the rest of Canada (although this is almost certainly untrue) and excluded BC until 1977. From 1977-86, I assumed a gradual, linear integration of the youngest demographic segment.
I've also changed the scale, so that it represents thousands of males of hockey age (it's weighted by age group, with 100% of males 25-34 considered of NHL age, and over half of those 18-24 considered of NHL age). It should be noted that the non-Canadian pool is an effective/implied amount. The actual hockey age population of the non-Canadian countries is much, much larger, but they are not fully devoted to hockey in comparison to Canada. This is all captured in the calculations, so no actual non-Canadian population data was used, only the % of non-Canadian goalies.
This gives an idea of the competition for goalie through the years. It's 1000s of males of hockey age per NHL team. One can see why the late 80s and early 90s is often referred to as a golden age of hockey. It's also apparent why the 70s were considered so diluted, given that the WHA siphoned off some talent. If/when I do a similar analysis of scoring forwards, I would expect the 90s to appear much stronger than they do in this analysis. It wasn't additional expansion or the lack of talent at forward that caused the "golden age" to end and scoring to decline dramatically, but other factors such as the lack of rules enforcement.
Year
Teams
1000/Team
1946
6
165
1947
6
168
1948
6
170
1949
6
172
1950
6
175
1951
6
177
1952
6
181
1953
6
185
1954
6
189
1955
6
194
1956
6
198
1957
6
201
1958
6
203
1959
6
206
1960
6
209
1961
6
212
1962
6
215
1963
6
218
1964
6
221
1965
6
224
1966
6
227
1967
6
236
1968
12
123
1969
12
127
1970
12
131
1971
14
115
1972
14
120
1973
16
109
1974
16
114
1975
18
106
1976
18
111
1977
18
114
1978
18
115
1979
17
127
1980
21
108
1981
21
110
1982
21
109
1983
21
108
1984
21
116
1985
21
120
1986
21
123
1987
21
122
1988
21
124
1989
21
123
1990
21
123
1991
21
123
1992
22
121
1993
24
114
1994
26
105
1995
26
107
1996
26
110
1997
26
117
1998
26
116
1999
27
109
2000
28
105
2001
30
102
2002
30
106
2003
30
110
2004
30
113
2005
30
126
2006
30
139
2007
30
141
2008
30
144
2009
30
151
2010
30
167
2011
30
173
2012
30
168
Last edited by Czech Your Math: 08-19-2012 at 04:01 PM.
I also excluded Peter Sidorkiewicz, Byram Dafoe and Olaf Kolzig from the group of non-Canadians, although they were born outside Canada. If anyone knows of any Czech, Finnish, Russian, Swedish or USA goalies which moved to Canada at a young age and primarily trained in Canada, then let me know, so that they can also be excluded.
I have a simplified model consisting of top scoring skaters, and will similarly exclude those who played multiple seasons in Canadian Juniors and/or minor leagues before the NHL. I excluded Paul McLean, Dany Heatley, Walt Tkaczuk, Claude Vilgrain, Willi Plett, Wojtek Wolski, Steve Thomas, Owen Nolan, Ivan Boldirev and Stan Mikita.
Last edited by Czech Your Math: 08-20-2012 at 11:50 AM.
This is what I've come up with so far for top scoring skaters. I used basically the same assumptions as in the OP, with a couple of changes. First, I shifted the age distribution more to the left (younger), to reflect the much younger peaks that forwards and offensive d-men have compared to goalies. Second, I measured the % of non-Canadian players than finished in the the top 2N in points (N = number of NHL teams, so currently that would be top 60 in points). Since this data is slightly more erratic than goalies' GP, I used a 5 year average instead of a 3 year.
Last edited by Czech Your Math: 11-08-2012 at 02:13 PM.
So, the goalie and top 2N scorer graphs have some similarities and some big differences. The top 2N scorers seems to have peaked, at least for now, and has started to decline, while the goalies took a while to get going and have really accelerated. I decided to average the results of each and graph the results:
I believe this gives a general idea of the total available hockey talent for these high impact positions. It suggests that the top talent from overseas was almost fully integrated about two decades ago and has remained relatively level since. I think these are two of the best ways to measure the effective non-Canadian talent pool, because top goalies and top scorers are more likely to come to the NHL, as opposed to 3rd/4th line type players who might choose to stay in their home countries.
Last edited by Czech Your Math: 08-21-2012 at 05:30 PM.
regarding the first chart: I've always been one to say that you can't just take a 1st team all-star and multiply him by some sort of factor to determine what that is like in modern terms as far as " historical impressiveness" goes. But for players who rank, say, 20th, or 40th, or 110th in the world at a given time, the spikes have smoothed and they should be improving over time on a curve that looks similar to this.
So, if we're talking about the 20th-best player in 1934, is it fair to say that the data in the first chart says that he was about as relatively good as the 100th-best player in 2012? the factors they both sit at are about 1.0 and 5.0.
regarding the first chart: I've always been one to say that you can't just take a 1st team all-star and multiply him by some sort of factor to determine what that is like in modern terms as far as " historical impressiveness" goes. But for players who rank, say, 20th, or 40th, or 110th in the world at a given time, the spikes have smoothed and they should be improving over time on a curve that looks similar to this.
So, if we're talking about the 20th-best player in 1934, is it fair to say that the data in the first chart says that he was about as relatively good as the 100th-best player in 2012? the factors they both sit at are about 1.0 and 5.0.
Thanks, I'm glad you find it valuable.
The first chart only applies to goalies. I would use the last (third) chart as a better general guide to general "premier" players. That chart would suggest the 20th best player in the early 30s may have been roughly equal to the 100th best player in a season during the past two decades. There's various assumptions made and I leave it up to the individual to evaluate and decide how to further adjust based on their own assumptions. One might think the 20th best player in 1934 was actually more comparable to the 80th best player in 2012 or the 120th best player. Either of those seem very possible, depending on what assumptions are made. If one tries to say the 20th best player in 1934 was equal to the 20th best player today, or even the 50th best player, I would say that is highly, highly improbable.
I also want to mention why the last chart is the best guide for premier players, and very possibly is best in general. First, it's an average of goalies (which are clearly defensive) and top scorers (which are mostly offensive). That gives it some balance. Since teams generally have two goalies and there are an average of two top scorers per team (by definition), that also gives it some balance. Perhaps more importantly, these are players which are:
A) easy to measure their quality/value compared to other players, using %GP for goalies and finishes in top 2N in points for scorers.
B) most likely to leave for the NHL, since they are highly valuable, and therefore should receive near-maximum playing time and salary.
C) of the quality most frequently compared to one another
Defensive defensemen and grinders are not measured so well by these metrics, so it's more difficult to capture their impact. That's one reason I haven't done a more comprehensive measure, the other being it would take much more time. My intuition is that defensive defensemen and grinders may have had significantly less impact than this chart suggests, but this would likely be due to a combination of factors. Differences in coaching and styles of play may be one factor. Difficulty in scouting, especially in earlier years, is another. Also, as quality level decreases, it seems more likely that there may be substantial barriers. For instance, a player may not want to leave his home country to face many challenges (country, culture, language, minor leagues, etc.) when the benefits are uncertain (pay, length of career, playing time, etc.). Also, teams may rather choose a player who they've scouted, trained, coached and watched for a much longer time than a player from overseas, whose value is much less certain to them. IOW, the player may say "I am guaranteed a large role, with good pay, for a long time in my home country, where I am comfortable" and the team may say "we have been grooming a Canadian player for this role for some time and know he can fill the role properly." OTOH, they can't so easily replace a quality goaltender or a top tier scorer. The inclusion of U.S. players in the non-Canadian pool mitigates these factors to some degree, but since they are (very roughly) ~25% of the non-Canadian pool, it certainly doesn't eliminate these potential barriers.
I agree that this is a very important study for historical research.
But the sharp increase in the slope of the line that corresponds exactly with the 1967 expansion has not been adequately explained
I've tried to explain that it's basically the "baby boom" generation reaching the age which are most frequent for NHL players. I'm not claiming it's 100% accurate in magnitude, due to assumptions that had to be made and the limits of the data. However, the number of teams in the NHL has absolutely no bearing on the calculations for these charts. Look at this info, and tell me whether you think there was an increase in NHL-age population from the mid-60s to the early 90s:
The charts both basically confirmed what many of us largely knew already, and were also surprising to me:
Goalies- I knew non-Canadian goalies took a while to really become a big factor (outside of Hasek and a few others), but until I saw it graphically, I didn't realize just how large a presence they had become over the past decade or so. Here is the % of GP by goalies from various countries:
Year
Cze
Fin
Rus
Swe
USA
Other
Total
1975
0%
0%
0%
0%
3%
0%
3%
1976
0%
0%
0%
0%
3%
0%
3%
1977
0%
0%
0%
0%
3%
0%
3%
1978
0%
0%
0%
0%
4%
0%
4%
1979
0%
0%
0%
0%
1%
0%
1%
1980
1%
1%
0%
4%
2%
0%
8%
1981
3%
2%
0%
2%
3%
0%
9%
1982
0%
0%
0%
0%
1%
0%
1%
1983
0%
1%
0%
2%
0%
0%
4%
1984
0%
1%
0%
2%
4%
0%
7%
1985
0%
1%
0%
4%
11%
0%
15%
1986
0%
0%
0%
0%
11%
0%
11%
1987
0%
2%
0%
0%
9%
0%
11%
1988
0%
2%
0%
0%
9%
0%
11%
1989
0%
2%
0%
0%
11%
0%
13%
1990
0%
1%
0%
0%
6%
0%
7%
1991
0%
1%
0%
0%
8%
0%
9%
1992
1%
1%
1%
0%
8%
0%
12%
1993
1%
0%
2%
2%
10%
0%
15%
1994
2%
0%
4%
1%
10%
0%
17%
1995
3%
0%
6%
2%
0%
0%
12%
1996
2%
0%
5%
2%
13%
0%
22%
1997
3%
0%
6%
3%
13%
0%
25%
1998
4%
0%
7%
3%
10%
0%
25%
1999
5%
0%
9%
3%
7%
0%
23%
2000
5%
0%
5%
3%
9%
0%
23%
2001
10%
1%
7%
3%
6%
1%
27%
2002
10%
3%
8%
5%
6%
1%
34%
2003
9%
6%
7%
4%
2%
2%
30%
2004
7%
9%
5%
3%
4%
5%
34%
2006
6%
11%
5%
4%
7%
6%
38%
2007
7%
15%
5%
7%
8%
4%
46%
2008
6%
14%
8%
7%
8%
5%
48%
2009
6%
14%
7%
6%
4%
5%
41%
2010
7%
14%
7%
7%
11%
5%
50%
2011
11%
15%
8%
8%
12%
2%
55%
2012
9%
14%
9%
8%
12%
4%
56%
Top 2N Scorers- I had looked at similar metrics (top 5/10 finishes in goals/points, etc.), so this wasn't shocking. The recent drop seems due to the Czech/Slovak and Russian impact having peaked, as they really started dropping 3-4 years ago. The U.S. isn't as strong at the top as it was in the 90s, but the depth hasn't shown any significant drop, so that may more of an aberration.
Sorry I haven't been around to comment much, but I think you're doing a great job - you certainly don't have any holes in this study that are easy for me to poke through
I think that research into the size of the talent pool is one of the most important areas of historical research
The one point of criticism I can offer is that you aren't making a distinction between native Canadian population and first generation immigrants who didn't grow up emerged in hockey culture, correct? Much of Canada's growth since WW2 has come from immigration, and I don't think first generation immigrants should be considered as strong a hockey producing pool as Canadians who grew up immersed in hockey culture
Sorry I haven't been around to comment much, but I think you're doing a great job - you certainly don't have any holes in this study that are easy for me to poke through
I think that research into the size of the talent pool is one of the most important areas of historical research
The one point of criticism I can offer is that you aren't making a distinction between native Canadian population and first generation immigrants who didn't grow up emerged in hockey culture, correct? Much of Canada's growth since WW2 has come from immigration, and I don't think first generation immigrants should be considered as strong a hockey producing pool as Canadians who grew up immersed in hockey culture
Yes, this should be accounted for, although it would be difficult.
The one point of criticism I can offer is that you aren't making a distinction between native Canadian population and first generation immigrants who didn't grow up emerged in hockey culture, correct? Much of Canada's growth since WW2 has come from immigration, and I don't think first generation immigrants should be considered as strong a hockey producing pool as Canadians who grew up immersed in hockey culture
That's a valid criticism. There probably should have been some sort of adjustment made for net immigration (immigrants minus emigrants). It might be reasonable to completely exclude the net immigrant population from the hockey population pool.
The data is rather incomplete, which complicates things. It appears there is yearly data for immigrants, but emigrant data is only available by the decade or (more recently) in periods of five years. There also does not appear to be age data for immigrants.
I certainly think this type of study could be improved, and your suggestion is one way in which it could be.
One means of estimating talent pools would be to go back to look at the youth registration records in the various countries. While this will not necessarily speak to NHL strength I think it can speak to what the skill level necessary. If there are 100K children playing hockey for a given age year and let's say, 60-70 openings for that are group in the NHL that will say something as compared to other similar numbers. Let us just assume that talent generally rises to the top.
The harder comparison will eras where "freedom of movement" and "tendency to move" is limited. Sure, the Euros play in NA now but they didn't used to... even now the KHL has an effect.
This may ignore the level of athlete who gets involved... you may generally have a stronger Canadian athlete pool because hockey is the thing to do... not so in the U.S. In either case, getting some metrics based upon age-group youth registration through the years may say something.
One means of estimating talent pools would be to go back to look at the youth registration records in the various countries. While this will not necessarily speak to NHL strength I think it can speak to what the skill level necessary. If there are 100K children playing hockey for a given age year and let's say, 60-70 openings for that are group in the NHL that will say something as compared to other similar numbers. Let us just assume that talent generally rises to the top.
The harder comparison will eras where "freedom of movement" and "tendency to move" is limited. Sure, the Euros play in NA now but they didn't used to... even now the KHL has an effect.
This may ignore the level of athlete who gets involved... you may generally have a stronger Canadian athlete pool because hockey is the thing to do... not so in the U.S. In either case, getting some metrics based upon age-group youth registration through the years may say something.
Will it? I'm not sure.
The issue with this approach is that in the "old days," most youth didn't register - they just went out to play in the frozen ponds.
Yes, it's difficult to say. Perhaps it's a wash, but I tend to believe there has been some compression of talent over the last ~35 years. Here is my general view (educated guess) as to how the NHL talent per team has changed since WWII:
before/during WWII- a relatively low population with often > 6 teams, which is then further diluted by WWII
late 40s to mid-50s- some players return from war, population increases, and talent per team roughly matches previous levels
mid-50s to early 60s- increase in roster size offsets continued population growth, so talent per team is diluted then makes it way back to roughly previous levels
60s- population increases further, and in the last few years before expansion, the talent really becomes compressed
late 60s to mid-70s- population increases further still, but massive and repeated expansion, as well as the WHA, serve to substantially dilute the talent per team
late 70s to mid-80s- NHL stops expanding, contract by one team, then absorbs WHA, while population continues to increase, all of which negate much of the previous dilution since expansion
mid-late 80s- Europeans trickle in further, and US becomes a significant source of talent again, which combined with lack of expansion further compresses talent
90s to present- hockey-age population seems to peak, while there is large influx of Euros/Russians... much of this is offset by expansion
Perhaps the overall talent per team is similar to the mid-late 80s, but I still tend to believe that there is more talent per team at the top. There were 21 teams then and 30 teams now, so the new talent would have to be ~30% of the current NHL for it to be roughly the same. The overall representation of US/Euro players probably hovers around that 30% mark, while the increase in their representation at the top has done so as well. Of course it varies by year and position. For instance, in the mid-late 90s, the top tiers of scoring forwards saw an increase of more than 30 percentage points, while expansion was still being completed (so the benchmark would be < 30%). Meanwhile, goalie representation was much less than 30% during that time. Since the lockout, top forwards have dipped under the 30% mark, while goalies have easily surpassed that level.
Ok, that sounds reasonable. Any chance of a study being done on it? I'd seriously like to read through it if one gets done.
As an aside, in an appendage to his book "The Game", Ken Dryden posted numbers regarding how many kids are playing lower-level hockey (ie the talent pool the pros draw from) as of 2003:
Now Dryden doesn't provide a footnote for his source for the data, but taking it at face value, is there any way that it can be shown just how much the talent pool has grown in proportion to how much the NHL has expanded?
