HFBoards Point Allocation - Original Six Era
 Register FAQ Members List Search Today's Posts Mark Forums Read
 Mobile Hockey's Future Become a Sponsor Site Rules Support Forum vBookie Page 2
 Notices Please do not post or solicit links to illegal game streams.

 The History of Hockey Relive great moments in hockey history and discuss how the game has changed over time.

# Point Allocation - Original Six Era

 09-27-2012, 04:56 PM #1 Hockey Outsider Registered User     Join Date: Jan 2005 Country: Posts: 3,783 vCash: 500 Point Allocation - Original Six Era Here is my attempt to apply Iain Fyffe's "Point Allocation" system to the NHL from 1953-1967. The system allocates the points a team earns in the standings to each player on the team, based on their on-ice contributions. The paper explains the methodology and shows the results. Comments (and suggestions to improve) are welcome. http://uploadfiles.ca/201PointAllocationOriginalSix.pdf (Also - if anyone has a suggestion for a more efficient way to store the file, let me know).
09-27-2012, 05:12 PM
#2
Registered User

Join Date: Aug 2006
Location: Brooklyn
Country:
Posts: 45,407
vCash: 500
Well done

Quote:
 We must calculate the “Goals Created” (or “GC”) for each player on each team9. GC equals one-half of the player’s goals, plus one-half of the player’s assists divided by the ratio of assists to goals on that team. In practical terms, the sum of Goals Created will equal the total goals scored for each team. As an example, Gordie Howe scored 49 goals and 46 assists in 1953. That season, there were 1.45 assists per goal in Detroit. Thus, Mr. Hockey gets credit for (49 / 2) + (46 / 2 / 1.45) = 40.4 goals created. In total, Red Wings scored 222 goals10, thus Howe earns 18.2% (40.4 / 222) of the team’s 43.5 points from Goal Creation. This equals 7.9 (0.182 * 43.5) Points Allocated to Offense (PAO). This is, according to Point Allocation, the greatest offensive season during the Original Six era11. 8
Fantastic method for determining "goals created"

I've never fully bought the unofficial shot totals and save percentages from the O6 period, so that's a weakness for me, when you are determining defensive value.

That fact that the top 2 defensemen based on the formula (Kelly and Pronovost) played on the same team for so long shows the limitation of the defensive point allocation that you acknowledge.

09-27-2012, 05:49 PM
#3
Registered User

Join Date: Aug 2006
Location: Brooklyn
Country:
Posts: 45,407
vCash: 500
HO, can you clarify this statement?

Quote:
 For the remainder of his career (1956-67), Sawchuk ranks last among goalies in Point Allocation, both in total and on a per-game basis.
What is the sample of goalies you are using here that Sawchuk ranks last in? Is it guys who started a minimum number of games? If so, what is the minimum? Thanks.

 09-27-2012, 05:53 PM #4 Iain Fyffe Hockey fact-checker     Join Date: Feb 2009 Location: Fredericton, NB Country: Posts: 3,469 vCash: 500 Solid work. Although the method for calculating the marginal rate for goals seems simplistic, it comes close to the appropriate number in my mind. I use a 44/56 split between offence and defence rather than 42/58. This is based on historical results. I'm sure I'll get around to publishing something on that at some point... I don't think there's any reason to be skeptical of the O6 save percentage numbers. As I understand them they were compiled from official NHL game reports. The league was recording shots on game sheets, they just weren't compiling the numbers at the time. That's what I recall, anyway, can't guarantee the accuracy of it. I've never been sure about assuming that goals and assists should be considered of equal importance in offence, which is what you're doing here if I understand it. I'm of the opinion that points probably fairly represent offensive contribution. Years in which there are more assists per goal might well reflect a greater importance on passing plays in scoring goals. When possible, I like to assume that hockey people at the time knew what they were doing, rather than applying an arbitrary adjustment. I doubt it makes a lot of difference in the end though.
09-27-2012, 06:15 PM
#5
Registered User

Join Date: Aug 2006
Location: Brooklyn
Country:
Posts: 45,407
vCash: 500
Quote:
 Originally Posted by Iain Fyffe I don't think there's any reason to be skeptical of the O6 save percentage numbers. As I understand them they were compiled from official NHL game reports. The league was recording shots on game sheets, they just weren't compiling the numbers at the time. That's what I recall, anyway, can't guarantee the accuracy of it. i.
In modern times, there is a fairly significant arena effect when it comes to shot recording, despite the fact that save % is an official stat, and one that GMs are increasingly focusing on when trying to build winning teams and awarding contracts.

In the O6 era, save perentage wasn't even an officially recognized stat by the NHL, and there was a much bigger emphasis on GAA. So recording shots accurately wouldn't have been a priority. And without the national telecasts there would have been less ability to standardize the definition of "a shot." So, if anything, I would expect the definition of a "shot" to vary even more during the Original 6 than it does today.

This is all theoretical. A rough test would involve looking at the team-by-team shots-allowed totals of the era and comparing them to the defensive reputations of the teams. If Johnny Bower's Leafs are allowing more shots than most teams, I would be highly skeptical of the data. Is all the data available on the yahoo hockey group?

 09-27-2012, 06:46 PM #6 Iain Fyffe Hockey fact-checker     Join Date: Feb 2009 Location: Fredericton, NB Country: Posts: 3,469 vCash: 500 To be more specific, I don't think there's much reason to be more skeptical of those numbers than modern numbers.
09-27-2012, 08:12 PM
#7
Hockey Outsider
Registered User

Join Date: Jan 2005
Country:
Posts: 3,783
vCash: 500
Quote:
 Originally Posted by TheDevilMadeMe Fantastic method for determining "goals created"
Thanks, but that was all Iain's work.

Quote:
 Originally Posted by TheDevilMadeMe HO, can you clarify this statement? What is the sample of goalies you are using here that Sawchuk ranks last in? Is it guys who started a minimum number of games? If so, what is the minimum? Thanks.
He ranked fifth out of the five goalies who appeared in at least 300 games (Hall, Plante, Sawchuk, Worsley, Bower). I don't mean to slam Sawchuk too much, but it's worth emphasizing that unlike Plante and Hall, he had a short, brilliant peak, followed by a longer period of fairly mediocre play.

Quote:
 Originally Posted by TheDevilMadeMe This is all theoretical. A rough test would involve looking at the team-by-team shots-allowed totals of the era and comparing them to the defensive reputations of the teams. If Johnny Bower's Leafs are allowing more shots than most teams, I would be highly skeptical of the data. Is all the data available on the yahoo hockey group?
I agree with Iain - the shot data is official (it's not like someone went back and watched all the games and recorded shots and saves, it was already published in newspapers). The NHL never bothered compiling the daily summaries into season and career totals - but there's no reason researchers can't do that today.

Take a look at "Complete Goalie Stats from 52-53 to 82-83.xls" posted by Hockeydad55 in the Hockey Analysis Group - all of the data I used is from that file.

Last edited by Hockey Outsider: 09-27-2012 at 08:22 PM.

09-27-2012, 08:20 PM
#8
Hockey Outsider
Registered User

Join Date: Jan 2005
Country:
Posts: 3,783
vCash: 500
Quote:
 Originally Posted by Iain Fyffe Solid work. Although the method for calculating the marginal rate for goals seems simplistic, it comes close to the appropriate number in my mind. I use a 44/56 split between offence and defence rather than 42/58. This is based on historical results. I'm sure I'll get around to publishing something on that at some point...
I'm curious to see what you found.

Quote:
 Originally Posted by Iain Fyffe I've never been sure about assuming that goals and assists should be considered of equal importance in offence, which is what you're doing here if I understand it. I'm of the opinion that points probably fairly represent offensive contribution. Years in which there are more assists per goal might well reflect a greater importance on passing plays in scoring goals. When possible, I like to assume that hockey people at the time knew what they were doing, rather than applying an arbitrary adjustment. I doubt it makes a lot of difference in the end though.
Correct - I'm trying to ensure that half of the offensive credit goes to goals (thus each goal is worth 0.5) and half of the offensive credit goes to assists (since there are up to two assists per goal, each assist is worth around 0.30). I like the symmetry in assuming that shooting and passing are equally valuable skills. Ultimately, I think that this has a fairly small impact (though obviously my approach would slightly favour goal-scorers like Hull, M. Richard, Geoffrion, etc).

09-27-2012, 09:51 PM
#9
Iain Fyffe
Hockey fact-checker

Join Date: Feb 2009
Location: Fredericton, NB
Country:
Posts: 3,469
vCash: 500
Quote:
 Originally Posted by Hockey Outsider Thanks, but that was all Iain's work.
I thought that it was. I guess I've changed my mind in the decade or so since I first did it.

09-27-2012, 11:25 PM
#10
tarheelhockey
Global Moderator

Join Date: Feb 2010
Location: The Triangle
Country:
Posts: 44,574
vCash: 1020
Quote:
 Originally Posted by Hockey Outsider I agree with Iain - the shot data is official (it's not like someone went back and watched all the games and recorded shots and saves, it was already published in newspapers).
On the other hand, that would be the obvious way to spot-check the stats... find a complete game tape, and see if the recorded shot total is at least reasonably close to what the tape shows. Shouldn't take more than a few hours to know whether the old numbers are garbage or gold.

 09-28-2012, 09:17 AM #11 pappyline Registered User     Join Date: Jul 2005 Location: Mass/formerly Ont Country: Posts: 4,265 vCash: 500 Interesting study but it seems to me it favors forwards on the better teams. Looking at 59-60, I find it hard to believe that Henri Richard is anywhere close to being the best forward that year. Ahead of Hull & Howe? Is he benefiting from playing on a packed team that ran away with first place that year?
09-28-2012, 09:46 AM
#12
Iain Fyffe
Hockey fact-checker

Join Date: Feb 2009
Location: Fredericton, NB
Country:
Posts: 3,469
vCash: 500
Quote:
 Originally Posted by pappyline Interesting study but it seems to me it favors forwards on the better teams. Looking at 59-60, I find it hard to believe that Henri Richard is anywhere close to being the best forward that year. Ahead of Hull & Howe? Is he benefiting from playing on a packed team that ran away with first place that year?
Yes, players on stacked teams will tend to benefit this way. This is because the system allocated the points a team earned to its players. But the fact is, the effect of having numerous good players is not additive, it's multiplicative. So players receive a benefit from having more good players on the team.

At times I've toyed with the idea of including a "synergy" element, essentially regressing teams' point totals to the mean somewhat, in order to account for this fact.

 09-28-2012, 04:01 PM #13 seventieslord Moderator     Join Date: Mar 2006 Location: Regina, SK Country: Posts: 28,309 vCash: 500 How do you know it is multiplicative and not additive? Just curious.
09-28-2012, 07:04 PM
#14
Iain Fyffe
Hockey fact-checker

Join Date: Feb 2009
Location: Fredericton, NB
Country:
Posts: 3,469
vCash: 500
Quote:
 Originally Posted by seventieslord How do you know it is multiplicative and not additive? Just curious.
To be honest I couldn't give you anything hard on that. However it's a theory that nicely explains why players on good teams get better results via point allocation than they would on bad teams.

 09-30-2012, 12:41 AM #15 seventieslord Moderator     Join Date: Mar 2006 Location: Regina, SK Country: Posts: 28,309 vCash: 500 Here's something I noticed: In 15 years, there are 75 instances of a player showing up in the top-5 for a season, and just five of those times, it's a defenseman. That just seems to be very proportionally wrong. This isn't the first point allocation system I've seen that has this limitation, obviously. It seems like maybe not enough defensive credit is given to the defensemen.
02-24-2016, 10:01 PM
#16
Hockey Outsider
Registered User

Join Date: Jan 2005
Country:
Posts: 3,783
vCash: 500
Quote:
 Originally Posted by seventieslord Here's something I noticed: In 15 years, there are 75 instances of a player showing up in the top-5 for a season, and just five of those times, it's a defenseman. That just seems to be very proportionally wrong. This isn't the first point allocation system I've seen that has this limitation, obviously. It seems like maybe not enough defensive credit is given to the defensemen.
I know that I'm bumping a thread from over three years ago, but I never responded to this valid question.

Since there's no ice time data, and no goals against at the individual level, there really isn't a good way to allocate defensive credit to skaters. I've allocated it on the basis of games played. To quote myself from page 4:

Ideally, I would have liked to allocate teams’ Points Allocated to Defense (PAD) in proportion to each player’s ice time. Since that isn’t possible, I’ve allocated PAD to each player in proportion to the number of games played, with a positional adjustment (forward games played are unadjusted and defensemen games played are double-weighted). This means that all teammates, who play the same number of games at the same position, earn the same PAD. Obviously, this isn’t realistic—for example, on the 1953 Detroit Red Wings, defensemen Red Kelly and Benny Woit both earn 3.9 PAD, since both played 70 games. If ice time was tracked, I’m sure it would show that Kelly played far more per game than Woit (and therefore contributed more to the team’s defense). However, since this data is unavailable, we’ll have to accept this as a limitation of the methodology.

Thus, the defensive contributions of every player is flattened. Exceptionally good (or bad) performances will get glossed over. This tends to hurt defensemen more than forwards. I think the system works quite well for measuring the performance of forwards, and reasonably well for goalies, but it distorts the ranking of defensemen relative to their peers (since defense isn't measured well) and underrates them relative to the other positions (since the other positions are getting full credit for their primary area of responsibility, while defensemen aren't).

(One other point - I did this project for fun. I think there's some value in it but am not suggesting this is the "one true method" for evaluating players from that era).

Forum Jump