There are plenty of cases where a particular player has played only a handful of games but has been on ice for a relatively large amount of goals that would suggest sometimes exceedingly high situational times. Usually, but obviously not always this leads to estimates that are too high. The old formula adjusts results down for players who haven't played the entire schedule (more for those who've played less games). Unfortunately it doesn't distinguish elite players from call-ups and star players tend to get underestimated. Since there are more instances of minor league call-ups this seemed like a good idea at the time.

In the current formula there are adjustment factors for ES, PP and PK, to bring outlier results closer to "expected" percentage of the total team situational time a player has spent on ice. i.e. If a player has several seasons of 30% of ESTOI and one season of 25% in the middle with his production and spot on the depth chard remaining the same the real value is likely to be somewhere in between and upping it a notch might make it 27-28%. I haven't done any studies on this, it's entirely based on looking at the numbers. A lot.
A preferable situation would of course be where this was done by the formula itself. The results for the current players are very good but there's no way to tell what is the case with manual adjustments pre 98.

Thanks for the details. Sounds like the old question "what population mean to regress to" is coming up again. Regressing to overall league average should work best for most players, but will underestimate stars.

For a similar example, an adjusted plus-minus system that assumes that all skaters deserve 20% of the credit for plusses and minuses will work well for most players - but not Bobby Orr or Wayne Gretzky.

Have you made adjustments based on the scoring level of players? For example, let's say a team has 4 forwards.

If you simply looked at the total number of goals, and you had 40 minutes to allocate among these 4 players, you would estimate Player A played 15 minutes, Player B and Player C played 10 minutes, and Player D played 5 minutes. Is that basically how your estimates work - calculating the minutes available and then dividing them up by total on-ice goals, with some regression to a mean? Or do the number of points they scored enter into the estimate at all, based on the assumption that higher scoring players tend to be on the ice for more total goals overall?

If you simply looked at the total number of goals, and you had 40 minutes to allocate among these 4 players, you would estimate Player A played 15 minutes, Player B and Player C played 10 minutes, and Player D played 5 minutes. Is that basically how your estimates work - calculating the minutes available and then dividing them up by total on-ice goals, with some regression to a mean? Or do the number of points they scored enter into the estimate at all, based on the assumption that higher scoring players tend to be on the ice for more total goals overall?

Yes, for ES it's basically the average of %TGF and %TGA with an additional factor that regresses the numbers towards desired results. For temp players this is lower than for regulars and for goons who don't play regular shift even lower.

I've been looking at GF- and GA-heavy players but haven't found anything useful so far. This may be a problem especially when GF or GA rates are vastly different from the rest of the team e.g. Bobby Clarke for a couple of years when his GA was miniscule but GF was very high. It's possible that these cancel each other out and it's likely that the error isn't huge (as in several minutes) but I don't know if it's consistent across the board and could be taken into account or something that varies case by case.

The thing about these estimates that hasn't been mentioned yet, is that (according to Iain Fyffe), the calculated results were compared to actual results for years in which the results were known, and were found to have a correllation of 0.96.

No one has ever said that they are perfectly accurate. They pass the smell test for sure, that is, they match what our perceptions are of these post-expansion players, and I haven't really seen a good example where they don't.

So there is no independent verification of the study and no one has been able to replicate the data.

Bolded above. Correlation that was produced was to be expected since the known data may have been the inspiration and source for the study. Since we do not know the details of the study ........
Assuming the same correlation between the estimates does not work due to the changes in coincidental penalty rules, roster sizes, shift length and other factors that impact TOI.

See posts 49 and 50. All we have are teaser snippets that do not balance. Factor in Stanfield's time playing the point and you have further drift.

Let some one produce the estimates for a complete team - say the Bobby orr Bruins in question or the complete NHL for a specific season 1967-68 thru 1971-72 and we'll have a look.

A situation that does lead to the numbers not balancing and one that's a little difficult to account for is when one or more players switch between forward and defense during a season (even during individual games) and the exact split is not known. Yet, the players are identified as either D or F.

This doesn't mean that the estimates are necessarily completely out of whack although D and F are handled a little differently. You'll end up with composite numbers for just like with the official TOI. That the numbers don't balance is mostly because a player who's switched positions has TOI on both D and F but is listed as one or the other and that isn't apparent to someone looking to add ice time totals.

A compromise solution would be possible where it's assumed that shifts are split 50/50 and the player is identified as a "switcher" but I don't know which players and which seasons would qualify. The improvements would likely be small except possibly in situations where someone plays regular shift on D but is a specialist/goon on F or vice versa.

The thing about these estimates that hasn't been mentioned yet, is that (according to Iain Fyffe), the calculated results were compared to actual results for years in which the results were known, and were found to have a correllation of 0.96.

This means nothing.

If the calculations used to conduct the estimate are calibrated to match the official data available (which only makes sense), how could you expect anything else but a high correlation?

If players were always used in the same fashion then fine we're reasonably sure.. but we all know they weren't used the same from 67-68 - 98-99 when official tracking began.

Quote:

Originally Posted by ssh

There are plenty of cases where a particular player has played only a handful of games but has been on ice for a relatively large amount of goals that would suggest sometimes exceedingly high situational times. Usually, but obviously not always this leads to estimates that are too high. The old formula adjusts results down for players who haven't played the entire schedule (more for those who've played less games). Unfortunately it doesn't distinguish elite players from call-ups and star players tend to get underestimated. Since there are more instances of minor league call-ups this seemed like a good idea at the time.

Does the average ice time sheet currently take into account individual players games played or does it work strictly from the TGF TGA etc.?

Quote:

In the current formula there are adjustment factors for ES, PP and PK, to bring outlier results closer to "expected" percentage of the total team situational time a player has spent on ice. i.e. If a player has several seasons of 30% of ESTOI and one season of 25% in the middle with his production and spot on the depth chard remaining the same the real value is likely to be somewhere in between and upping it a notch might make it 27-28%. I haven't done any studies on this, it's entirely based on looking at the numbers. A lot.

A preferable situation would of course be where this was done by the formula itself. The results for the current players are very good but there's no way to tell what is the case with manual adjustments pre 98.

Thanks for all the information ssh.

Not that I am doubting how much you have looked at the numbers but what is the "expected" percentage?

How much subjective massaging was done? I definitely agree that it would be better if the work was done by the formula itself.

It just may not be possible given the slim pickings in the data up until more recently..

Quote:

A few things come to mind right now.

Total team situational ice times obviously have an effect as player TOI is calculated as a percentage of that. The model I use currently has a stdev for ES of about 0.8% and about 2.6% for PP and PK for seasons 99-2011. 97-98 is problematic. The model gives consistently higher special teams results, mostly around 5% with several above 10%. The worst case is the 98 Avs, their PP estimate is 22% higher than what's expected based on the sum of the player's PPTOI. Either there was something odd going on with a few teams at the time or there's some really bad data. I really don't know if this is something that should be taken into account for the seasons before that.

Team performance may change a lot during the season. This may be due to injuries, trades, changes in goaltending or just "normal" ups and downs in player and team performance. For example someone who played with the 93-94 Sabres in front of Fuhr in late 93 would have drastically different GA numbers than someone who got to play with Hasek in the following spring.

Players who spent a long time in a stable environment (same team without a lot of turnover) are easier to evaluate than a very fragmented career. It's not easy to tell if a borderline player plays a handful of games on a good team and gets first line/pairing ice time.

The relationship between actual percentage spent in a particular situation and the rough estimate based on TGF and TGA data is usually fairly consistent. For example top offensive forwards tend to play a little less on ES and PP, defensive players a little more on ES and PK. ES and PP seem to be more predictable, there is fluctuation in PKTOI that I don't know how to estimate.

There's probably more. I'll comment if something comes up.

So there is no independent verification of the study and no one has been able to replicate the data.

Bolded above. Correlation that was produced was to be expected since the known data may have been the inspiration and source for the study. Since we do not know the details of the study ........
Assuming the same correlation between the estimates does not work due to the changes in coincidental penalty rules, roster sizes, shift length and other factors that impact TOI.

See posts 49 and 50. All we have are teaser snippets that do not balance. Factor in Stanfield's time playing the point and you have further drift.

Let some one produce the estimates for a complete team - say the Bobby orr Bruins in question or the complete NHL for a specific season 1967-68 thru 1971-72 and we'll have a look.

Not sure what you mean since in post 50, Iain thoroughly debunked what you said in post 49. When I do the math correctly (total TOI times total GP) for each defenseman and sum it up, I get 119. Which makes sense considering Stanfield's 2 minutes a game on the PP.

Quote:

Originally Posted by BraveCanadian

This means nothing.

If the calculations used to conduct the estimate are calibrated to match the official data available (which only makes sense), how could you expect anything else but a high correlation?

If players were always used in the same fashion then fine we're reasonably sure.. but we all know they weren't used the same from 67-68 - 98-99 when official tracking began.

You're nitpicking. How different do you think it really is? Teams play their best players the most. They play their grunts and thugs the least. It's always been that way.

Teams do use their players "somewhat" differently, but this should not change the fact that more GF and GA are a symptom of being on the ice more often.

Yes or no, if you're on the ice more often you will have more GF and GA?

Quote:

Does the average ice time sheet currently take into account individual players games played or does it work strictly from the TGF TGA etc.?

No, the sheet doesn't know specifically which games the players played, hence the Fuhr/Hasek example ssh cited.

You're nitpicking. How different do you think it really is? Teams play their best players the most. They play their grunts and thugs the least. It's always been that way.

Shift lengths have changed, TV timeouts have been introduced, the number of defensemen carried on a typical game roster has changed yadda yadda..

Quote:

Yes or no, if you're on the ice more often you will have more GF and GA?

In general, sure. However the players we are most often interested in are not the average players.

We're generally interested in the more outstanding players and those are the ones that score and get scored upon at completely different rates than even the "average" player of their position on the depth chart.

To use an example that ssh brought up: does that therefore mean that Bobby Clarke didn't play much when he had so few GA?

Quote:

No, the sheet doesn't know specifically which games the players played, hence the Fuhr/Hasek example ssh cited.

I'm more interested in a player who normally chews up half the game but misses 5 or 10 games on the schedule and goes back to being a 25 minute player on average because they had 0 minutes in those games.

To use an example that ssh brought up: does that therefore mean that Bobby Clarke didn't play much when he had so few GA?

just like in the Orr thread when you forgot that GA was also a component, you're now forgetting GF was also a component. Clarke's GF was very high, which made his total ESG (F+A) still the highest on the team, indicating what we know to be true - that he got the most icetime among all forwards on the team.

Quote:

I'm more interested in a player who normally chews up half the game but misses 5 or 10 games on the schedule and goes back to being a 25 minute player on average because they had 0 minutes in those games.

no, of course not, it is all done per game. it is looking at his total GF+GA, divided by game.

Last edited by seventieslord: 07-17-2012 at 03:38 PM.

just like in the Orr thread when you forgot that GA was also a component, you're now forgetting GF was also a component. Clarke's GF was very high, which made his total ESG (F+A) still the highest on the team, indicating what we know to be true - that he got the most icetime among all forwards on the team.

I didn't forget. I was simply pointing out how your statement doesn't get a simple yes or no response.

That is why there is so much manual manipulation in these numbers and guesswork.

Quote:

no, of course not, it is all done per game. it is looking at his total GF+GA, divided by game.

I would hope so but that isn't what you indicated.

Not that you know, because you don't have the calculations.

Shift lengths have changed, TV timeouts have been introduced, the number of defensemen carried on a typical game roster has changed yadda yadda..

In general, sure. However the players we are most often interested in are not the average players.

We're generally interested in the more outstanding players and those are the ones that score and get scored upon at completely different rates than even the "average" player of their position on the depth chart

IIRC the estimated numbers showed an increase in the 1992-93 minutes played for star players, so it is picking up on the TV timeout change from that season.

I agree with your second point about outlier players, as well as your earlier post about the need to use out-of-sample data to test the accuracy, not the same data that was used to develop it.

But I suspect that any changes to the methodology are going to move things by a couple of minutes at the most. The underlying GF/GA data is pretty good.

IIRC the estimated numbers showed an increase in the 1992-93 minutes played for star players, so it is picking up on the TV timeout change from that season.

OT, but could this be the primary reason that stars began scorin a higher percentage of their teams' offenses? (and thus became overrated by traditional adjusted stats?)

IIRC the estimated numbers showed an increase in the 1992-93 minutes played for star players, so it is picking up on the TV timeout change from that season.

Good stuff. I don't actually have the sheet atm because I toasted my computer a while back but that is good.

Quote:

I agree with your second point about outlier players, as well as your earlier post about the need to use out-of-sample data to test the accuracy, not the same data that was used to develop it.

Right. This is a no brainer.. unfortunately other than more recent seasons we don't have any data to check the current calculation methods against. (Not to mention we don't have the method anyways)

So that isn't going to do us much good for our accuracy before 99.

Quote:

But I suspect that any changes to the methodology are going to move things by a couple of minutes at the most. The underlying GF/GA data is pretty good.

Agreed, I don't expect we're going to be seeing players with really big swings..

My main concern has been that for a while that these numbers were being relied upon without anyone even knowing where they came from. Even basing other work in part on the ice time numbers.

For anyone that is trying to take any of this work seriously that is a pretty big problem.

Last edited by BraveCanadian: 07-17-2012 at 03:20 PM.

OT, but could this be the primary reason that stars began scorin a higher percentage of their teams' offenses? (and thus became overrated by traditional adjusted stats?)

If the calculations used to conduct the estimate are calibrated to match the official data available (which only makes sense), how could you expect anything else but a high correlation?

If players were always used in the same fashion then fine we're reasonably sure.. but we all know they weren't used the same from 67-68 - 98-99 when official tracking began.

The effect player usage has probably isn't as big as some people think aside from the fact that in the 70's it wasn't uncommon that some players were on ice for practically the whole time their team spent on PP or PK.

IMO a much bigger issue is that in the past there could be vast differences in defensive ability between players on the same team. Poor defensive players would have relatively high GA/game and TOI would be overestimated. Then, particularly in the early and mid 80's forwards scored with sometimes silly efficiency. Plugger/goon lines could have a total of say 2 shots on goal per game but score on 20-25% of their shots. If the scoring lines aren't having a strong year the goon line may score a disproportionate amount of goals relative to their ice time.

Quote:

Does the average ice time sheet currently take into account individual players games played or does it work strictly from the TGF TGA etc.?

Both the old and the new model are built around TGF/TGA rates but regress to mean. The regression factor is weighted more for players who've played fewer games.

Quote:

Not that I am doubting how much you have looked at the numbers but what is the "expected" percentage?

How much subjective massaging was done? I definitely agree that it would be better if the work was done by the formula itself.

It just may not be possible given the slim pickings in the data up until more recently..

In the old version there's less massaging, and only on ES.
In the new one, less would be preferable. Although it's rather systematic and I'd rather not make adjustments that aren't based on the data even if the results would be more accurate there's bound to be some personal bias towards the official numbers since I'm fairly familiar with them.

As for the expected percentages on ES and with forwards tend go from around 15% for goons, 25-30% for 2nd and 3rd liners and high 20's to mid 30's for top line talent. The variation is surprisingly small. The top defensemen are around 40%
On PP it depends more on what kind of talent is available. The top units tend to play 55-65%, the rest goes to the 2nd unit and usually a fair number of bottom 6 players and talented call-ups. Players with overwhelming talent, F or D can go higher.
PK is much more difficult.

The type of massaging I've tested lately is generally of the following kind: keeping average situational times in line, keeping average situational percentages in line, adjusting top ES and PP times so that the resulting percentages remain a few percentage points below the raw TGF/TGA averages and throughout the line-up generally towards to numbers mentioned above. The adjustments are typically about a minute up or down, more in extreme cases.

EDIT: It takes way too long to compose these messages.

I didn't forget. I was simply pointing out how your statement doesn't get a simple yes or no response.

That is why there is so much manual manipulation in these numbers and guesswork.

You may or may not be right, but the Clarke example does nothing to further that. He had the most GF+GA, he had the most icetime. where is the manual manipulation and guesswork?

Quote:

I would hope so but that isn't what you indicated.

Not that you know, because you don't have the calculations.

I don't have the exact calculations, but the math behind them is very easy. Prior to the "x-th line adjustment" it is as simple as determining what percentage of the team's total situational time they played (based on situational GF/GA), determining what percentage of each game the team played in each situation (based on team PPOF and PPOA), and determining total minutes played. Obviously that final answer would then be divided by GP to get a "per game" result. i don't need to "have the calculations" to know that the creators of the formula would have remembered to include the most elementary of all the calculations. You don't think they'd have seen Mario Lemieux at 8 minutes a game in the 1994 season and figured something was up?

You may or may not be right, but the Clarke example does nothing to further that. He had the most GF+GA, he had the most icetime. where is the manual manipulation and guesswork?

The fact that we reasonably think he had the most icetime isn't the issue. It is how much and how did we arrive at that how much.

The manual manipulation and guesswork is inserting factors that "look right" to make the numbers jive with what is "expected".

It might all be reasonable and fairly accurate, but it is subjective.

Surely even you can see this..

Quote:

I don't have the exact calculations, but the math behind them is very easy. Prior to the "x-th line adjustment" it is as simple as determining what percentage of the team's total situational time they played (based on situational GF/GA), determining what percentage of each game the team played in each situation (based on team PPOF and PPOA), and determining total minutes played. Obviously that final answer would then be divided by GP to get a "per game" result. i don't need to "have the calculations" to know that the creators of the formula would have remembered to include the most elementary of all the calculations.

Do you ever get tired of how smart and infallible you are?

Like I said, you don't have the actual background info... just assumptions. Apparently it isn't quite so elementary:

Quote:

Originally Posted by ssh

Both the old and the new model are built around TGF/TGA rates but regress to mean. The regression factor is weighted more for players who've played fewer games.

Quote:

Originally Posted by seventieslord

You don't think they'd have seen Mario Lemieux at 8 minutes a game in the 1994 season and figured something was up?

Do you ever get tired of how smart and infallible you are?

Like I said, you don't have the actual background info... just assumptions. Apparently it isn't quite so elementary:

Nice one. Sigh..

Like I said, aside from the "massaging" as you call it, the underlying assumptions made to get to the end results are all very reasonable and logical. The regression applied to players with less GP is "massaging" above and beyond what I described to you. The only reason this "massaging" would be done, would be to create results that are more realistic.

If Lonny Bohonos plays 10 games as a call up and gets GF/GA figures that make it appear that he's a 20 minute a game player for those 10 games, applying regression to that number will bring it more in line. This is a good thing.

As it applies to star players like Mario, I agree it could skew him, but the creators saw it as the lesser of two evils due to there being far fewer instances of Mario than Bohonos.

It should be noted that as it applies to a player like Mario, in seasons where he had fewer games, he probably did have fewer minutes per game as well, as a result of occasionally "playing hurt" and being "worked back into the lineup" following an injury. I'm not saying drastically fewer, but perhaps 2-3 minutes less, which the sheet appears to capture.

Not sure what you mean since in post 50, Iain thoroughly debunked what you said in post 49. When I do the math correctly (total TOI times total GP) for each defenseman and sum it up, I get 119. Which makes sense considering Stanfield's 2 minutes a game on the PP.

The provided data as submitted by Iain Fyffe:

Originally Posted by Iain Fyffe View Post
In 1968/69, my estimates have the Bruins top pair averaging 28 minutes each, the second pair 24 minutes each, and the fifth defenceman playing about 18.5 minutes. They used almost exactly 5.0 defencemen per game.

The system uses team stats, not league stats, to build the estimates, so as to consider the context in which each individual player played.
-------------------------------------------------------------------------------
You can generate 119 if you fudge the numbers and make certain assumptions that are not supported.

Iain presented the data on a pairing basis plus a 5th dman as opposed to per player. So when Orr and Green were missing games how were the pairings affected? And who else played defense

The Bruins dmen are listed as combining for 371 games out of a possible 380 games. So if you take the % which is 97.63 you get 119.6 minutes But Iain is careful because he clearly states they used almost exactly 5 dmen per game, providing wiggle room below and aboce 5 dmen per game.

Now if you use the same data for RWs you get 291 games.Bruins were a three line team that season, yielding 225 games. Note Ed Westfall is listed clearly as a RW yet he came into the NHL out of junior/minors as a dman. In the games where the Bruins dressed 4 dmen or had filler types, Ed Westfall played some as a dman.

So the 122.5 minutes per game is more reflective of the team situation and that is before Fred Stanfield"s role is factored in.

Like I said, aside from the "massaging" as you call it, the underlying assumptions made to get to the end results are all very reasonable and logical. The regression applied to players with less GP is "massaging" above and beyond what I described to you. The only reason this "massaging" would be done, would be to create results that are more realistic.

If Lonny Bohonos plays 10 games as a call up and gets GF/GA figures that make it appear that he's a 20 minute a game player for those 10 games, applying regression to that number will bring it more in line. This is a good thing.

As it applies to star players like Mario, I agree it could skew him, but the creators saw it as the lesser of two evils due to there being far fewer instances of Mario than Bohonos.

It should be noted that as it applies to a player like Mario, in seasons where he had fewer games, he probably did have fewer minutes per game as well, as a result of occasionally "playing hurt" and being "worked back into the lineup" following an injury. I'm not saying drastically fewer, but perhaps 2-3 minutes less, which the sheet appears to capture.

Underlying assumptions is another way of saying incomplete research was done. Fact of the matter is that in the data under discussion the presenter was caught omitting the Fred Stanfield and Ed Westfall effect on the data. Properly presented with appropriate notes or explanations would have given needed credibility to the data.

As is, you are left trying to spin doctor the data and you get caught with the Westfall shortage.

First of all, Ed Westfall was only a defenseman in his first three NHL seasons, which were all pre-expansion, and hockey-reference acknowledges this. He doesn't affect the post-expansion TOI calculations in any way.

The rest of it, it's really difficult to make sense of it. If you have five defensemen and you have two of them on the ice at all times, then they are playing an average of 24 minutes per game. that is exactly what I calculate when I take their total GP (371) divided by their minutes (9076) - it balances.

The Stanfield factor does throw it off a bit, unless I am interpreting it incorrectly. because playing the point on the PP, he's techically a "D" for those 2.3 minutes every game, so instead of 120 D minutes to go around, there are 117.7. Leaving Orr's minutes untouched (because we know he played the whole PP) that leaves 2.3 minutes to come from the likes of Green, Smith, Awrey, Doak and Gauthier.

The Stanfield case isn't exactly common, it should be pointed out. But one example and we have people saying "the system is broken! too many assumptions! throw it out!" as though it is suddenly not useful for any team in any season.

The Bruins dmen are listed as combining for 371 games out of a possible 380 games. So if you take the % which is 97.63 you get 119.6 minutes But Iain is careful because he clearly states they used almost exactly 5 dmen per game, providing wiggle room below and aboce 5 dmen per game.

Now if you use the same data for RWs you get 291 games.Bruins were a three line team that season, yielding 225 games. Note Ed Westfall is listed clearly as a RW yet he came into the NHL out of junior/minors as a dman. In the games where the Bruins dressed 4 dmen or had filler types, Ed Westfall played some as a dman.

So the 122.5 minutes per game is more reflective of the team situation and that is before Fred Stanfield"s role is factored in.

Westfall's listed position is largely irrelevant. The time he might have spent on defense is factored in his total ice time and the per game average is somewhere between how much he played on F and D. The "missing" ice time isn't missing at all, it just isn't listed separate from his stats when playing forward.

Stanfields specific role in Bruins' power play is also quite irrelevant. Whether it was him or some other forward who played point doesn't matter. All that matters is that when a team that uses 4 (or 5) forwards on the power play everyone who's on the ice gets credit for the goal scored and everyone who isn't doesn't. If the situational ice times are calculated separately there shouldn't be any issue.

First of all, Ed Westfall was only a defenseman in his first three NHL seasons, which were all pre-expansion, and hockey-reference acknowledges this. He doesn't affect the post-expansion TOI calculations in any way.

The rest of it, it's really difficult to make sense of it. If you have five defensemen and you have two of them on the ice at all times, then they are playing an average of 24 minutes per game. that is exactly what I calculate when I take their total GP (371) divided by their minutes (9076) - it balances.

The Stanfield factor does throw it off a bit, unless I am interpreting it incorrectly. because playing the point on the PP, he's techically a "D" for those 2.3 minutes every game, so instead of 120 D minutes to go around, there are 117.7. Leaving Orr's minutes untouched (because we know he played the whole PP) that leaves 2.3 minutes to come from the likes of Green, Smith, Awrey, Doak and Gauthier.

The Stanfield case isn't exactly common, it should be pointed out. But one example and we have people saying "the system is broken! too many assumptions! throw it out!" as though it is suddenly not useful for any team in any season.

Inaccurate from both ends. They list Ed Westfall as a defenseman his first three season but he also played RW. The O6 teams would play a minimum of 5 dmen per game especially if one was a fighter - Ted Green on the Bruins.

1964-65 Bruins if you exclude Westfall from the dmen, show:

303 out of 350 Games for dmen. No other Bruin listed as a forward could drop back and play defense. Similarly throughout his Bruin career.

Furthermore HR lists Ron Stewart exclusively as a RW, yet in the mid/late fifties he played defense, sufficiently to earn AST votes as evidenced in the sticky on this board.

Westfall's listed position is largely irrelevant. The time he might have spent on defense is factored in his total ice time and the per game average is somewhere between how much he played on F and D. The "missing" ice time isn't missing at all, it just isn't listed separate from his stats when playing forward.

Stanfields specific role in Bruins' power play is also quite irrelevant. Whether it was him or some other forward who played point doesn't matter. All that matters is that when a team that uses 4 (or 5) forwards on the power play everyone who's on the ice gets credit for the goal scored and everyone who isn't doesn't. If the situational ice times are calculated separately there shouldn't be any issue.

Your explanation reduces the total of minutes played by defensemen per game in the pre regular season overtime era from 120 minutes per game to a lower number while increasing the forward time accordingly.

What is relevant is that proper and complete research was not done and this is getting spun to try and justify the discrepancies.

You have taken the science of mathematics and replaced it with an art form where data gets overlooked, massaged, fudged and semanticized to justify a lack of mathematical precision.

Play nice. If you want to stay in the thread (or the forum in general), make your points without ad hominem attacks, generalizations, or personal considerations.

Inaccurate from both ends. They list Ed Westfall as a defenseman his first three season but he also played RW. The O6 teams would play a minimum of 5 dmen per game especially if one was a fighter - Ted Green on the Bruins.

1964-65 Bruins if you exclude Westfall from the dmen, show:

303 out of 350 Games for dmen. No other Bruin listed as a forward could drop back and play defense. Similarly throughout his Bruin career.

Furthermore HR lists Ron Stewart exclusively as a RW, yet in the mid/late fifties he played defense, sufficiently to earn AST votes as evidenced in the sticky on this board.