Best point producers of all time, a new take

[Deas]

Hello,

I've done some research I thought I'd share with you.

I've looked at the top players in points per game (P/G) for every NHL season that’s been played. I list the player that finished first and see how many percent higher his P/G was than the player that finished second in P/G.

This gives an interesting overview, and one quickly sees it’s no coincidence how familiar some names become throughout the study.

It shows us both who was most dominant production wise, and also for how big a portion of their careers they stayed at certain levels of production.

I haven’t found a study with this approach, nor does it seem to be taken into consideration very often. HockeyReference.com’s excellent stat sheets do include highlighting if a player’s P/G for each seasons was 1st in the league or not, but I haven’t seen a more direct focus on this aspect anywhere.

There are three sections on the page:

- About. What I've done and what I consider the pros and cons with this approach, along with how I've handled minimum requirement of games played per season.

- All P/G winners and win margins. A sheet of all seasons' P/G winners and win margins.

- Three different rankings of best point producers. 1: Numbers of first place finishes. 2: Listing the most doimant P/G win margins. 3: A more thorough ranking looking at numbers of wins, top 5 and 10 finishes, consistency within top 10 and win margins.

I've posted it on wordpress.

nhlbestscorers.wordpress.com

Thoughts?

Click to read more...

 

[bathdog]

Hello,

I've done some research I thought I'd share with you.

I've looked at the top players in points per game (P/G) for every NHL season that’s been played. I list the player that finished first and see how many percent higher his P/G was than the player that finished second in P/G.

This gives an interesting overview, and one quickly sees it’s no coincidence how familiar some names become throughout the study.

It shows us both who was most dominant production wise, and also for how big a portion of their careers they stayed at certain levels of production.

I haven’t found a study with this approach, nor does it seem to be taken into consideration very often. HockeyReference.com’s excellent stat sheets do include highlighting if a player’s P/G for each seasons was 1st in the league or not, but I haven’t seen a more direct focus on this aspect anywhere.

There are three sections on the page:

- About. What I've done and what I consider the pros and cons with this approach, along with how I've handled minimum requirement of games played per season.

- All P/G winners and win margins. A sheet of all seasons' P/G winners and win margins.

- Three different rankings of best point producers. 1: Numbers of first place finishes. 2: Listing the most doimant P/G win margins. 3: A more thorough ranking looking at numbers of wins, top 5 and 10 finishes, consistency within top 10 and win margins.

I've posted it on wordpress.

nhlbestscorers.wordpress.com

Thoughts?

Thanks for compiling the lists.

I usually don't participate in comparisons with oldtimers as I don't feel educated enough doing so, so I won't comment on the lists in depth.

From a statistical standpoint I'd say comparing only over 2nd place finish, while it clearly holds some value, is still flawed as it doesn't attempt normalize varying competition at all. I'd also say, saying 30% of a regular season (25 games) is enough sample size, is not close to reality. I could literally go through the most productive players and find dozens of 25-game spans that by far exceed their career averages, scoring oftentimes comes in clusters.

 

[Deas]

Thanks for compiling the lists.

I usually don't participate in comparisons with oldtimers as I don't feel educated enough doing so, so I won't comment on the lists in depth.

From a statistical standpoint I'd say comparing only over 2nd place finish, while it clearly holds some value, is still flawed as it doesn't attempt normalize varying competition at all. I'd also say, saying 30% of a regular season (25 games) is enough sample size, is not close to reality. I could literally go through the most productive players and find dozens of 25-game spans that by far exceed their career averages, scoring oftentimes comes in clusters.

Thank you for commenting. I agree with you about comparing to the 2nd place player only. I do mention this as a potential flaw of this approach, how it's hard to measure the level of the competition between different seasons. Comparing the first place finish to P/G average of the top 20 or something like that would be very interesting (hypothetically the top 20 of 1955 could still suck compated to the top 20 of 1975, however). Kind of like golf does with strokes ahead of the field average. It would take a ton of time though. What I've done so far certainly was time consuming enough. I think it's a great idea, however. A young Lemieux for example "only" finished 2nd for a few years because of his timing in coming in to the league with prime Gretzky being on another planet. Even a few seasons before starting racking up his own first place finishes in P/G Lemieux was way ahead of the top 20 average.

Concerning minimum requirements of games played my initial suspicion was exactly in line with what you're saying, but it proved to not be a factor really. I did not end up with any first place winner being there just because of few games played. Neither would there be any weird upsets if I lowered the requirement even lower. The only one would be one win for Paul Kariya, and he wasn't a slouch exactly. Doing this study showed me the P/G a player has over about 30 games is close to the number he'd likely have had he played 70 games (at least among those being factors in this study, it's really a best of the best thing). I was actually tempted to lower the limit 2 or 3 games as that would give Lemieux two more wins by big margins. That in a time of his career when he won with huge margins anyway, if he played more games. There is no player that would happen to win a season without being competitive any other years even if I lowered it as low as say 15-20 games.

Part of this is related to one of the bigger points of this survey. I want to know who was best at outproducting his own competition, if healthy. I don't want to punish injuried players. Career totals are easily found anywhere. I wanted to get a bigger picture than that.

Some people who argue staying healthy is part of the game and that they don't care if someone scored 50 points in 35 games while the Art Ross winner had say 80 in 80 games, and that P/G is speculative, won't like this study. But season and to further extent career totals wasn't what I was curious about. And speaking about P/G being speculative: As mentioned this survey has showed me that really isn't the case. Assuming that a player's P/G would be lower after 70 than 30 games, however, that more games carry a risk of lowering one's pace but not the chance of increasing one's pace - that is speculative.

And about "I usually don't comment on comparisons with old timers"; that's the main advantage of this study, how what era a player played in, or different eras' varying goals/G isn't a factor. This is a measurement of who fared how in points production compared to his own competition, and it treats everyone the same.

 

[bathdog]

Concerning minimum requirements of games played my initial suspicion was exactly in line with what you're saying, but it proved to not be a factor really. I did not end up with any first place winner being there just because of few games played. Neither would there be any weird upsets if I lowered the requirement even lower. The only one would be one win for Paul Kariya, and he wasn't a slouch exactly.

But you don't know they'd sustain that pace. There are cases where the difference is so large that it's a fair assumption, but there are those a lot closer as well.

Doing this study showed me the P/G a player has over about 30 games is close to the number he'd likely have had he played 70 games (at least among those being factors in this study, it's really a best of the best thing).

Care to back this up?

Some people who argue staying healthy is part of the game and that they don't care if someone scored 50 points in 35 games while the Art Ross winner had say 80 in 80 games, and that P/G is speculative, won't like this study.

Health is part of the game.

I'd take the 80 in 80 guy every day, and I'm a bit of a peak guy. The 93 in 65 (same ppg) guy, I'll listen.

And speaking about P/G being speculative: As mentioned this survey has showed me that really isn't the case. Assuming that a player's P/G would be lower after 70 than 30 games, however, that more games carry a risk of lowering one's pace but not the chance of increasing one's pace - that is speculative.

Sure, if the pace doesn't strongly deviate from anything the player has previously done over a full season (going to assume there aren't cases where it's significantly lower while still leading the league in ppg).

And about "I usually don't comment on comparisons with old timers"; that's the main advantage of this study, how what era a player played in, or different eras' varying goals/G isn't a factor. This is a measurement of who fared how in points production compared to his own competition, and it treats everyone the same.

I've learned that's a popular way to look at it around here, but I still refrain from making too in depth statements about players I can only compare stats from, when there are those that can do both that and having had the chance of following their career. They're better suited.

 

[Deas]

But you don't know they'd sustain that pace. There are cases where the difference is so large that it's a fair assumption, but there are those a lot closer as well.

Care to back this up?

I do back it up below. I don't assume a player's pace will go down or up with more games played, I assume it'd be the same. Anything else is speculation. We can dig up examples of both happening. I don't get why so many assume a lowered pace but never an increase,. Look at Thornton's last 10 games his fantastic 05/06 season.

With the guys who are factors in this study it turned out to hold up. There is no one securing a 1st place finish due to low amount of games without being a factor in more seasons as well. Raising the required amount would cut a few 1st place finishes, however, by players when you look at their seasons before and after that season with lower games played, are still at around the same P/G.

Health is part of the game.

I'd take the 80 in 80 guy every day, and I'm a bit of a peak guy. The 93 in 65 (same ppg) guy, I'll listen.

Well of course it is. I phrased myself a bit unclearly given what my main point is. If I'm a GM with a crystal ball I too take the 80/80 player before the 35/50 player. Well, depending on the timing of the injury. If healthy down the stretch and if your team is in the playoffs anyway I take the 35/50 guy first. Also, the 35/50 guy combined with his replacement might be better than the 80/80 guy.

Those are speculative details however. My main point is that from the perspective of who's the best point producer, that's the 35/50 guy, barring a substansial decrease of pace had he played more games. But again, with these elite players that showed to not be an issue.

Sure, if the pace doesn't strongly deviate from anything the player has previously done over a full season (going to assume there aren't cases where it's significantly lower while still leading the league in ppg).

See above. It turned out no one with significance in the study had a problem with maintaining pace in neighbouring seasons (with more games played). No one "stole" rankings in this study due to the 30% games played requirement being too low. Again, it still really punished Lemieux as 1 and 3 games more played in two seasons very, very likely (like, definitely) would've given him two more first place finishes. But the line must be drawn somewhere. And I don't want to go too low. If we make it say 15 games I'm sure what you suspect would happen a few times; a "nobody, in this context, "stealing" ranking.

I've learned that's a popular way to look at it around here, but I still refrain from making too in depth statements about players I can only compare stats from, when there are those that can do both that and having had the chance of following their career. They're better suited.

Well of course it's a popular way to look at things. Anything else punishes players in eras with less possibilites to score.

If we don't use this approach one could say "hey Howe isn't near the top players in P/G", altough that's irrelevant as he's still the second or third most dominant point producer of all time compared to his peers, depending on how one interprets the outcome of this study (and other studies too of course, I don't pretend to have a scoop about Howe, but he's a good example on this specific point).

That's why I like lists with scoring win margins, number of Art Ross:s etc. What I've done is take it one step further, by looking at P/G instead of season totals, and comparing everyone in the same way. Unlike some studies where a fan of a player finds one way to treat that players results, but without doing the same comparison with all other players. I've seen this with "adjusted stats" and when trying to pitch how Lemieux produced better than Gretzky. That's something I might make sub post about.

Where would you rank Jesse Owens as a 100 m runner? His best time is like 1200th all time by now, but his achievements vs his own competition of course puts him high among the all time best.

And again, thank you for great comments. I don't want to be too defensive about my thinking in this study There are flaws of course, as I point out in the About section.

 

[bathdog]

I do back it up below. I don't assume a player's pace will go down or up with more games played, I assume it'd be the same.

People tend to make decisions on a small set of data points.

Anything else is speculation. We can dig up examples of both happening. I don't get why so many assume a lowered pace but never an increase,. Look at Thornton's last 10 games his fantastic 05/06 season.

Assuming a player will score points he didn't score is the definition of speculation.

I like pace stats, and peak, but I'm not fine with certainity saying Player A is X% better than Player B when basing it on a very small set on games. For instance, I'm comfortable saying Crosby was likely to have beat Sedin for the scoring title in 2011, but I'm not comfortable quantifying by what amount.

Many assume a lowered pace because someone leading the league in ppg is a very high ppg to start with. Streaks happen.

If you lead the league in ppg at time of your injury you're obviously not in the middle of a cold streak, unless you're Gretzky perhaps.

Also, the 35/50 guy combined with his replacement might be better than the 80/80 guy.

Those are speculative details however. My main point is that from the perspective of who's the best point producer, that's the 35/50 guy, barring a substansial decrease of pace had he played more games. But again, with these elite players that showed to not be an issue.

I'll still take the 80/80 guy over the 35/50 + AHL player.

Of course the guy with the higher pace is the better offensive player, but not necessarily more valuable.

See above. It turned out no one with significance in the study had a problem with maintaining pace in neighbouring seasons (with more games played).

Is that entirely correct?

Crosby 11-13 avg 1.61ppg, his neighbouring seasons ~1.33ppg. Now I do think that was his peak years, but that's a large jump.

Gaborik 08/09 avg 1.35ppg, neighbouring seasons 1.08ppg and 1.13ppg.

Coffey 94/95 avg 1.29ppg, neighbouring seasons 0.96ppg and 0.97ppg.

Roberts 95/96 avg 1.20ppg, neighbouring seasons 0.50ppg (though only 8GP, season before that 1.15ppg but in higher scoring season) and 0.80ppg.

Elias 05/06 (yes this was high scoring season but so was 06/07) avg 1.18ppg, neighbouring seasons 0.99 and 0.92.

Stamkos 16/17 avg 1.18ppg, neighbouring season 0.83ppg.

Savard 03/04 avg 1.16ppg, neighbouring seasons 0.75ppg and 1.18ppg (high scoring 05/06).

Lupul 12/13 avg 1.13ppg, neighbouring seasons 1.02ppg and 0.64ppg.

That 10 of the top 20, or 50%, of the highest ppg producers that played between 15 and 45 games since 94/95-16-17. They may not all be in your study, but you use a very specific set of parameters.

I'd say you can make an argument short seasons with high ppg's can be unsustainable.

Well of course it's a popular way to look at things. Anything else punishes players in eras will less possibilites to score.

If we don't use this approach one could say "hey Howe isn't near the top players in P/G", altough that's irrelevant as he's still the second or third most dominant point producer of all time compared to his peers, depending on how one interprets the outcome of this study (and other studies too of course, I don't pretend to have a scoop about Howe, but he's a good example on this specific point).

That's why I like lists with scoring win margins, number of Art Ross:s etc. What I've done is take it one step further, by looking at P/G instead of season totals, and comparing everyone in the same way. Unlike some studies where a fan of a player finds one way to treat that players results, but without doing the same comparison with all other players. I've seen this with "adjusted stats" and when trying to pitch how Lemieux produced better than Gretzky. That's something I might make sub post about.

Most measurements has some flaws, you can make arguments from at least a few different angles imho.

Where would you rank Jesse Owens as a 100 m runner? His best time is like 1200th all time by now, but his achievements vs his own competition of course puts him high among the all time best.

I don't carry enough knowledge on the subject tbh. Both Bolt and Johnson have been awesome though!

And again, thank you for great comments. I don't want to be too defensive about my thinking in this study There are flaws of course, as I point out in the About section.

We all have different opinions, otherwise it wouldn't be interesting.

 

[Ingvar]

Yup, 30 games cut-off seems arbitrary and, if I had to guess, too small to be reliable. There is a way to check it though - one would just need to take full (or reasonably full if there isn't enough data) seasons and check how 30 game pace compares to the full season PPG. Over 82 game season it already gives 53 batches of consecutive games so you can count deviation over one player's season and then you can do it for a lot of players in a lot of seasons to see how the average looks. You can even take different amounts of consecutive games and compare averages until you find sample size which is acceptable - good stuff.

It's all rather easy if you already have an established database and probably too much work for a fan project if you don't. However, serious business such as comparisons of players in different eras demands proper commitment

 

[Deas]

People tend to make decisions on a small set of data points.

Assuming a player will score points he didn't score is the definition of speculation.

I like pace stats, and peak, but I'm not fine with certainity saying Player A is X% better than Player B when basing it on a very small set on games. For instance, I'm comfortable saying Crosby was likely to have beat Sedin for the scoring title in 2011, but I'm not comfortable quantifying by what amount.

Fair point. Again, not claiming the approach of this study is perfect, but I find it more fair than other ways I've seen. If specifically "best point producer" is what we want to look at, it's certainly more fair than career totals, and to a lesser extent even season totals (as that punishes injuried players).

I'll still take the 80/80 guy over the 35/50 + AHL player.

Of course the guy with the higher pace is the better offensive player, but not necessarily more valuable.

This. Your sentence that I've bolded. That's exactly this study. I'm not saying the players listed in this study are the best or most valuable players (everyone can draw their own conclusions). For starters it includes forwards only (with Orr as the lone exception, the freak), and it can't measure the value of two way forwards adding more value than their points. It´s strictly a measurement of who were the best point producers compared to their competition.

Is that entirely correct?

Crosby 11-13 avg 1.61ppg, his neighbouring seasons ~1.33ppg. Now I do think that was his peak years, but that's a large jump.

Gaborik 08/09 avg 1.35ppg, neighbouring seasons 1.08ppg and 1.13ppg.

Coffey 94/95 avg 1.29ppg, neighbouring seasons 0.96ppg and 0.97ppg.

Roberts 95/96 avg 1.20ppg, neighbouring seasons 0.50ppg (though only 8GP, season before that 1.15ppg but in higher scoring season) and 0.80ppg.

Elias 05/06 (yes this was high scoring season but so was 06/07) avg 1.18ppg, neighbouring seasons 0.99 and 0.92.

Stamkos 16/17 avg 1.18ppg, neighbouring season 0.83ppg.

Savard 03/04 avg 1.16ppg, neighbouring seasons 0.75ppg and 1.18ppg (high scoring 05/06).

Lupul 12/13 avg 1.13ppg, neighbouring seasons 1.02ppg and 0.64ppg.

That 10 of the top 20, or 50%, of the highest ppg producers that played between 15 and 45 games since 94/95-16-17. They may not all be in your study, but you use a very specific set of parameters.

I'd say you can make an argument short seasons with high ppg's can be unsustainable.

What I saw about the elite producers not fluctuating alot year to year based on amount of games is corret to quite some extent, yes. Some interesting anomalies showed years we happen to know a certain player was hampered by an injury, so it certainly is telling.

Again, it's a best of the best list. The players you list are mostly irrelevant in this study. There can be variations year by year, of course. Bernie Nicholls sais hi. I haven't claimed non elite scorers can't have bigger variations in P/G. Among the elite players, however, I didn't see the chosen games played requirement being much of a factor.

Take the player relevant in this context among those you list above, Crosby. Yes, he peaks in P/G 10-13, in shortened seasons. 11/12 is also excluded in the study, as he´s below the games played requirement level. 12/13 he plays about 75% of the amount of games of those who play all games, so it's a non issue.

More to the point, however; the seasons before and after those years, he's still elite. What you bring up is exactly what I've been saying. With the very elite the amount of games played didn't have a big affect on their ranking. On their win margin, sure it could be a factor (we'll never know) but ranking/position wise it wasn't much of a factor in this study.

I'll show you the individual page for Crosby (haven't added the individual sheets to the study yet).


Sidney Crosby

Entered league at age 18.

05/06: 6
06/07: 1 9.28% more productive than Joe Thornton
07/08: 2
08/09: 3
09/10: 4
10/11: 1 26.97% more productive than Daniel Sedin
11/12: TFG (1st, had he played 3 more games, 15.76% more productive than Evgeni Malkin)
12/13: 1 24.48% more productive than Martin St. Louis
13/14: 1 8.30% more productive than Evgeni Malkin
14/15: 1 0.55% more productive than Tyler Seguin
15/16: 4
16/17: 2

Retired at age x

11 seasons included in survey.

Number of seasons as 1st: 5
Number of seasons in top 5: 10
Number of seasons in top 10: 11
Years in/outside top 10 ratio: 11/0, 100%
Span in years between first and last top 10: 11 *still active

He's pretty sick I hope he gets another 1st place finish, so he ties Gordie Howe at third all time. If not it'll be frustrating that he very likely would've gotten another win in 11/12 if not for injuries. He certainly could've won this year. He was 1st with quite some margin for a long time, but cooled off (which is not evidence that a player might not as well get hotter by the end with more games played, which there are many examples of too ).

Parts of your last post I didn't quote are what I consider nuances of details where we can only agree to disagree or didn't really seem to disagree that much, or my thoughts on it was covered in this post anyway. Do tell me if you feel I "dodged" something.

 

[Deas]

Yup, 30 games cut-off seems arbitrary and, if I had to guess, too small to be reliable. There is a way to check it though - one would just need to take full (or reasonably full if there isn't enough data) seasons and check how 30 game pace compares to the full season PPG. Over 82 game season it already gives 53 batches of consecutive games so you can count deviation over one player's season and then you can do it for a lot of players in a lot of seasons to see how the average looks. You can even take different amounts of consecutive games and compare averages until you find sample size which is acceptable - good stuff.

It's all rather easy if you already have an established database and probably too much work for a fan project if you don't. However, serious business such as comparisons of players in different eras demands proper commitment

Good point. Not sure if I´ve saved data in a structure to do what you suggest in a way that wouldn't take forever, but I might look at it.

With that said, I think I can argue the chosen games played requirement level isn't a huge factor. This study showed me how the very elite producers (which is what this focuses on) were very consistently in line with their P/G at a given time of their career, regardless of if they played 35 or 75 games. Have you read my take on this above and below? I think you might have overreacted to the games played requirement level.

As posted, no one who's not very much a factor other seasons too stole rankings because I squeezed them in with only 25 games played. There were more cases actually where I felt I punished a player who very probably would've had another P/G 1st place finish had I lowered the limit 5 games to 20. Looking at their seasons before and after, their pace with fewer games, or at least rank in pace, wasn't noticeably different from their neighbouring seasons with more games played.

The only one that´s close to be what would've felt like a suspected benefactor of few games played is Paul Kariya, had I credited him with a 1st place finish in 97/98 when he had a P/G of 1.409 in 22 games. That would be somewhat of a stand out or anomalie for him. He was in no way a slouch though. He had 3 top 10:s and 2 top 5:s in the 13 seasons he's included in the study.

Otherwise it's players like Lemieux (twice) and Crosby (once) who'd be credited with more 1st place finishes if the limit is lowered to 20, and they truly back up those paces (or at least rank of their pace) in their neighbouring seasons with more games played.

Put it this way: Can you point to someone who's been credited with a 1st place finish who you feel very likely wouldn't have won that year had he played more games?

Spontanously I think those are very few, if any. And if it has happened, the surrounding top producers are still credited with a top 5 or 10 finish, which is included in my final rank (the third list under "Summary of the Greatest).

Another point: I feel now we are nitpicking into too small details. If one is to look at P/G instead of totals, in order to not punish injuried playes, one has to accept a small portion of speculation. I think it's very small though, as I haven't encountered any "anomalie 1st place finishers" because of few games played. The player with a 1st place finish that surprised me the most is Todd Bertuzzi in 01/02. But he played 72 games, so he earned it. It's a fact he had the highest P/G among the players playing 25 games or more.

Going back to a previous point, the same season (11/12) is an argument for actually lowering the limit, as Lemieux missed out with 1 gamed played at 24. Who honestly believes his P/G would´ve been lower than Todd Bertuzzi's had he played more games? (assuming reasonable health level, which in his case at that point admittedly is a stretch, but still).

I think this speculation factor is actually more punishing to elite players who's just barely missed 25 games, rather than too leniant as in allowing the "wrong" player a 1st place finish with only 30-50 games played.

 

[Big Phil]

I'm not terribly focused on PPG. It is a nice stat, sometimes it puts things into perspective and I am not saying it doesn't have its lustre to it, but the thing is, if Brent Burns played 60 games this year he probably is projected to have at least 90. Yet he ended with 76 and it probably cost him the Norris this year. However, if he played 60 games we'd have assumed he could have cracked the 40 goal mark if he were healthy. You just never know when a player is going to have a cold streak so I always prefer giving someone credit for that they did, rather than what they may have done.

 

[Deas]

I'm not terribly focused on PPG. It is a nice stat, sometimes it puts things into perspective and I am not saying it doesn't have its lustre to it, but the thing is, if Brent Burns played 60 games this year he probably is projected to have at least 90. Yet he ended with 76 and it probably cost him the Norris this year. However, if he played 60 games we'd have assumed he could have cracked the 40 goal mark if he were healthy. You just never know when a player is going to have a cold streak so I always prefer giving someone credit for that they did, rather than what they may have done.

Your approach risk discrediting a player however, by assuming he'd drop pace.

It'd be interesting to see how the stars' p/g look in their first 15 games of a season compared to their last 15. If there's a big difference that's a factor for sure. The cases I've looked at have shown examples of both.

When looking at the very elite point producers however (which is what this study focus on), there aren't any upsetting anomalies. The study doesn't hand out any 1st place finishes in p/g to players who were not among the top producers in neighbouring seasons in which they played more games.

I think the opposite is more of a factor actually. Lemieux for example very likely would've had two additional 1st place finishes if he hadn't played below the required amount of games played. Looking at his p/g rank and win margins in neighbouring seasons it's a no brainer.

 

[BenchBrawl]

Something like this has already been done in various formats, but it dealt with outliers and gave a result for every player compared to the 2nd PPG scorer, or an artificial benchmark according to some rules (put in place to deal with outliers).

PPGVsX and ReplacementVsX.You can find those in a thread created by Hockey Outsider IIRC.

http://hfboards.mandatory.com/threads/reference-vsx-comprehensive-summary-1927-to-2017.2215905/

replacement can be found somewhere in the ATD section, not sure where it is anymore.

 

[Johnny Engine]

Something like this has already been done in various formats, but it dealt with outliers and gave a result for every player compared to the 2nd PPG scorer, or an artificial benchmark according to some rules (put in place to deal with outliers).

PPGVsX and ReplacementVsX.You can find those in a thread created by Hockey Outsider IIRC.

http://hfboards.mandatory.com/threads/reference-vsx-comprehensive-summary-1927-to-2017.2215905/

replacement can be found somewhere in the ATD section, not sure where it is anymore.

One thing of note about replacement VsX; it was set up to be applied specifically to the ATD fantasy game - though it can be modified for more general use.

Basically, the idea is that a missing player forces you to use an inferior player for the time that he's out, but while he plays, the full value of his offensive contribution is condensed into the games he did play.

So you'd take Lindros's tradional VsX score - without looking let's say it's somewhere around 90 - and use it as a starting point.
Then you take the percentage of the available schedule he missed in his career - again to completely pull a number out of the air let's say 0.20 - and multiply it by a VsX score you've chosen to represent r. Then you add those totals together.

On an ATD team, you'd be replacing Lindros with a guy like Jason Arnott or Eric Staal - guys who hover around a VsX of 60 typically. So Lindros's 90 + 0.20 * 60 get's you something like 102. That's a more realistic estimate of his statistical offensive value than treating him the same as someone who maintained the same PPG over a full season. However, in real life r is someone a lot worse than Eric Staal. A real-life AAAA forward might be more likely to put up a single-season VsX score of 20-25 if he got into a full NHL season, so using that in the formula, Lindros ends up with something like 95.

The problem here is that there's no real life numerical basis at all, just the smell test of what a reasonable observer thinks a replacement player is. That's not to say you couldn't develop one, but we haven't done that in a way that handles multiple eras with any finesse, to my knowledge.

 

[BenchBrawl]

One thing of note about replacement VsX; it was set up to be applied specifically to the ATD fantasy game - though it can be modified for more general use.

Basically, the idea is that a missing player forces you to use an inferior player for the time that he's out, but while he plays, the full value of his offensive contribution is condensed into the games he did play.

So you'd take Lindros's tradional VsX score - without looking let's say it's somewhere around 90 - and use it as a starting point.
Then you take the percentage of the available schedule he missed in his career - again to completely pull a number out of the air let's say 0.20 - and multiply it by a VsX score you've chosen to represent r. Then you add those totals together.

On an ATD team, you'd be replacing Lindros with a guy like Jason Arnott or Eric Staal - guys who hover around a VsX of 60 typically. So Lindros's 90 + 0.20 * 60 get's you something like 102. That's a more realistic estimate of his statistical offensive value than treating him the same as someone who maintained the same PPG over a full season. However, in real life r is someone a lot worse than Eric Staal. A real-life AAAA forward might be more likely to put up a single-season VsX score of 20-25 if he got into a full NHL season, so using that in the formula, Lindros ends up with something like 95.

The problem here is that there's no real life numerical basis at all, just the smell test of what a reasonable observer thinks a replacement player is. That's not to say you couldn't develop one, but we haven't done that in a way that handles multiple eras with any finesse, to my knowledge.

The reason I introduced PPGVsX is because it is informative about the offensive value of a player on a per-game basis in a way that "standard" VsX is not.Standard VsX punishes players for missing games, which is good, and so it informs us of the offensive production of a player throughout an entire season, but if we want to know how impactful a player is on a per-game basis ,we needed PPGVsX.The two are to be looked at side by side, and not mixed in any way beyond that.

I feel it is more informative to have both numbers side by side, because even in the ReplacementVsX case, you will have situations where say players X and Y, with X being a superior per-game player, and both would have the same score.The information of who was the better per-game player is thus lost in the fusion.OTOH, I get that it's more convenient because it wraps the whole thing under a single number to see the "total value", with both PPG and point totals considered.

I know this post is a mess, I'm sleepy.

 

[Johnny Engine]

The reason I introduced PPGVsX is because it is informative about the offensive value of a player on a per-game basis in a way that "standard" VsX is not.Standard VsX punishes players for missing games, which is good, and so it informs us of the offensive production of a player throughout an entire season, but if we want to know how impactful a player is on a per-game basis ,we needed PPGVsX.The two are to be looked at side by side, and not mixed in any way beyond that.

I feel it is more informative to have both numbers side by side, because even in the ReplacementVsX case, you will have situations where say players X and Y, with X being a superior per-game player, and both would have the same score.The information of who was the better per-game player is thus lost in the fusion.OTOH, I get that it's more convenient because it wraps the whole thing under a single number to see the "total value", with both PPG and point totals considered.

I know this post is a mess, I'm sleepy.

Yeah, I think in the end the utility of a crunched number versus side-by-side comparison of multiple sets of data comes down to what you want to use it for, although it's always better to have more information at your disposal.
The main takeaway of my post, however, is the forewarning that r is a completely made-up number, and that if one were to stumble upon the RVsX calculations that Hockey Outsider did for us in the ATD thread a couple of years ago, you'd see results in which r is based on fringe ATD forwards - which is not a useful classification outside this one particular fantasy league.

 

[daver]

I like looking at the PPGs of the Top Ten scorers only along with their actual point totals to gauge the strength of an Art Ross win. Unless you were a Crosby or a Jagr, you would have likely 70 plus games under your belt to finish in the Top Ten.