Reference - VsX comprehensive summary (1927 to 2023)

[BenchBrawl]

The VsX benchmarks rules as stated by Sturminator in the OP:

1. First preference is to use the #2 scorer

2. If #3 points/#2 points < .90, I use the #3 scorer, unless...

3. There is a gap of greater than 10% anywhere else in the top-5 - following the same method as above: [small #]/[large #] < .90.

At that point, I take the first gap, and identify the upper outlier group (top 3 or 4 or 5 above which the gap occurs), and then go down into the scoring table until I reach a number of players which equals: [size of outlier group] * 2. The benchmark is set as an average of the scoring of these players.

4. If any player in the top-5 is more than 7% below the player above him and more than 7% above the player below him, his score is taken as the benchmark. [this is the Bathgate Rule]

5. The Wartime Fudge, as discussed in posts 131 and 159 of this thread. The new benchmark numbers are shown in parentheses next to the originals.

6. The Orr Rule, as discussed in posts #163 to #182. The new benchmark numbers are shown in parentheses next to the originals.

 

[BenchBrawl]

.

 

[Hockey Outsider]

The 12 main tables on the first page have been updated for the 2023-24 season. All data has been compiled in good faith. Let me know if you see any errors.

Kucherov won the Art Ross trophy. He climbed to 20th place all-time for best seven-year peak (he's right behind Ovechkin, and just ahead of Sakic). His ten-year prime score will continue to rise (he's only played ten years which includes a forgettable rookie campaign, and his injury-shortened 2022 season). His assist totals, especially as a winger, are impressive, but he still hasn't surpassed Howe, Jagr or a few others (Bathgate being perhaps the most forgotten of the elite playmaking wingers).

MacKinnon continues to rise. He ranks right between Bossy and Malkin in terms of seven year peaks. He already ranks ahead of Messier based on ten year prime, and that will presumably continue to rise as some of his weaker earlier seasons drop off (ie 2014 through 2017).

McDavid remains in 5th place for the best seven-year peak (he inched slightly closer to Lemieux). He's now in 11th place for the best ten-year peak (which is amazing because he's only played nine years, one of which was his injury-shortened rookie campaign). Barring a catastrophic career-ending injury, McDavid will eventually end up in at least 5th place (and possibly higher) based on best ten year prime. McDavid now has the 7th best seven-year peak for assists.

Matthews led the league in goals for the third time in four seasons. He now has the T-18th best seven-year peak (which will only increase, as some of his weaker seasons drop off).

Draisaitl's season (7th in scoring) felt forgettable. Still, it helped push his numbers higher. He now has the 17th highest seven-year peak (above Bathgate, Ovechkin, and Sakic). How much of that is attributable to McDavid, I suspect, will be discussed for a long time.

This was only the 14th best season of Crosby's career in terms of (adjusted) points, so it doesn't affect his seven- or ten-year scores. But he's now moved into 4th place all-time in terms of career points. He'll never catch up to Howe and Gretzky, but we'll see if he plays long enough to catch up to Jagr. Crosby now ranks 20th all-time in VsX career goals, and 10th all-time in VsX assists.

Ovechkin is 8th place all-time in terms of career points. He remains in 2nd place in terms of career goals. He won't pass Howe, but he could become only the second player in history to reach 1,000 VsX goals.

A few other veterans with improved career (point) totals - Kane climbs to 23rd, Malkin rises to 26th, Kopitar now ranks 42nd, and Stamkos is in 44th (26th in goals).

 

[The Panther]

To @Hockey Outsider:
I was wondering about something simple, and I apologize if this is a stupid question (math not being my strong point). Do you think it would be more functional to do -- instead of 'VsX" -- more like a "Vs.-Average"? Let me explain:

Rather than identifying a (well-selected, but still with considerable subjectivity) benchmark individual scorer each season, could the individual totals be calculated vs. the average of a certain number of top scorers?

For example, for the 1930s (or whatever) to 1967, could the benchmark simply be the average of the top-10 scorers in the League? From 1968 to 1991 (or whatever), the top 15 scorers? From 1992 to today, the top 20 scorers? (Or something like that?)

There would still be the question of whether to include extreme outliers, but my inclination is to include them. There's something uncomfortable to me about eliminating anything from the actual results. By doing averages instead of an individual-player-as-benchmark, outliers could (theoretically) be included in the totals.

Examples of "average" benchmarks:
2024 average of top-20 = 104 points
1984 average of top-15 = 116 points
1954 average of top-10 = 55 points

Just a thought. Forgive my stupidity if this is silly...

 

[The Macho King]

To @Hockey Outsider:
I was wondering about something simple, and I apologize if this is a stupid question (math not being my strong point). Do you think it would be more functional to do -- instead of 'VsX" -- more like a "Vs.-Average"? Let me explain:

Rather than identifying a (well-selected, but still with considerable subjectivity) benchmark individual scorer each season, could the individual totals be calculated vs. the average of a certain number of top scorers?

For example, for the 1930s (or whatever) to 1967, could the benchmark simply be the average of the top-10 scorers in the League? From 1968 to 1991 (or whatever), the top 15 scorers? From 1992 to today, the top 20 scorers? (Or something like that?)

There would still be the question of whether to include extreme outliers, but my inclination is to include them. There's something uncomfortable to me about eliminating anything from the actual results. By doing averages instead of an individual-player-as-benchmark, outliers could (theoretically) be included in the totals.

Examples of "average" benchmarks:
2024 average of top-20 = 104 points
1984 average of top-15 = 116 points
1954 average of top-10 = 55 points

Just a thought. Forgive my stupidity if this is silly...

I think the idea is that outliers skew the averages by virtue of being outliers, and by including them in an average, you lessen the impact of those outlier seasons. I tend to agree with that for the record.

I would be interested to see what the results look like with the median "star player" (as defined by the highest scoring player on a team - so in a 32 team league the median star player would be the 16 scorer, 6 team league it would be 3, 30 team league it would be 15, etc.). I think that may fall apart in the O6 era just by virtue of being *so* small a number, but I'd still be interested in seeing what the data pops out.

I'm way too lazy to do it myself though.

 

[BraveCanadian]

To @Hockey Outsider:
I was wondering about something simple, and I apologize if this is a stupid question (math not being my strong point). Do you think it would be more functional to do -- instead of 'VsX" -- more like a "Vs.-Average"? Let me explain:

Rather than identifying a (well-selected, but still with considerable subjectivity) benchmark individual scorer each season, could the individual totals be calculated vs. the average of a certain number of top scorers?

For example, for the 1930s (or whatever) to 1967, could the benchmark simply be the average of the top-10 scorers in the League? From 1968 to 1991 (or whatever), the top 15 scorers? From 1992 to today, the top 20 scorers? (Or something like that?)

There would still be the question of whether to include extreme outliers, but my inclination is to include them. There's something uncomfortable to me about eliminating anything from the actual results. By doing averages instead of an individual-player-as-benchmark, outliers could (theoretically) be included in the totals.

Examples of "average" benchmarks:
2024 average of top-20 = 104 points
1984 average of top-15 = 116 points
1954 average of top-10 = 55 points

Just a thought. Forgive my stupidity if this is silly...

People have played with this idea before where you use a number of players per league size to compare a player to the “average first liner” for example.

I think it generally gave results similar to VsX but it takes a lot more work so VsX took the reins. I could be mistaken.

 

[MadLuke]

vs. the average of a certain number of top scorers?

That one of the main idea behind project like this, versus using a VsX approach:

Adjusted point - using elite Canadians player scoring

With different previous at adjusting players season, the latest attempt has been the following with those goals: 1) Only use the top players in the league to automatically adjust for scoring distribution change over time, if rules change, style of play, power play habit make scoring more...

forums.hfboards.com

Going from the top 10 players in the league up to 32 during 1986-1990 (expected peak Canadian top talents) window.

Using an average for example, the benchmark (before being adjusted for 82 games season) for the season
86-87: 89 pts
87-88: 95 pts +6.7%
88-89: 97 pts +2.1%
89-90: 98 pts +1%
90-91: 92 pts -6.1%

Raw Vsx with second position player would have has a benchmark
86-87: 108 pts
87-88: 149 pts +38%
88-89: 168 pts +13%
89-90: 129 pts -23.2%
90-91: 131 pts +1.5%

I think it generally gave results similar to VsX but it takes a lot more work so VsX took the reins. I could be mistaken.

More work than raw VsX for sure, but once you start having sub rules like gap more than 10% between player of position x and y, what do we do with Mario-Wayne (and now Mc-Drai some season), it can became similar or less work.

 

[BraveCanadian]

That one of the main idea behind project like this, versus using a VsX approach:

Adjusted point - using elite Canadians player scoring

With different previous at adjusting players season, the latest attempt has been the following with those goals: 1) Only use the top players in the league to automatically adjust for scoring distribution change over time, if rules change, style of play, power play habit make scoring more...

forums.hfboards.com

Going from the top 10 players in the league up to 32 during 1986-1990 (expected peak Canadian top talents) window.

Using an average for example, the benchmark (before being adjusted for 82 games season) for the season
86-87: 89 pts
87-88: 95 pts
88-89: 97 pts
89-90: 98 pts
90-91: 92 pts

Raw Vsx with second position player would have has a benchmark
86-87: 108 pts
87-88: 149 pts
88-89: 168 pts
89-90: 129 pts
90-91: 131 pts


More work than raw VsX for sure, but once you start having sub rules like gap more than 10% between player of position x and y, what do we do with Mario-Wayne (and now Mc-Drai some season), it can became similar or less work.

Yes this idea and (others like it) has been done a few times on the boards.

I fundamentally disagree with the Canadian thing it gets too myopic.

I do think that doing scores based on front liners - however they are defined - and scaling that with league size is a good start but obviously still has many limitations. There is no way to account for how strong the league is at a particular time in relation to another other than estimates.

 

[jigglysquishy]

I think Ontario is a better base line than Canada. Quebec used to produce half the league's stars, now it's a weak hockey program. Atlantic Canada wasn't pumping out stars until the 2000s. BC has slowly risen to be the second hockey power in Canada.

Ontario will be a more neutral benchmark.

 

[Namba 17]

Rather than identifying a (well-selected, but still with considerable subjectivity) benchmark individual scorer each season, could the individual totals be calculated vs. the average of a certain number of top scorers?

I did smth like this (not exactly like, though) here The best goalscorers in the NHL - 2
The main problem with what you suggested is "certain" number. As long as it is just random "pretty" number like 10, 15 or 20 it doesn't make more sense than VxV.
Now I think, that, probably a ratio between #1 scorer and league average will give you smth more informative, but have no time to check so far.

 

[MadLuke]

s long as it is just random "pretty" number like 10, 15 or 20 it doesn't make more sense than VxV.

I feel like a small pretty number would automatically do make more sense than VsX, say comparing to the average 2-5 scorer in the league instead of comparing to the number 2.

Outside added work (specially if it is by hands) what the downside ?

 

[The Macho King]

I feel like a small pretty number would automatically do make more sense than VsX, say comparing to the average 2-5 scorer in the league instead of comparing to the number 2.

Outside added work (specially if it is by hands) what the downside ?

I don't think any of these are theoretically fleshed out enough to say one is superior to the other. I kind of want to find one. The other thing is none of these will be objectively better than the others, but the comparison of what various numbers spits out will probably help us sniff out any that seem off.

Maybe I'll make it a project. I got passed over for a promotion due to some nepotism bullshit so I kinda feel like doing the bare minimum for awhile.

 

[WarriorofTime]

Yes this idea and (others like it) has been done a few times on the boards.

I fundamentally disagree with the Canadian thing it gets too myopic.

I do think that doing scores based on front liners - however they are defined - and scaling that with league size is a good start but obviously still has many limitations. There is no way to account for how strong the league is at a particular time in relation to another other than estimates.

And injury factor, if half the top 10 scorers get injured the next year, and each only play 40 games. The 1st line numbers are skew-y because you have 40 games of that guy's production and 40 games of replacement 1st liner's production and overall your top whatever number is going to be thrown off, without it saying enough about the league's difficulty for a healthy 1st liner to produce.

 

[The Macho King]

Alright so here is my proposed methodology. I'm going to do a sampling of years to see what we get from different generations. I'll do the most recent season (2023-24), Then roll a 10 sided die to get one season a decade back through the 50s. At some later date I'll probably do from consolidation through the war, but I figured the first blush I want to see how well this scales for league size and scoring environments.

I am going to set the benchmark at the league median star player (operating on the assumption that there is one star player per team on average). Not trying to quibble or split hairs on that - it's just terminology for ease of communication. So in a 32 team league, the median would be the average of #16 and #17. 6 team league average between 3 and 4. 21 team league it will be the 11th highest scorer, etc.

From there, I'll identify by a factor how much separation there is between the top scorers and that benchmark. I'll probably represent it as a factor (i.e. 1.6 times higher than the median scorer). If there's a better presentation idea (scaling as a point value somehow?) let me know.

And... that seems like it should give us some barometer on how impressive scoring a season is. Outliers shouldn't throw it off, and it should control for different scoring environments. Then, compare it to VsX and see if the different computations either confirm each other or raise some interesting questions. I'm just going to do it for points at first but might expand to goals and assists if it seems worth it.

The one potential problem I think I see at this early stage is the O6 era. But I won't know if that works or not until we see it.

Please let me know what you think. The goal isn't to undermine VsX, just provide a different data set and see if it sparks any discussion.

 

[jigglysquishy]

People have done similar work and it falls apart in 1967-1975.

When the league expands faster than talent expands it breaks.

 

[Professor What]

People have done similar work and it falls apart in 1967-1975.

When the league expands faster than talent expands it breaks.

Could the Orr-led Bruins have something to do with it too? I mean, there was a specific rule set forward in VsX because they broke the scale, and I can't help but notice that the dates coincide.

 

[Namba 17]

I feel like a small pretty number would automatically do make more sense than VsX, say comparing to the average 2-5 scorer in the league instead of comparing to the number 2.

Outside added work (specially if it is by hands) what the downside ?

Downside is the same as with VxV - what does it mean
Why do you take 10 players and not 9? Or 12?
Why the difference in VxV is 10%? What does 10 mean in that case? Why Bathgate rule is 7%?

 

[WarriorofTime]

Why not do 2.5X which is probably gonna be less noisy than 1X? 2-3 players getting opportunity and controlling out some of the injury variance seems reasonable.

Expansion era is always gonna be funky as you had half real teams and half minor league teams but kind of is what it is as far as looking at value over replacement value at the time.

 

[Dingo]

I think Ontario is a better base line than Canada. Quebec used to produce half the league's stars, now it's a weak hockey program. Atlantic Canada wasn't pumping out stars until the 2000s. BC has slowly risen to be the second hockey power in Canada.

Ontario will be a more neutral benchmark.

this actually sounded ridiculous to me at first glance, and i think i really like it now (soon after)

 

[jigglysquishy]

One of the beautiful aspects of the VsX system is that it recognizes that hockey history is more complicated than a (League Size x Y) formula.

Pre-Consolidation needs a special rule. WW II needs a special rule. Bobby Orr needs a special rule. Even bring in a special rule for WHA era talent depletion and Soviet stars being gone.

But a one size fits all approach will always break for the former three.

 

[The Macho King]

People have done similar work and it falls apart in 1967-1975.

When the league expands faster than talent expands it breaks.

What does breaking mean? Just for a quick back of the envelope math the benchmark points for 75 would be 95.5 (18 team league, so average 95 and 96 from 9th and 10th in the league). This would put Bobby Orr's season as a 1.413.

Compared to this season, for instance, the benchmark is 90. That would put Kucherov's season at a 1.6. Jamie Benn's Art Ross season would be 1.216. Gretzky's 1986 would be 2.067 (Mario's second place finish that season is 1.356). Crosby's 07 Art Ross is at 1.379. Kucherov's first Art Ross comes in as a 1.454.

Obviously these aren't complete tables, but is there anything there that immediately doesn't pass the sniff test? Orr's season is most notable because it was a Dman winning the Art Ross, and from a quick look it seems to be a solid if not legendary Art Ross season. And that seems.... right to me? Gretzky obviously has bonkers Art Ross seasons. Jamie Benn's is the weakest we've seen so far. Kucherov's look pretty solid so far based on the limited data points.

I don't think this is the end all be all, but I think these numbers may tell us two things. One, it will give us an idea of how impressive seasons stack up across generations and hopefully can avoid a lot of bending over backwards to control for Gretzky and Mario. Two, I think it *may* be a pretty good indicator of league health as measured by offensive talent.

Maybe it all falls apart when the numbers are compiled and we get a fuller look at the data, but I think it could still tell us *something*.

 

[jigglysquishy]

What does breaking mean? Just for a quick back of the envelope math the benchmark points for 75 would be 95.5 (18 team league, so average 95 and 96 from 9th and 10th in the league). This would put Bobby Orr's season as a 1.413.

Compared to this season, for instance, the benchmark is 90. That would put Kucherov's season at a 1.6. Jamie Benn's Art Ross season would be 1.216. Gretzky's 1986 would be 2.067 (Mario's second place finish that season is 1.356). Crosby's 07 Art Ross is at 1.379. Kucherov's first Art Ross comes in as a 1.454.

Obviously these aren't complete tables, but is there anything there that immediately doesn't pass the sniff test? Orr's season is most notable because it was a Dman winning the Art Ross, and from a quick look it seems to be a solid if not legendary Art Ross season. And that seems.... right to me? Gretzky obviously has bonkers Art Ross seasons. Jamie Benn's is the weakest we've seen so far. Kucherov's look pretty solid so far based on the limited data points.

I don't think this is the end all be all, but I think these numbers may tell us two things. One, it will give us an idea of how impressive seasons stack up across generations and hopefully can avoid a lot of bending over backwards to control for Gretzky and Mario. Two, I think it *may* be a pretty good indicator of league health as measured by offensive talent.

Maybe it all falls apart when the numbers are compiled and we get a fuller look at the data, but I think it could still tell us *something*.

If you using 0.5Y (with Y being number of teams in the league) here is what it looks like

Season	Y	Y Points	Art Ross/Y Points	2nd place/Y Points
1964-65	3rd	76	1.15	1.09
1965-66	3rd	78	1.24	1.00
1966-67	3rd	70	1.39	1.14
1967-68	6th	75	1.16	1.12
1968-69	6th	87	1.44	1.23
1969-70	6th	74	1.62	1.34
1970-71	7th	79	1.92	1.76
1971-72	7th	93	1.43	1.25
1972-73	8th	93	1.40	1.12
1973-74	8th	84	1.71	1.45
1974-75	9th	96	1.41	1.32

You can push this back and forward too, but you get an explosion with expansion. Orr explains him and the Bruins, but it doesn't explain Clarke's offensive dominance. Going from Vs3 to Vs9 in only 9 years is too rapid.

 

[WarriorofTime]

If you using 0.5Y (with Y being number of teams in the league) here is what it looks like

Season	Y	Y Points	Art Ross/Y Points	2nd place/Y Points
1964-65	3rd	76	1.15	1.09
1965-66	3rd	78	1.24	1.00
1966-67	3rd	70	1.39	1.14
1967-68	6th	75	1.16	1.12
1968-69	6th	87	1.44	1.23
1969-70	6th	74	1.62	1.34
1970-71	7th	79	1.92	1.76
1971-72	7th	93	1.43	1.25
1972-73	8th	93	1.40	1.12
1973-74	8th	84	1.71	1.45
1974-75	9th	96	1.41	1.32

You can push this back and forward too, but you get an explosion with expansion. Orr explains him and the Bruins, but it doesn't explain Clarke's offensive dominance. Going from Vs3 to Vs9 in only 9 years is too rapid.

But doesn’t that explain relative dominance to league and overall value in league? It would make sense if you double a league overnight, the best players are now more valuable to have compared to before. Would be the same if NHL had 64 teams next year. Of course the “haves”/“have not” factor makes it almost like two separate leagues, especially with an expansion division existing.

 

[jigglysquishy]

But doesn’t that explain relative dominance to league and overall value in league? It would make sense if you double a league overnight, the best players are now more valuable to have compared to before. Would be the same if NHL had 64 teams next year. Of course the “haves”/“have not” factor makes it almost like two separate leagues, especially with an expansion division existing.

When comparing scoring across eras it doesn't help.

If you're comparing Cook to Hull to Ovechkin, the league doubling in size overnight shouldn't artificially push one era ahead.

 

[WarriorofTime]

When comparing scoring across eras it doesn't help.

If you're comparing Cook to Hull to Ovechkin, the league doubling in size overnight shouldn't artificially push one era ahead.

There's no real way to statistically control out "the League was weak and watered down", only to create a fair statistical model that says "this is roughly the equivalent of what x number of points looked like in a given year." There's no real model that can say "the early 80s was a weak period because there were too many teams and the overall talent pool feeding into it from the junior/minor ranks wasn't deep or geographically diverse enough to sustain it in a strong enough manner." That's just my opinion.

People have to use their own analysis/subjectivity for how they consider the relative value of that. Although at least for the first couple of years, you could perhaps control out O6/Expansion Division and almost look at them certainly. There are other times this occurs and a useful step back on various schedules is worth examining, look at 2011-12 for instance, where you have the conundrum of al the top scorers being in the Eastern Conference but the best teams being in the Western Conference.. and then seasons like 2012-13 and 2020-21 where special rules basically meant there wasn't 1 unified NHL for the season.