Reference - VsX comprehensive summary (1927 to 2023)

[BenchBrawl]

It is the VsX measure that does that, not me I am fine with this "implicit equalizing" if we are comparing one season vs one season, but the more seasons come into play, the more I am inclined to do an adjustment. At 7 seasons, as in average VsX for best 7 seasons, I am barely putting up with this "implicit equalizing". At the career level, I am positive we have to have an adjustment.

Thoughts on how to better do this adjustment (other than multiply O6 VsX by 70/82=0.85) are welcome.

The whole point of VsX is to equalize the format across eras so we can make a direct comparison.

Stripped to its essence, your point is that a generic Art Ross winner is more valuable at 82 games than at 70 games. Hey, this is probably true on some level. Like playing Poker, in the long run the best players will win the money, but if we only play a few short sessions, weaker players can win.

There are others questions like that: for example, all else being equal, is the generic Art Ross winner more valuable if he had to score 150 points compared to another who had to only score 90 points? In the sense that, "he still had to physically make the moves and score those 50 points", even if a point is easier to come by in his era?

Or maybe there's an ugly factor for those playing in low-scoring era: is there more inherent variance as to who will be scoring the points, if the games are very low scoring? So like low games-played schedules, a low scoring environment is harder on the best players to show their real identity? This is basically the games played argument transfered to points totals.

There's so many angles like that we could rip appart and deconstruct, because inherent in those methods is circularity. We cannot dig and dig and hope to find the Holy Grail of all analysis (though we can find better and better ones). That's because as deep as you go in one direction, the other side can always override you due to the circularity.

The circularity is roughly this: An imaginary league with three players (X,Y,Z). The value of player X is determined against the value of his peers Y and Z. But the value of Y is determined against X and Z. And the value of Z against X and Y.

Now do that with another set (A,B,C), none of which played in the same era as X, Y or Z. So now, how do we know how B compares to Z ?

That's why continuity (playes who played against two generations) and the eye-test are crucial, to give some sort of sanity-check to the method.

 

[Johnny Engine]

Zuluss is right if they are positing that more games offers better sample to work with. Particularly when you get down to looking at something like points among defensemen in the 30s...it gets downright glitchy.
I think it's fine to take the results of such a metric less and less seriously, the fewer games they represent. It's possible that 1953's Gordie Howe could have slowed down in the last 12 games and failed to crack 100 points. But it's also possible that the goalies he shot against could have fallen to pieces in those extra 12 games and allowed Gordie to have a 120 point season.
What Zuluss is suggesting doesn't exactly make sense because VsX isn't a counting stat (or even an adjusted counting stat) and shouldn't be treated as such. But feel free to apply a sense of entropy-based skepticism to older results.

 

[Zuluss]

Stripped to its essence, your point is that a generic Art Ross winner is more valuable at 82 games than at 70 games. Hey, this is probably true on some level. Like playing Poker, in the long run the best players will win the money, but if we only play a few short sessions, weaker players can win.

This effect actually goes the other way: over a longer season, luck will run out, and if #1 is truly outstanding and #2 is just the luckiest of run-of-the-mill stars, the 1-2 gap will be larger in a 82-game season vs. a 70-game season. So comparing VsX from just one peak season across two players from two different eras seems fair: the player from the 82-game-seasons era works more, but also gets the chance to create more separation.

The problems start when the "truly outstanding" part wears out - nobody is truly outstanding over all of his 7 best seasons (well, except for maybe Big4 guys). Almost everyone is struggling to keep up with #2 by their 7th-best season, there is no talk about "you play more, you create more separation" - maybe you will create it the other way, #2, who is now a better player, will get further away from you. So outside of peak seasons, longer seasons do not benefit players - but the "extra work", "extra games" part remains.

It is not about Art Ross winners - all VsX measures in this thread (7-year average, 10-year average, career sum) are rewarding players for keeping up with #2, not as much beating #2 - how many players with VsX10 > 100 do we have? (Seven). And the longer you keep up, in terms of number of games, the better - but that's not how the measure works, it says "the longer you keep up in terms of number of seasons, no matter how long they are, the better" - and that sounds a bit wrong.

There are others questions like that: for example, all else being equal, is the generic Art Ross winner more valuable if he had to score 150 points compared to another who had to only score 90 points? In the sense that, "he still had to physically make the moves and score those 50 points", even if a point is easier to come by in his era?

That one does not seem likely - in a high-scoring era, SV % is lower and defense stops you less often. You still make the same moves, but the puck hits the goalie or a defenseman less frequently.

Or maybe there's an ugly factor for those playing in low-scoring era: is there more inherent variance as to who will be scoring the points, if the games are very low scoring? So like low games-played schedules, a low scoring environment is harder on the best players to show their real identity? This is basically the games played argument transfered to points totals.

It is convergence to the mean regardless, LLN/CLT work the same way irrespective of whether the mean is high or low. To the contrary, if the mean is high, the variance will be greater, but this is resolved by normalizing, which VsX does.

 

[BenchBrawl]

.

 

[Hockey Outsider]

Why is 110 the 2019 benchmark? 110/116 is 0.948 so shouldn't McDavid's 116 be the benchmark?

The "outlier" rule says that if the #1 scorer leads the #2 scorer by at least 10%, then we treat the #3 scorer as the benchmark. Since Kucherov was more than 10% ahead of McDavid, we used Kane as the benchmark.

Since the thread was bumped, I will ask my question too: can we really compare career VsX points/goals across eras?

O6 era had 70-game seasons, and currently we have 82-game seasons. Yet #2 in goals gets 50 goals regardless of the era. Effectively, O6 players are gifted 12 games they did not play and assigned pro-rated goals over those games - post-expansion players play extra 12 (and sometimes I think even extra 14) games each season, but that does not help them one bit in terms of career goals/points against O6 players.

There have been some good responses to this already. What I'll add is I don't think any HOH regular really relies on career goals/assists/points for anything serious. It still doesn't deal with the "accumulation" problem. For example, Patrick Marleau and Mike Gartner still rank ahead of Mario Lemieux and Mike Bossy in career goals. I present career numbers because I think they're somewhat interesting, but nobody should be using them for any serious analysis. (To be honest, if it would take me more than about fifteen minutes to update the career goal/assist/point tables annually, I probably wouldn't bother - it's not worth much more effort than that).

 

[seventieslord]

Whether you go by points, adjusted points, vsx or something else entirely, career accumulation is always the wrong way to go. Use some sort of standard that looks at as players best seasons, not something that allows them to pad their total whether they were great in their prime or not.

 

[decma]

Whether you go by points, adjusted points, vsx or something else entirely, career accumulation is always the wrong way to go. Use some sort of standard that looks at as players best seasons, not something that allows them to pad their total whether they were great in their prime or not.

If a player is still adding value beyond his best seasons, then I think that additional value needs to be taken into account in assessing his overall contribution. There is a big difference between padding totals at an Andreychuck-like pace and still contributing decent (if not spectacular) totals beyond your 7th or 10th best season. I don't see the merit in disregarding non-peak seasons when they fall into the latter category.

 

[Hockey Outsider]

The 12 main tables on the first page have been updated for 2020. (The benchmarks are 48 goals, 63 assists, and 97 points). Note that I didn't make any adjustments to account for the fact that teams played an unequal number of games. The spread is small (a three game difference from highest to lowest), and to make an adjustment for this factor implies an unrealistic level of precision in this approach.

Also, I noticed that my 2019 benchmarks for assists and points were slightly too low, and that's been corrected as well.

Some observations for 2020:

• Ovechkin led the league in goals for the 9th time (tied). He now ranks 5th all-time in career VsX goals (rapidly closing in on the 800 goal milestone), and is one big season away from moving into 3rd place. He has the 4th most productive ten-year goal-scoring prime; the reason he's not even higher is because his margins of victory have usually been small. He now ranks 23rd all-time in career VsX points.
• Crosby had another injury-shortened season and didn't change his peak rankings in any category. He ranks 24th all-time in career VsX points (10 points behind Ovechkin). But he crossed three VsX career milestones this year - 500 goals, 900 assists, and 1,300 points.
• Like his teammate, Malkin also had an injury-shortened season (although a more productive one). Malkin reached a VsX career milestone - 1,100 points. He now has the 25th most productive ten-year offensive peak (moving past Bryan Trottier).
• It's only a matter of time before Connor McDavid starts shooting up the rankings. His five-year offensive peak (which I don't publish the table for) is already in the top forty all-time, and that's including his injury-shortened rookie season. His seven- and ten-year peak rankings could, eventually, match Jagr's.
• Patrick Kane, at age 31, is well into his prime. That allows him to soar up the rankings as some of his weaker seasons get pushed out of the seven- and ten-year calculations. He now ranks 25th all-time in seven-year offensive peak, and 24th all-time in ten-year offensive peak. He's 62nd in career VsX points and should be in the top 25 within three years (barring major injuries).
• The same is true for Steve Stamkos. He's now 50th all-time in career VsX goals scored, and is within striking distance of 500. He has the 13th best seven-year goal-scoring peak, though we'll see if he still has the ability to push his ranking there. He's closing in on 900 career VsX points.
• Joe Thornton now ranks 9th in career VsX points. He moved ahead of Sakic and Yzerman. He now ranks 5th in VsX assists and, even with his reduced ability, could move into 3rd place with one more season (behind only Gretzky and Howe).
• It's too early for him to appear on these lists, but Nathan MacKinnon is putting together a great offensive resume. He surpassed 500 career VsX points.

 

[ted2019]

Thanks for all of your hard work on these.

 

[Hawkey Town 18]

@Hockey Outsider

For vs.X for goals/assists, are the 7 years being used the same 7 years used for each player's vs.X points score?

To me it seems this should be the rule, otherwise you're getting a distorted picture of the total offensive package a player brings.

Same would obviously apply for 10 years or any other length of time.

 

[Hockey Outsider]

@Hockey Outsider

For vs.X for goals/assists, are the 7 years being used the same 7 years used for each player's vs.X points score?

To me it seems this should be the rule, otherwise you're getting a distorted picture of the total offensive package a player brings.

Same would obviously apply for 10 years or any other length of time.

The tables for goals and assists are looking at goal-scoring and playmaking in isolation. In other words - I'm showing a player's seven or ten best years for goals, even if that isn't their seven or best ten seasons overall.

Example - Joe Thornton's 2nd best goal-scoring season was 2001. But it was only his 13th best season overall. That season gets included when I calculate his best seven years for goal-scoring, even though it's excluded from his best seven years overall.

In this sense, the numbers benefit a player who has distinct goal-scoring and playmaking peaks.

 

[Hockey Outsider]

This entire thread has been updated for 2021. I'll give my standard disclaimer - all data has been presented in good faith, but if you see any errors, let me know.

A few people have been working on improvements to VsX. As I've said in other threads (but it's worth repeating here) - if someone can improve the method, I'm all for it. Still, I figured there's some value in presenting the numbers under the "generally accepted" formula, at least until there's a consensus on what the new method should look like.

McDavid should shoot up these charts over the next years. Looking at players' best five years, he currently has the 8th highest result all-time (behind only Grezky, Esposito, Howe, Orr, Lemieux, Jagr, and Lafleur).

Kane quietly had another very strong year. He now has the 20th highest seven-year prime and 19th highest ten-year prime. He's 41st and rising fast in VsX career point. (To be clear, I've often said that career totals, even when adjusted, aren't usually a good way to evaluate a player, but I figured there may be some interest in this).

Crosby narrowly improved his ten year result (but didn't move up from 9th place). He now ranks 20th in VsX career points. (Crosby, and Ovechkin for that matter, are in their late primes - not likely to push their seven- or ten-year scores much higher, but they can certainly boost their career totals).

Ovechkin now ranks 4th in VsX career goals (virtually tied for 3rd place), and 22nd in VsX career points.

Thornton now ranks 4th in VsX career assists (a few helpers out of 3rd place), and 8th in VsX career points. This may have been his last season.

 

[Czech Your Math]

Can you please post Jagr's year by year VsX for at least his best 10 years?

 

[Hockey Outsider]

Can you please post Jagr's year by year VsX for at least his best 10 years?

Sure:

Player	Year	Points	Benchmark	VsX
Jaromir Jagr	2001	121	96	126
Jaromir Jagr	1996	149	120	124
Jaromir Jagr	1999	127	107	119
Jaromir Jagr	2006	123	106	116
Jaromir Jagr	1998	102	91	112
Jaromir Jagr	2000	96	94	102
Jaromir Jagr	1995	70	70	100
Jaromir Jagr	2002	79	90	88
Jaromir Jagr	1997	95	109	87
Jaromir Jagr	2004	74	87	85
Jaromir Jagr	2007	96	114	84
Jaromir Jagr	1994	99	120	83
Jaromir Jagr	2014	67	87	77
Jaromir Jagr	2016	66	89	74
Jaromir Jagr	2003	77	104	74
Jaromir Jagr	2008	71	106	67
Jaromir Jagr	1993	94	148	64
Jaromir Jagr	2013	35	57	61
Jaromir Jagr	1992	69	116	59
Jaromir Jagr	2012	54	97	56
Jaromir Jagr	2015	47	86	55
Jaromir Jagr	2017	46	89	52
Jaromir Jagr	1991	57	115	50
Jaromir Jagr	2018	7	102	7

[TBODY] [/TBODY]

 

[jigglysquishy]

I've been thinking a lot about VsX recently and I would like to take a stab at another rule that will help to "normalize" the benchmark. If we assume the benchmark is the "normal" second place player then there are some instances where the benchmark is more of a "normal" first place player. I really love VsX as a system and think it could benefit from a different perspective (and subsequent rule) to help improve some of the variance.

We already have the rule where if #3/#2 <0.9 then #2 is an outlier. Love it.

But we keep seeing seasons where having #2 or #3 as the benchmark doesn't quite feel right. So I have another way of looking at the benchmark.

If a 10% lead is considered an outlier, then wouldn't a 5% lead be considered a "normal" range for first place above second place? So, if #3/#1 is <0.9 AND #3/#2 is <0.95, then we treat #3 as the benchmark. Afterall, this would be a season where #1 is leading by a wide margin, and #2 is leading the pack by a "normal" margin. So #2 would be a normal Art Ross winner, not a normal second place. Keep this for the entire top 10. It enables more than 1 or 2 outlier seasons as we no longer look at #1 as an outlier, but rather any VsX larger than 1 as being a "normal Art Ross winner". And the benchmark becomes more reflective of the "pack".

Here is every season after 1979-80.

Season	OG Benchmark	New Benchmark	Change
1979-80	119	106	Rule 3 artificial number -> New Rule at #4
1980-81	135	119	#2->New Rule at #4
1981-82	147	119	#2->New Rule at #6
1982-83	124	118	#2->New Rule at #4
1983-84	121	119	Orr rule applies evenly, though I've adjusted the benchmark to #3 as Coffey is not a "normal Art Ross Winner"
1984-85	135	117	#2->New Rule at #6. Oilers players at #1,#2, and #5 is kind of an Orr rule anyways
1985-86	141	131	#2->New Rule at #4. Oilers players at #1,#3, and #4 is kind of an Orr rule anyways
1986-87	108	107	Oilers break the rule again so I have gone with Lemieux at #3 instead of Kurri at #2
1987-88	131	121	#3->New Rule at #4. Savard was the "normal Art Ross winner" so Hawerchuk is the benchmark
1988-89	139	115	Rule 3->New Rule at #5. Recognizing that Lemieux, Gretzky, Yzerman, and Nicholls are all outliers
1989-90	129	113	#2-> New Rule at #5
1990-91	115	115	No change
1991-92	116	109	Rule 3 -> #4
1992-93	148	142	Lemieux is not a normal Art Ross winner, so LaFontaine is treated as it. So Oates at #3 is the benchmark
1993-94	120	112	#2-> New Rule at #3
1994-95	70	70	No change
1995-96	120	120	No change
1996-97	109	99	#2->New Rule at #3
1997-98	91	91	No change
1998-99	107	101	#2->New Rule at #3
1999-00	94	94	No change
2000-01	96	96	No change
2001-02	90	85	#2->New Rule at #3
2002-03	104	104	No change
2003-04	87	87	No change
2005-06	106	106	No change
2006-07	114	108	#2->New Rule at #3
2007-08	106	98	#2->New Rule at #3
2008-09	110	97	#2->New Rule at #4
2009-10	109	95	#2->New Rule at #4
2010-11	99	99	No change
2011-12	97	84	#2->New Rule at #4
2012-13	57	57	No change
2013-14	87	87	No change
2014-15	86	86	No change
2015-16	89	89	No change
2016-17	89	89	No change
2017-18	102	102	No change
2018-19	116	110	#2->New Rule at #3

[TBODY] [/TBODY]

I didn't do the last two seasons because HFboard won't let me add new rows to the table. But no changes occur.

Things I really like about this method:

• It recognizes season's like McDavid in 2019, Stamkos in 2012, Crosby/Ovi in 2010, Malkin in 2008, and Selanne 1999 as "outlier" seasons where they would have been the normal Art Ross winner.
• It recognizes that multiple "Art Ross caliber" seasons can happen at once. For years like 2008-09, where three all-time greats are all their peak play, it makes sense that all three would have been a normal Art Ross winner.
• There's less variance year-over-year, which makes more sense than huge jumps around

Thing I don't like

• Gretzky completely breaks this. And the way scoring was in the high 80s means the benchmark is frequently at the #4-#6 spot.
• I don't think Pat LaFontaine's 148 points is a reasonable benchmark for second place so I had to cheat to adjust, but I can't actually demonstrate it mathematically
• Gretzky breaking the game means his dominance becomes even more evident in this model, which perhaps is not right.

We have the Andy Bathgate rule of 7% above/7% below. This is a variation of that.

I took a quick look at 1945-1979, and the rule works fairly well from 1945 to about 1970 and then is broken the whole decade. A combination of Orr defying possibility and expansion craziness. I don't offer a solution. But I think Sturmonator creating the Orr rule is reflective that the entire era needs fudging.


All data has been provided in good faith, but it's all manual so please correct any errors.

A quick look at how selected VsX changes

Person	VsX7	VsX7 Adj	Change
Wayne Gretzky	155.6	167.9	+12.3
Mario Lemieux	135.2	148.7	+13.5
Jaromir Jagr	114.2	115.2	+1.0
Alex Ovechkin	98.4	103.7	+5.3
Sidney Crosby	102.4	106.8	+4.4
Evgeni Malkin	93.7	99.4	+5.7
Patrick Kane	96.2	98.6	+2.4
Connor McDavid	89.2	90.0	+0.8

[TBODY] [/TBODY]

 

[jigglysquishy]

If my math is correct, McDavid is now sitting at a Vsx7 of 104.5, including his injury shortened rookie year. That's already overtaking Crosby.

A strong year next year can really vault him up.

 

[Hockey Outsider]

If my math is correct, McDavid is now sitting at a Vsx7 of 104.5, including his injury shortened rookie year. That's already overtaking Crosby.

A strong year next year can really vault him up.

I'll post the tables over the next few days, but that's exactly right - McDavid is up to 104.5 over his best (and only) seven years.

That puts him tied with Guy LaFleur for 10th place. Obviously, this includes his injury-shortened rookie season. Barring a serious injury, he'll pass Beliveau, Hull and Mikita next season.

 

[Hockey Outsider]

This entire thread has been updated for 2022. I'll give my standard disclaimer - all data has been presented in good faith, but if you see any errors, let me know.

To quote myself from last year -A few people have been working on improvements to VsX. As I've said in other threads (but it's worth repeating here) - if someone can improve the method, I'm all for it. Still, I figured there's some value in presenting the numbers under the "generally accepted" formula, at least until there's a consensus on what the new method should look like.

McDavid is skyrocketing up these lists. At age 25, he's already tied for 10th in best seven years (in point production). That includes his injury-shortened rookie season. Barring injury, he should end up in 7th place as early as next year.

Ovechkin now ranks 2nd in career goals. He's still well behind Howe, but he's already easily surpassed Gretzky by this metric. Ovechkin ranks 5th in goal production over his best seven years, and 4th in goal production over his best ten years. Ovechkin also ranks 16th in career points (and is in the top twenty for both seven- and ten-year peaks).

Crosby pushed his career totals a bit higher. He's now 15th in points.

Kane (again) quietly had a very good year offensively. (Would you have guessed he was 3rd in scoring over the past four seasons?) He's now 32nd in points.

This was, believe or not, one of Stamkos's best seasons in terms of overall point production. He's in the top 15 in seven- and ten-year peaks for goal-scoring. VsX already has him over 500 goals (in fact, he's over 550 goals) and 1,000 points. He has his weaknesses, but I think he's generally been underrated.

Thornton is (presumably) at the end of his Hall of Fame career. His paltry five assists were enough to tie him with Jaromir Jagr for 3rd all-time. He just barely squeezed past Ron Francis for 7th in points.

 

[jigglysquishy]

Thank you very much for posting this.

I have to double take a lot of the McDavid stats. I've never seen someone climb this high this quickly.

Taking a deeper look at all the tables, I'm continually floored by Gretzky.

He twice had a Vsx of 1.694. That's the equivalent of 195 points in the 21-22 season.

 

[Batis]

Also interesting to note that the 7-year VsX of Nikita Kucherov jumped from 80.1 to 85.7 in spite of him missing more than 40 percent of the season. His previous seventh best season was his rookie year where he only scored 18 points in 52 games so his 69 points in 47 games which placed him second only to McDavid on the points per game list was enough to make that big of a jump.

 

[FBOTL]

Hello, the link listed to Sturminator's analysis (method to calculate VsX) is invalid. Could you please direct me to a link that explains how to calculate VsX for player categories. I appreciate the data included below, but it only goes to 250 players. I'm looking for the top 500 players in goals, assists and points. Thanks

 

[Black Gold Extractor]

Hello, the link listed to Sturminator's analysis (method to calculate VsX) is invalid. Could you please direct me to a link that explains how to calculate VsX for player categories. I appreciate the data included below, but it only goes to 250 players. I'm looking for the top 500 players in goals, assists and points. Thanks

Here is the link.

 

[Hockey Outsider]

Hello, the link listed to Sturminator's analysis (method to calculate VsX) is invalid. Could you please direct me to a link that explains how to calculate VsX for player categories. I appreciate the data included below, but it only goes to 250 players. I'm looking for the top 500 players in goals, assists and points. Thanks

Just to confirm, you're looking for the players ranked 251st to 500th in each of goals, assists and points, right? Which categories (7 year, 10 year, career totals)? Let me know as I should have time to do this tonight or Friday.

 

[FBOTL]

Just to confirm, you're looking for the players ranked 251st to 500th in each of goals, assists and points, right? Which categories (7 year, 10 year, career totals)? Let me know as I should have time to do this tonight or Friday.

OMG you're the best!!! all 500 players (1 to 500). Just career totals for goals, assists, and points for all 500 players. If it's easy, please include 7 and 10 year career totals for goals, assists, and points as well. Can't thank you enough!!! Thanks.

 

[Dingo]

OMG you're the best!!! all 500 players (1 to 500). Just career totals for goals, assists, and points for all 500 players. If it's easy, please include 7 and 10 year career totals for goals, assists, and points as well. Can't thank you enough!!! Thanks.

He is actually the best.

(still think this format needs to be vs #5ish instead of #2)