Post-consolidation VsX Benchmarks

[Hockey Outsider]

Same thing with ten year weighted average:

Rank|Player|Goals
1 | Bobby Hull | 66.2
2 | Maurice Richard | 62.5
3 | Gordie Howe | 62.2
4 | Phil Esposito | 57.6
5 | Wayne Gretzky | 55.1
6 | Brett Hull | 52.4
7 | Jean Beliveau | 51.3
8 | Mario Lemieux | 51
9 | Mike Bossy | 50.2
10 | Frank Mahovlich | 48
11 | Nels Stewart | 48
12 | Jaromir Jagr | 47.4
13 | Bernie Geoffrion | 47
14 | Bill Cook | 46.8
15 | Stan Mikita | 46.3
16 | Charlie Conacher | 46.3
17 | Ted Lindsay | 46.1
18 | Teemu Selanne | 45.4
19 | Marcel Dionne | 45.1
20 | Pavel Bure | 44.3
21 | Roy Conacher | 44.3
22 | Ilya Kovalchuk | 44.2
23 | Alex Ovechkin | 44.2
24 | Norm Ullman | 44
25 | Steve Yzerman | 43.8
26 | Jarome Iginla | 43.7
27 | Peter Bondra | 43.2
28 | Luc Robitaille | 43.2
29 | Guy Lafleur | 42.4
30 | Busher Jackson | 41.7
31 | Jari Kurri | 41.4
32 | Joe Sakic | 41.3
33 | Keith Tkachuk | 41.2
34 | Bryan Hextall | 41
35 | Howie Morenz | 40.9
36 | Andy Bathgate | 40.7
37 | Sweeney Schriner | 40.6
38 | Marty Barry | 40.4
39 | Brendan Shanahan | 40.3
40 | John Bucyk | 40.3
41 | Toe Blake | 40.1
42 | John LeClair | 39.9
43 | Mike Gartner | 39.7
44 | Syl Apps Sr | 39.6
45 | Max Bentley | 39.1
46 | Marian Hossa | 39.1
47 | Michel Goulet | 38.9
48 | Syd Howe | 38.8
49 | Yvan Cournoyer | 38.7
50 | Camille Henry | 38.5
51 | Dany Heatley | 38.4
52 | Dickie Moore | 38.2
53 | Mats Sundin | 38
54 | Aurel Joliat | 37.9
55 | Pat LaFontaine | 37.8
56 | Joe Nieuwendyk | 37.7
57 | Cecil Dillon | 37.7
58 | Rick Martin | 37.7
59 | Alexander Mogilny | 37.6
60 | Lorne Carr | 37.6
61 | Theoren Fleury | 37.6
62 | Rod Gilbert | 37.4
63 | Dino Ciccarelli | 37.2
64 | Eric Lindros | 37.2
65 | Paul Kariya | 37
66 | Markus Naslund | 36.9
67 | Dave Keon | 36.9
68 | Lanny McDonald | 36.8
69 | Bill Mosienko | 36.7
70 | Jean Ratelle | 36.6
71 | Herb Cain | 36.4
72 | Doug Bentley | 36.4
73 | Alex Delvecchio | 36.4
74 | Mike Modano | 36.3
75 | Joe Mullen | 36.2
76 | Rick Nash | 36.1
77 | Tony Amonte | 36.1
78 | Dale Hawerchuk | 36
79 | Kenny Wharram | 36
80 | Mark Messier | 36
81 | Patrick Marleau | 35.8
82 | Sergei Fedorov | 35.8
83 | Jeremy Roenick | 35.8
84 | Darryl Sittler | 35.7
85 | Dave Andreychuk | 35.5
86 | Cam Neely | 35.5
87 | Ziggy Palffy | 35.5
88 | Steve Shutt | 35.4
89 | Gordie Drillon | 35.4
90 | Henri Richard | 35.4
91 | Milan Hejduk | 35.4
92 | Pierre Turgeon | 35.4
93 | Garry Unger | 35.3
94 | Woody Dumart | 35.3
95 | Ted Kennedy | 35.2
96 | Dit Clapper | 35.1
97 | Ken Hodge | 34.9
98 | Marian Gaborik | 34.8
99 | Bryan Trottier | 34.7
100 | Bill Barber | 34.7
101 | Glenn Anderson | 34.7
102 | Milt Schmidt | 34.6
103 | Mark Recchi | 34.6
104 | Gilbert Perreault | 34.4
105 | Pat Verbeek | 34.4
106 | Martin St. Louis | 34.4
107 | Vincent Lecavalier | 34.3
108 | Bernie Nicholls | 34.2
109 | Owen Nolan | 34.2
110 | Sid Abel | 34.1
111 | Alexei Yashin | 34
112 | Bill Guerin | 33.8
113 | Bill Cowley | 33.8
114 | Rick Vaive | 33.8
115 | Jacques Lemaire | 33.7
116 | Patrik Elias | 33.7
117 | Miroslav Satan | 33.6
118 | Rick Middleton | 33.5
119 | Daniel Alfredsson | 33.5
120 | Dean Prentice | 33.4
121 | Peter Stastny | 33.3
122 | Don McKenney | 33.2
123 | Reggie Leach | 33.1
124 | Steve Larmer | 33.1
125 | Tod Sloan | 33
126 | Cooney Weiland | 32.8
127 | Jean Pronovost | 32.8
128 | Ray Sheppard | 32.8
129 | Gaye Stewart | 32.7
130 | Harry Watson | 32.6
131 | Tim Kerr | 32.6
132 | Brian Bellows | 32.5
133 | Rick MacLeish | 32.3
134 | Ron Ellis | 32.1
135 | Paul Thompson | 32.1
136 | Lynn Patrick | 31.8
137 | Gary Roberts | 31.7
138 | John Ogrodnick | 31.7
139 | Bun Cook | 31.7
140 | Dick Duff | 31.6
141 | Alex Kovalev | 31.6
142 | Sid Smith | 31.6
143 | Johnny Gottselig | 31.6
144 | Stephane Richer | 31.5
145 | Glen Murray | 31.5
146 | Denis Savard | 31.5
147 | Rick Tocchet | 31.5
148 | Pierre Larouche | 31.4
149 | Elmer Lach | 31.4
150 | Eric Staal | 31.2
151 | Dennis Hull | 31.1
152 | Danny Gare | 31
153 | Larry Aurie | 31
154 | John MacLean | 30.9
155 | Ryan Smyth | 30.9
156 | Bob Nevin | 30.7
157 | Herbie Lewis | 30.7
158 | Brian Propp | 30.7
159 | Daniel Sedin | 30.5
160 | Pavol Demitra | 30.5
161 | Bobby Rousseau | 30.5
162 | Clint Smith | 30.4
163 | Dave Taylor | 30.3
164 | Ralph Backstrom | 30.2
165 | Bob Pulford | 30.2
166 | Claude Provost | 30.2
167 | Simon Gagne | 30.1
168 | Hooley Smith | 30.1
169 | Bill Goldsworthy | 30.1
170 | Jim Pappin | 30
171 | Charlie Simmer | 30
172 | Bobby Orr | 29.9
173 | Vic Hadfield | 29.9
174 | Kevin Stevens | 29.8
175 | Thomas Vanek | 29.8
176 | Adam Graves | 29.8
177 | Rick Kehoe | 29.7
178 | Butch Keeling | 29.6
179 | Geoff Sanderson | 29.5
180 | Tomas Sandstrom | 29.5
181 | Jimmy Ward | 29.5
182 | Olli Jokinen | 29.5
183 | Vincent Damphousse | 29.4
184 | Don Marshall | 29.4
185 | Petr Sykora | 29.3
186 | Rod Brind'Amour | 29.3
187 | Henrik Zetterberg | 29.3
188 | Greg Adams | 29.2
189 | Mike Bullard | 29
190 | Baldy Northcott | 29
191 | Peter McNab | 29
192 | Steven Stamkos | 28.9
193 | Jason Arnott | 28.9
194 | Dennis Maruk | 28.9
195 | Claude Lemieux | 28.8
196 | Andy Hebenton | 28.8
197 | Steve Thomas | 28.8
198 | Ray Whitney | 28.8
199 | Joe Thornton | 28.7
200 | Frank Boucher | 28.7
201 | Jeff Carter | 28.5
202 | John Sorrell | 28.4
203 | Geoff Courtnall | 28.4
204 | Paul MacLean | 28.4
205 | Ray Ferraro | 28.4
206 | Petr Nedved | 28.3
207 | George Armstrong | 28.3
208 | Shane Doan | 28.3
209 | Red Kelly | 28.2
210 | Dave Gagner | 28.1
211 | Ed Litzenberger | 28.1
212 | Pavel Datsyuk | 28.1
213 | Pit Martin | 28.1
214 | Paul Henderson | 28
215 | Sidney Crosby | 28
216 | Danny Grant | 27.9
217 | Murray Oliver | 27.9
218 | Buzz Boll | 27.8
219 | Bobby Clarke | 27.8
220 | Butch Goring | 27.7
221 | Bill Thoms | 27.7
222 | Wilf Paiement | 27.6
223 | Wendel Clark | 27.5
224 | Hec Kilrea | 27.5
225 | Brian Rolston | 27.4
226 | Pete Mahovlich | 27.4
227 | Doug Mohns | 27.3
228 | Ivan Boldirev | 27.3
229 | Ed Olczyk | 27
230 | Bobby Bauer | 27
231 | Ray Getliffe | 27
232 | Corey Perry | 27
233 | Paul Coffey | 26.9
234 | Todd Bertuzzi | 26.9
235 | Grant Warwick | 26.8
236 | Doug Gilmour | 26.8
237 | Vyacheslav Kozlov | 26.8
238 | Johnny Peirson | 26.8
239 | Tony Leswick | 26.8
240 | Vic Stasiuk | 26.7
241 | Eddie Wiseman | 26.7
242 | Ron Francis | 26.6
243 | Rene Robert | 26.6
244 | Petr Klima | 26.5
245 | John McKenzie | 26.5
246 | Mike Knuble | 26.5
247 | Brian Sutter | 26.5
248 | Harry Oliver | 26.5
249 | Bernie Federko | 26.5
250 | Tony Tanti | 26.4

 

[Rob Scuderi]

Can someone explain the calculations to get these scores? I know how to get their raw VsX score, but what do you multiply by/divide by in order to get these scores? I know weights are involved, but I'm not sure what they are.

The weights in order are 18, 19, 20, 19, 18, 17, 16.

Take a player's seven best seasons and multiply each score by the corresponding weight. Then take the sum of those seven weighted scores. Divide that sum by 127 and you get the weighted score.

 

[TheDevilMadeMe]

HO, while you're here, can you fix the tables you posted in posts 84 and 85 to take into account the adjustments to the formula (which were a result of many of us thinking the benchmark in a couple of years in the late 50s was set too low)?

 

[Hockey Outsider]

Updated - unless I missed something, the benchmark for the 1957 and 1959 seasons have been raised. This has a minor negative impact on the results for players who peaked then (primarily Beliveau, Bathgate, Litzenberger and Howe).

 

[Sturminator]

I'm going to try to work my way through what I think are reasonable fudges for the wartime players on a case-by-case basis. What follows is a re-post of a conversation about Toe Blake from this year's ATD thread.

--------------------------------------------------------------

It's possible that Blake's 1945-46 statistics are majority distorted like his 1944-45 ones. But I think it's just as possible that it was a legitimate late career spike year for him, similar to how his Hart year was an early career spike year (though not as good as his Hart year obviously).

In 1944-45, the Punch Line finished 1-2-3 in scoring, way above everyone else, Blake was the third, and Lach and Richard finished 1-2 in Hart voting.

But in 1945-46 (after most but not all players returned from the War), Blake actually led the Montreal Canadiens in scoring - 50 points for Blake, 48 for Richard, 47 for Lach. All 3 members played all 50 games: http://www.hockey-reference.com/teams/MTL/1946.html. Hard to tell anything from Hart voting, as only Montreal's goalie finished in the top 5, and we don't have anything past that for that year.

Not to necro an old conversation, but I think this is a fair point. One thing I'm trying to do this year is work up "adjusted" VsX stats for the players affected by the war. I have given up trying to do this systematically as the war hit each player at a different point in his career, but I think there is progress that can be made by evaluating the players as individuals. In Blake's case, the question of how we handle his finishes is a tricky one. Of this group:

75, 107, 84, 73, 83, 82, 77, 92, 96, 79

...I think the 92 from 1944-45 and the 79 from 1946-47 as the third wheel on the great Habs line can be safely thrown out, but that leaves us with the question of what to do with his 96% VsX finish in 1945-46. It was still a war-weakened league, and he was surrounded by all-time great linemates, but as you say, he also led his line in scoring that season. I'd say a fair appraisal of that season would be to just split the difference between his 7-season prewar peak average and the 96 that was his actual finish.

So his prewar peak average was an 83. Split the difference between 83 and 96, and you've got a 90. Insert that into his top-7 weighted VsX, dropping the 92 and the 79, and what you've got is a 7-season weighted VsX score of 85.7, which seems about right for Blake.

 

[Sturminator]

Ok, I've got an idea of how to handle the war years. I'm going to post the new benchmarks I've thought up here, and then work my way through how they affect the various wartime players. Any comments or critiques are welcome, as is help in crunching the numbers should anyone be so inclined.

1942-43: I don't see this season as particularly problematic. In all likelihood, the scoring champ, Doug Bentley, would have been at or around the top of "the pack" (so, he'd have been the benchmark, himself), in a completely normal season. The old benchmark used the averaging method because there is a big gap between the top three and the next group, but I think this is distorted. The problem in this year is not the guys at the top, but rather the crappy next group. Their crappiness is what forced us to use the averaging method for this season (due to the gap between #3 and #4), but I think that method is misguided in this case, and we should simply use Cowley's #2 score as the benchmark. For 1942-43, I think the benchmark should be set to a standard Vs2, using Cowley's 72 points.

1943-44: I think this is the single most problematic season of the war years. Herb Cain didn't only win the scoring title this year, but he won it by a healthy margin. I think this year is an absolute disaster, and my solution is a relatively tough one. I think we can use Bill Cowley, who was the #2 scorer and our benchmark from the previous season, as our benchmark for this season as well, prorated for the number of games he played.

What do I mean by this? Well, we simply assume that Cowley's offensive production remained constant from 1942-43 to 1943-44. That seems fairly reasonable, and is perhaps even generous to the 1943-44 season given that Cowley was evidently hurt during the year and missed 25% of the games. So we assume that a Cowley who plays 48 games (which he had played in 1942-43) is our benchmark, and we calculate from that how many points he would have reached at his scoring rate for 1943-44 had he played 48 games. For 1943-44, we arrive at a benchmark of 95 points.

Now, this is pretty harsh. Herb Cain, in his career year, goes from a VsX score of 106 to a score of 86. Doug Bentley goes from being the benchmark (and a score of 100) to a score of 81. And so on down the line. Among hall of famers, this obviously hurts Doug Bentley the most, but I don't think that is inappropriate considering that he is the one and only hall of famer who was playing at his real peak this season, and he got easily outscored by Herb Cain.

1944-45: Elmer Lach's first scoring title. The trick with this season is that Lach also won a scoring title in a full-strength league, so we cannot dismiss out of hand the idea that he was a worthy #1 scorer in this season. What does look suspicious, however, is the margin of his victory.

What I propose is simple: assume that Lach's offensive production was just as good in 1944-45 as it was in 1947-48, and by extension assume that a normalized benchmark scorer would have been behind Lach's year-end scoring totals by the same margin. Lach won the 1947-48 Art Ross trophy by a single point, 61 to 60, over Buddy O'Connor in the latter's Hart Trophy season. So...

61/60 [1947-48 margin of victory] = 1.02
80/1.02 [Lach's 1944-45 scoring output/above margin] = 78

I propose 78 as a benchmark for the 1944-45 season.

1945-46: Some argue that this is not really a war year, but I disagree. The league was clearly still quite weak, Apps only played 40 games, Schmidt was not in hockey shape, and the scoring leaderboard is just a hot mess. What to do?

Luckily, we can repeat the process we used for the 1944-45 season here. Max Bentley won the scoring championship in consecutive seasons starting in 1945-46. The problem in this year is that his margin is much bigger than in the following season in a full-strength league. So if we assume that Max is the legitimate scoring champ but recalculate the margin between #1 and #2 to establish our benchmark for the 1945-46 season, this is what we get:

72/71 [1946-47 margin of victory] = 1.01
61/1.01 [Bentley's 1945-46 scoring output/above margin] = 60

I propose 60 as a benchmark for the 1945-46 season.

------------------------------------------------------------------

Questions, comments or personal attacks are welcome. I really hate asterisks, and the problem of the war years has been bothering me ever since we started down this road with the VsX project. I want to do the war years players justice without making them look better than they are, and I also want to differentiate them from one another, as one asterisk is not necessarily the same as the next. Anyway, let me know what you all think.

I will start working my way through the adjusted numbers for the wartime stars using the above benchmarks.

 

[Sturminator]

Bill Cowley's top scoring seasons under the new benchmarks (war years in bold):

141, 100, 95, 93, 89, 83, 78, 75*

*8th best season, so would be dropped in a 7-season weighted average

7-season weighted VsX with new benchmarks = 97.4

 

[Sturminator]

Oh, one other piece of work to attend to: Doug Bentley and Roy Conacher's assist totals in the 1948-49 season. There is evidence that the Chicago scorekeepers were awarding phantom assists during Blackhawks home games in this season, and if you take a look at the assist tables, the separation between these two and the pack is simply galling, especially with respect to Conacher, who was not known as a playmaker.

Again, though, Bentley led the league in assists on one other occasion, so if we assume that he was just as good this year and no better, then we can calculate his adjusted assist totals by the same margin of victory and alter our benchmark (Doug was the benchmark scorer this season) accordingly. So...

37/36 [Doug's 1947-48 margin of victory in assists] = 1.03
29 * 1.03 [3rd place Paul Ronty's assist finish * Doug's previous margin of victory] = 30

Doug actually tallied 43 assists during the 1948-49 season. If we shave 13 assists off of his total (and that of Roy Conacher) to get him down to his previous margin of victory, Doug falls from 66 to 53 points, and Ted Lindsay/Sid Abel are now in 2nd place at 54 points. Roy Conacher also falls from 68 to 55 points, still in first place, but by a much saner margin.

I propose no change to the benchmark for the 1948-49 season (it was already Abel/Lindsay at 54), with 53 and 55 being the end-of-season scoring totals for Doug Bentley and Roy Conacher, respectively.

 

[Sturminator]

Doug Bentley's top scoring seasons under the new benchmarks (affected years in bold):

101, 98, 95, 87, 81, 77, 67, 66*

*8th best season, so would be dropped in a 7-season weighted average

7-season weighted VsX with new benchmarks = 87.2

 

[Sturminator]

Max Bentley's top scoring seasons under the new benchmarks (affected years in bold):

102, 102, 97, 94, 90, 76, 59

7-season weighted VsX with new benchmarks = 89.4

*in Max Bentley's case, the biggest thing dragging his score down is his weak 7th best season score. We should consider that he lost two prime seasons to the war, and spent the latter part of his career playing on a lower line in Toronto without much in the way of scoring talent on his wings. If we make his 7th best season equal to his 6th best season (which is a fair fudge, I think, without overdoing it), Bentley's VsX score under the new benchmarks jumps up to 91.6, which is closer to his true offensive value, in my opinion.

 

[Sturminator]

Elmer Lach's top scoring seasons under the new benchmarks (affected years in bold):

103, 102, 94, 81, 78, 76, 70

7-season weighted VsX with new benchmarks = 86.8

 

[Sturminator]

Syd Howe's top scoring seasons under the new benchmarks (affected years in bold):

100, 100, 86, 82, 76, 75, 68

7-season weighted VsX with new benchmarks = 84.3

 

[Sturminator]

Toe Blake's top scoring seasons under the new benchmarks (affected years in bold):

107, 86, 84, 83, 83, 82, 73

7-season weighted VsX with new benchmarks = 85.3

 

[Canadiens1958]

1944-45 & 1945-46 NHL Assist Rule Change

The following about the June 1945 NHL Rule change about according assists may be of interest:

http://hfboards.mandatory.com/showpost.php?p=79176659&postcount=1

 

[TheDevilMadeMe]

HO, thanks for the update. Sturminator, I like what you're doing with the war years; by far the best attempt I've seen at coming up with a reasonable standard.

The following about the June 1945 NHL Rule change about according assists may be of interest:

http://hfboards.mandatory.com/showpost.php?p=79176659&postcount=1

I think that figuring out what to do with the years when the assist rules were in flux might be a next worthwhile step, but those years don't really stick out like a sore thumb like the war years did.

 

[Canadiens1958]

Rule Changes

HO, thanks for the update. Sturminator, I like what you're doing with the war years; by far the best attempt I've seen at coming up with a reasonable standard.



I think that figuring out what to do with the years when the assist rules were in flux might be a next worthwhile step, but those years don't really stick out like a sore thumb like the war years did.

I would suggest that over time it will be demonstratable that rule changes thru the O6 era had the greatest impact. Should have initial 1945-46 assist impact by mid week.

 

[Sturminator]

Get a hold of the current vsX sheet. Take a look from season to season and look at how many players meet a certain threshold: 65, 70, 75, 80, etc, and you'll see wild swings in that number, not because scoring changed drastically from year to year, not because the number of players hitting relative points benchmarks changed drastically, but because the benchmark from one year to another is drastically different due to the rigid rules being applied regarding which player is the benchmark.

Actually no, there are not. The work on this has already been done, and was part of the vetting of the system last year. The reader can judge for himself how well the system tracks to the expansion of talent in the NHL over time (since expansion, at least).

http://hfboards.mandatory.com/showpost.php?p=60959637&postcount=48

http://hfboards.mandatory.com/showpost.php?p=61159825&postcount=50

My old table - number of players meeting certain percentages of VsX threshold broken into blocks of time:

Years|>150%|>125%|>110%|>105%|>100%|>95%|>90%|>85%|>80%|>75%|>70%|>65%|>60%|>55%|>50%
2000-01 -- 2011-12 |0|0|0|1|1|3|6|9|14|21|30|42|56|74|98
1990-91 -- 1999-00 |0|0|1|1|1|4|5|9|13|18|25|34|44|59|77
1980-81 -- 1989-90 |1|1|1|1|1|4|4|6|9|12|19|26|37|51|68
1967-68 -- 1979-80 |0|0|1|2|2|3|5|7|9|14|19|25|34|44|58

You can't expect any system to be smooth from year-to-year because league scoring and individual player performance are not. You can expect a system to be smooth over time and track with our basic intuitions about the expansion of the talent pool, which VsX does.

second, vsX doesn't necessarily show who was a better producer or better offensively or whatever else you want to call it. it's a composite measure of a player's per-game production and their ability to play a lot of games.

You could have saved yourself a lot of typing and ended the post there. Yes, VsX measures offensive production in full seasons, not on a per-game basis, which obviously punishes those players who cannot stay healthy. How appropriate one considers this method depends on how much one values health, I suppose. I value it pretty highly, and think poor health is ultimately the reason why Pierre Turgeon will not make this list. Had be been able to stay healthy and done anything in the playoffs, he'd already be in; he certainly didn't lack the talent.

 

[Canadiens1958]

VsX

VsX 7 year scores for the available NHLers

Remember that VsX-7 is a measure of a player's best 7 regular seasons from a point-scoring standpoint compared to his peers, nothing more, nothing less.

Rank|Player|Score
31 | Dale Hawerchuk | 85.9
32 | Denis Savard | 85.4
33 | Eric Lindros | 85.4
34 | Alex Delvecchio | 84.9
35 | Gilbert Perreault | 84.6
36 | Darryl Sittler | 84.1
37 | Henrik Sedin | 82.8
38 | Clint Smith* | 82.6
39 | Mats Sundin | 82.3
40 | Doug Gilmour | 82.3
41 | Pierre Turgeon | 82.3
42 | Mike Modano | 81.7
43 | Jeremy Roenick | 81.5
44 | Ted Kennedy | 81.5
45 | Sergei Fedorov | 81
46 | Evgeni Malkin | 80.7
47 | Bernie Nicholls | 80.3
48 | Cooney Weiland | 79.4
49 | Pavel Datsyuk | 78.9
50 | Pat LaFontaine | 78.8
51 | Hooley Smith | 78.8
52 | Doug Weight | 78.6
53 | Brad Richards | 78.4
54 | Phil Watson | 78.1
55 | Alexei Yashin | 77.6
56 | Bernie Federko | 77.3
57 | Vincent Lecavalier | 77.2
58 | Henrik Zetterberg | 76.7
59 | Joe Primeau | 76
60 | Don McKenney | 75.8
61 | Jacques Lemaire | 75.5
62 | Jason Spezza | 75.2
63 | Phil Goyette | 74.9
64 | Vincent Damphousse | 74.2
65 | Bill Thoms | 74.2
66 | Marc Savard | 73.9
67 | Eric Staal | 73.8
68 | Neil Colville | 73.2
69 | Dave Keon | 73.2
70 | Rod Brind'Amour | 72.8
71 | Tod Sloan | 72.8
72 | Kent Nilsson | 71.7
73 | Dennis Maruk | 71.7
74 | Rick MacLeish | 71.6
75 | Buddy O'Connor* | 71.3
76 | Patrick Marleau | 71
77 | Ed Litzenberger | 70.8
78 | Joe Nieuwendyk | 70.5
79 | Pete Mahovlich | 70.3
80 | Billy Taylor* | 69.5
109 | Steven Stamkos | 63.6 N/A | Guy Carbonneau | N/A

*wartime star

Note on Stamkos: His score is so low because he only played 5 seasons (4 spectacular, 1 basically a write-off), so he gets zeros for his 6th and 7th best seasons. His 7 scores that are used to calculate the 7 year total: 100, 100, 91.92, 87.16, 48.81

Note on Carbonneau: He's so far behind the rest, it isn't worth listed. He's also the type of player that VsX should not be used for, since VsX is pinned to the top scorers in the league who receive ample PP time. If anyone wants to compare Carbonneau's offense to others, adjusted even strength points would be the best method, I think.

Note on Colville: He spent part of his career as a defenseman, so these numbers likely underrate him somewhat

More to the point, the VsX stat should not be used in the first place because it gives a false presentation of what actually happened.

Specifically it is a strictly offensive stat or is used strictly as such because it does not look at VsX against the best defensive teams or match-up, so filler or slush time points receive equal value to premium points. No different but more sophisticated than GWG. GWG overrates the value of the second goal in a 9-1 win while VsX overates the value of the points generated by the winning teams 9th goal. GWG benefits one player while VsX benefits up to three players. VsX may be viewed as three times as damaging.

VsX as presently used does not differentiate between two, three and four line rotations or other positional quirks or unbalanced situations. The ability and knowledge to adjust for these factors is present but it is not applied.

So VsX is reduced to being a lazy stat. Save the effort. Might as well rely on raw stats with adjustments.

You unwillingness to include Carbonneau drives home the last point very well.

 

[Sturminator]

VsX as presently used does not differentiate between two, three and four line rotations or other positional quirks or unbalanced situations. The ability and knowledge to adjust for these factors is present but it is not applied.

Actually no, it is not at all clear how one should adjust for different icetime/line rotation situations, as TOI does not track directly to scoring opportunities. Our knowledge of the specifics of line combinations and icetime for every team throughout hockey history is also quite limited. As far as game situations go, if you want to sift through the boxscores of every game in hockey history to figure out who scored what against whom and in which game situation, be my guest.

VsX is useful as a starting point for discussion of offensive value, not as the end-all and be-all of any analysis. Your objection amounts to the following:

1) It doesn't integrate every possible relevant data point.
2) All of the data is available (false).
3) Therefore, it is worthless.

VsX is an algorithm. Of course it doesn't perfectly handle every nuance of every goal scored, nor does anyone believe that it does. If you have a better system in mind, by all means, enlighten us.

 

[MXD]

I think everyone relying on the VsX charts know they come with caveats...

It's a very good metric to measure what it actually purports to measure. I wouldn't use TOI to evalutate how fast a skater is, nor will I use GAA to determine who's the best faceoff man.

In other words you now readily admit that insufficient data was considered and not enough work was done before the algorithm was presented but the situation is not the fault of the developers of the algorithm but the readers of the data are to blame.

I never claimed that all of the data is available. My claim is that all of the available data is not being considered or used to insure accuracy. The number of players a team used at a position is available. The size of game day skater rosters for each season with in season adjustments - before and after a specific date is known. The TG/G and scoring range is readily available for multiple seasons across eras to the start of the NHL. Trivial work but not used.

The raw data is a sufficient starting point. Just like hockey does not need another faceoff spot for the opening faceoff, the stat community does not need another starting point to move along.

In fact hiding behind a starting point position means others are required to do the necessary lackey work to dismiss something that is not acceptable in the first place.

Sadly it took the application/use of the VsX during most of the project to realize the weaknesses of the algorithm. The project became a VsX laboratory experiment unwittingly.

Did it taint the project? No.

Did it facilitate understanding? No.

What did it do? Well it focused on a few analytical weaknesses that are present. The inability or unwillingness to look at the impact of factors like roster size - including in season variations, before/after December 1, home/away, etc, player rotations thru a position, impact on scoring of rule changes - a goal was/is always a goal throughout the history of hockey but an assist was not always an assist. This impacts relative scoring since rebound or continuation or second(briefly third) assists varied over the history of hockey.

Will continue this in the next few days in an appropriate thread if necessary.

 

[Sturminator]

In other words you now readily admit that insufficient data was considered and not enough work was done before the algorithm was presented but the situation is not the fault of the developers of the algorithm but the readers of the data are to blame.

How the system functions is quite simple. It is explained in a single post and quickly understandable. So yes, I do expect the people looking at the data to understand its limitations, as the are obvious with just a little bit of thought.

What you are doing is tantamount to criticizing the theory of conservation of energy because general relativity showed there was more going on. Of course there is more going on. Anyone with an understanding of hockey can perceive that, but it is impossible to model these factors with incomplete data.

The human mind can only hold so many variables in play at one time. We need shorthand methods to aid us in our judgments, and we have always had them. The old shorthand method was comparing top-X finishes on leaderboards, or comparing raw scoring or what have you...all considerably worse methods of evaluation across eras because they do not account for the changing size of the league or the talent pool. VsX is a new and improved shorthand, nothing more. The only person who ever thought it made any loftier claims...is you.

 

[Canadiens1958]

Sad

How the system functions is quite simple. It is explained in a single post and quickly understandable. So yes, I do expect the people looking at the data to understand its limitations, as the are obvious with just a little bit of thought.

What you are doing is tantamount to criticizing the theory of conservation of energy because general relativity showed there was more going on. Of course there is more going on. Anyone with an understanding of hockey can perceive that, but it is impossible to model these factors with incomplete data.

The human mind can only hold so many variables in play at one time. We need shorthand methods to aid us in our judgments, and we have always had them. The old shorthand method was comparing top-X finishes on leaderboards, or comparing raw scoring or what have you...all considerably worse methods of evaluation across eras because they do not account for the changing size of the league or the talent pool. VsX is a new and improved shorthand, nothing more. The only person who ever thought it made any loftier claims...is you.

Sad to see that you suddenly have such a poor opinion of your own efforts. Properly used the VsX has a lot of untapped potential.

NHL, lower scouting has already factored in a lot of the so-called variables that you superficialize in their évaluations using basic grade school math.

 

[unknown33]

NHL, lower scouting has already factored in a lot of the so-called variables that you superficialize in their évaluations using basic grade school math.

You talk about it all the time. Would you mind show us those superduper complex evaluations?

 

[Sturminator]

NHL, lower scouting has already factored in a lot of the so-called variables that you superficialize in their évaluations using basic grade school math.

The data that is available to professional scouts today has not been available for very long, at all. We most certainly could model individual performance more efficiently for players competing today, but the same does not hold true for older generations. The data is simply not there. Too much more complexity in the system would also render it opaque to the casual observer and turn it into a "magic box" whose accuracy people would simply have to "trust" because they could not see all of the gears moving.

The system as it stands is universal, easy to understand, and it passes the smell test quite well. That it is less than perfect is not much of an objection. It succeeds at precisely what it sets out to do, and is a large improvement on what came before it. No, I don't think little of it; I indirectly compared myself to one of the great scientists who predated Einstein in the last post. {Mod}

If you wish to do the hard work of parsing every single box score from 1926 to today in order to tease out high-leverage performance in various game situations and then figure out a way to model that performance...go for it. {Mod}

 

[Canadiens1958]

Simple

You talk about it all the time. Would you mind show us those superduper complex evaluations?

Just applied percentages with team subjective weight given to each attribute.