All-in-one hockey value stats

[sjaustin77]

I would like to measure who is the best player and I don't think any of the systems do that. They all seem to measure some type of point contribution, point shares, or value toward wins. They are OK at doing that but I think there are obvious flaws in all of them especially if you are trying to measure who the best player is. I would also do forwards, D and goalies separately. I don't think any one number can accurately measure all three positions.

Here are some of the flaws that I see in addition to or reiterating what others have mentioned:

1. The value given to goals vs assists as has been pointed out. I think assists are undervalued even at the correct proportion. I see no reason to arbitrarily assign an assist .5 value of a goal instead of about .6. Forwards should probably be even higher.

2. Most (if any) don't take into account quality of competition. Without it you are lacking huge context.

3. Depending on the system not enough or too much weight is given to special teams and Shootouts. Very little to no weight should be given to shootouts. Weighting of special teams should be in proportion to their time on ice and contribution to goals. About 83% of ice time is at even strength. Only about 23% of goals are on the PP or SH. Plus/Minus a couple of % points in any given year.

This combined with point two give some very skewed results. Any systems that give players like Kaberle, Rafalski, and McQuaid more defensive value than Chara, etc are just plain wrong in my opinion.

4. Too complicated - only as good as the input. Systems should use concrete stats instead of advanced stats which may or may not be accurate. There are flaws with some of the advanced stats so you are starting with incorrect input data. I have problems with systems based on shots over production. Which means I don't really like some of the Corsi stats.

I would use a much more basic formula and tweak as necessary.

1. Start with overall production. Very good indicator of the best point producers.
2. Adjust for PP and PK time. Given equal production the player with more PK time and less PP time is usually the better player.
3. Adjust for Quality of Competition - which I would determine based more on production than any shots based or +/- stat.

Factor in some other stats weighted how you see fit to add defensive value. Maybe Zone starts, Blocked shots, Hits

I think this would give a very good list of the best players without getting too complicated or relying on flawed input.

 

[Iain Fyffe]

Still it is just for making a point, with no other information I take the guy with more goal than assists for the same point totals.

I'd prefer the more balanced player. To me, a player with 50G-30A probably needs a playmaker to be at his most effective, while someone with say 35G-45A plays a balanced game and is not dependent on having either someone to pass him the puck, or someone to pass the puck to.

The sedins are not as good together that a 200 goal scorer, because of lot of their points are on the same goal.

That's because they play on the same line. If Rick Nash was on that line, a lot of his points would be on the same goals as well. That's how lines work.

 

[Iain Fyffe]

I would like to measure who is the best player and I don't think any of the systems do that. They all seem to measure some type of point contribution, point shares, or value toward wins.

There's a reason for this: the best players, by definition, contribute the most towards a team winning. That's the point of hockey: to win hockey games. The players that contribute the most toward that are therefore the best players.

3. Depending on the system not enough or too much weight is given to special teams and Shootouts. Very little to no weight should be given to shootouts. Weighting of special teams should be in proportion to their time on ice and contribution to goals. About 83% of ice time is at even strength. Only about 23% of goals are on the PP or SH. Plus/Minus a couple of % points in any given year.

I agree with shootouts, but not for this reason. I tend to ignore them because they're generally a coin flip, and they involve a very limited subset of hockey skills.

It's very easy to get the weight for special teams correct, because we know how much time is spent on special teams, and how many goals are scored in each situation.

4. Too complicated - only as good as the input. Systems should use concrete stats instead of advanced stats which may or may not be accurate.

Sorry, but the "concrete" stats may or may not be accurate either. Goals scored may seem concrete, but the goals stat is as flawed as any other hockey stat.

1. Start with overall production. Very good indicator of the best point producers.
2. Adjust for PP and PK time. Given equal production the player with more PK time and less PP time is usually the better player.

If you're going to adjust for PP and PK time, why not just start with situational production in the first place? Don't start with points, start with ESP, PPP and SHP.

Factor in some other stats weighted how you see fit to add defensive value. Maybe Zone starts, Blocked shots, Hits

Ah, there's the rub. Just add something for defence. Sounds easy, doesn't it?

 

[MadLuke]

I'd prefer the more balanced player. To me, a player with 50G-30A probably needs a playmaker to be at his most effective, while someone with say 35G-45A plays a balanced game and is not dependent on having either someone to pass him the puck, or someone to pass the puck to.


That's because they play on the same line. If Rick Nash was on that line, a lot of his points would be on the same goals as well. That's how lines work.

Would you trade a unbalanced 180 goal scorer 0 assist for both Sedins ? They do about 180 points ? Not me.

 

[Iain Fyffe]

Would you trade a unbalanced 180 goal scorer 0 assist for both Sedins ? They do about 180 points ? Not me.

No, because one 180-point player is worth substantially more than two 90-point players, in general. Regardless of how said points were distributed.

But there's no such thing as a 180G-0A player, or anything even close to it. Try to keep the examples realistic if you expect to have a meaningful discussion.

 

[sjaustin77]

Thanks for responding. You have obviously done a lot of work on these things. Have you done rankings using your system for the past couple of years? I would be interested to see if it matches closely with mine (which is incomplete and unpublished - more just for fun) and GVT.

Off Topic - but I'm very interested in your TOI work for a few players. Is there a link that you can point me to that shows each year's ice time for Gretzky, Lemieux, Orr and Bourque? Or if not and you have that information could you PM me?

There's a reason for this: the best players, by definition, contribute the most towards a team winning. That's the point of hockey: to win hockey games. The players that contribute the most toward that are therefore the best players. Best player can have more than 1 meaning. I don't believe that best player necessarily equals contributing the most to winning. I think Crosby is clearly the best but because of injuries he hasn't contributed the most. I also think players like Kovalchuk get rated too high because of ridiculous ice time. I think there are a number of players who score less points that are better players. I also don't see the defense that some others see. He does play PK but I'm not sure he should.

Teams with depth, that roll 4 lines may have some better players than where the contribution or point shares rank them in comparison to players that get more ice time, powerplay time, more offensive zone starts and easier quality of competition. I think these things need to be accounted for in any good system.


I agree with shootouts, but not for this reason. I tend to ignore them because they're generally a coin flip, and they involve a very limited subset of hockey skills. They do contribute to winning in the regular season so I can see why some systems use them for that but as far as best player it does nothing for me. Like you said - coinflip, limited subset of skills and I'll add very small sample sizes with great variance year to year.

It's very easy to get the weight for special teams correct, because we know how much time is spent on special teams, and how many goals are scored in each situation. So maybe your system does this and I think that is great if it does. To me most systems don't weight things properly. We have all the data, it should be easy to give a better weighting.


Sorry, but the "concrete" stats may or may not be accurate either. Goals scored may seem concrete, but the goals stat is as flawed as any other hockey stat. Why do you say the goal stat is flawed? Maybe it is but I can't agree that it is as flawed as many of the advanced stats. Goals are scored in many different ways so they will have different values but I don't see how they aren't accurate or concrete. Maybe tangible is a better word. How are goals more flawed than using systems and formulas based on shots or adjusted +/-, and assigning arbitrary percentages of value, etc.? I think using advanced stats in formulas is much more flawed than using goals, assists, points, TOI, etc. Never going to get me to agree that goals is as flawed as some of these other stats that formulas use.


If you're going to adjust for PP and PK time, why not just start with situational production in the first place? Don't start with points, start with ESP, PPP and SHP. I could do that. One of my favorite stats for comparing players is ES Points/60. Production is just a quick look at the overall production number. I think taking the different rates and adding them up would give the overall production number. So I would give a preliminary ranking of players based on production and then they would move up or down with adjustments for the different situations and as I add defensive components.


Ah, there's the rub. Just add something for defence. Sounds easy, doesn't it?

It doesn't sound easy and I'm not saying that it is. Defense is hard to gauge with one number. I think I can compile a pretty good list based on the numbers - Ice time, SHTOI, +/-, Blocked Shots, Hits, Points scored against while on/off the ice, zone starts, etc. The hardest part is Quality of Competition. I'm not sure there is a number that I fully trust yet. I just think things can be done simpler and as good or better with the more concrete/tangible stats and by weighting things better.

I appreciate all the work others have put in. Much of it is beyond my scope of knowledge. Given all that work and presumably smarter people doing it than me I wonder why they choose to do some things the way they do. I'm obviously not the only one to point out what we think are obvious flaws.

GVT matches up very well with how I would rank the players so maybe that uses most of the stuff that I see as flaws in other systems. I had Mike Smith with the Vezina, Smith or Malkin as MVP, Bergeron with the Selke and Chara with the Norris. The only one that GVT has differently than me is Karlsson over Chara.

 

[MadLuke]

No, because one 180-point player is worth substantially more than two 90-point players, in general. Regardless of how said points were distributed.

But there's no such thing as a 180G-0A player, or anything even close to it. Try to keep the examples realistic if you expect to have a meaningful discussion.

You right, concentration is always better in offense.

What about Pavel Bure with 2 assist less vs Doug Weight in 00-01 ?

Or for a better systematic way to look at it:

http://www.hockey-reference.com/leaders/goals_created_season.html

Look the goal created leader of the 92-93 season, Mogilny and Selanne are above player with more points because of their goal > assist ?.

 

[Czech Your Math]

You right, concentration is always better in offense.

What about Pavel Bure with 2 assist less vs Doug Weight in 00-01 ?

Or for a better systematic way to look at it:

http://www.hockey-reference.com/leaders/goals_created_season.html

Look the goal created leader of the 92-93 season, Mogilny and Selanne are above player with more points because of their goal > assist ?.

Most would agree that goals are generally more valuable than assists, although it may not be the case for all players/teams.

The problem is that no one can agree how much more valuable they are. HR gives goals 2x value of assists, presumably because there are two assists possible on each goal. GVT uses 1.5x assists as a presumably arbitrary compromise. The actual ratio has varied between ~1.6 & 1.75 over the past 50+ seasons, generally drifting upwards over time.

At least everyone can agree that for each point by a player, a goal has been scored.

 

[MadLuke]

At least everyone can agree that for each point by a player, a goal has been scored.

I do not agree, each goal by a player a goal has been scored, not for each point in a way.

Add all players point on a team and compare it to the team goals total, we will see that a goal was not scored for each point by a player

 

[Czech Your Math]

I'm trying to refine a crude all-in-one metric using limited data. It does not use any actual ice time, nor any breakdown of points by situation, so it can be used for any season since expansion.

I had seen Point Allocation before, but recently looked at it again and found it very interesting, especially the assumptions made and the relative simplicity of the metric. Since Iain's PA listed his numbers for a couple of teams, so I decided to start with one of those, the '76 Canadiens. I guess I'll call it Total Value for now, for lack of a better term. I'd appreciate any feedback, whether positive or negative. I'm really interested in how the player values compare from player to player and position to position. I just included PA for fun. I'll do the same for the '01 Red Wings in the near future. TV1 & TV2 are just two very similar versions of the same system, so also curious if one appears better than the other.

PLAYER | Pos | TV1 | TV2 | PA
Lafleur | F | 13.0 | 13.2 | 12.5
MahovlichP | F | 11.4 | 11.3 | 10.2
Dryden | G | 11.3 | 11.2 | 11.8
Lapointe | D | 11.0 | 10.9 | 8.1
Shutt | F | 9.7 | 10.1 | 9.0
Savard | D | 7.9 | 7.6 | 7.1
Cournoyer | F | 7.2 | 7.4 | 6.5
Robinson | D | 7.0 | 6.9 | 5.4
Lambert | F | 6.1 | 6.2 | 8.2
Lemaire | F | 5.7 | 5.7 | 5.6
Risebrough | F | 4.4 | 4.4 | 5.9
Wilson | F | 4.0 | 4.0 | 4.0
Awrey | D | 4.0 | 3.9 | 5.2
Jarvis | F | 4.0 | 3.8 | 4.6
Gainey | F | 3.7 | 3.8 | 5.1
Larocque | G | 3.7 | 3.6 | 3.1
Bouchard | D | 2.9 | 2.9 | 4.1
Van Boxmeer | D | 2.5 | 2.5 | 1.5
Tremblay | F | 2.4 | 2.4 | 3.7
Roberts | F | 2.4 | 2.5 | 3.7
Nyrop | D | 1.7 | 1.7 | 1.4
Chartraw | D | 1.1 | 1.1 | 0.8
Goldup | F | (0.0) | (0.0) | 0.1
Shanahan | F | (0.0) | (0.0) | 0.1

 

[TheDevilMadeMe]

Someone like Gainey is going to be vastly underrated by a stat like this, unless it includes quality of competition as a factor. And good luck calculating QComp for historical players

 

[Czech Your Math]

Someone like Gainey is going to be vastly underrated by a stat like this, unless it includes quality of competition as a factor. And good luck calculating QComp for historical players

Thanks for the feedback, you are probably right. The only way to increase his value would be giving extra weight for involvement in the PK (as an indicator of defensive ability), and I will look into doing so after first calculating the values for the '01 Wings.

 

[Czech Your Math]

Here is the metric for the 2001 Wings:

PLAYER | Pos | TV1 | TV2 | PA
Lidstrom | D | 11.4 | 11.0 | 9.1
Fedorov | F | 9.6 | 9.8 | 8.8
Shanahan | F | 9.4 | 9.5 | 8.4
Lapointe | F | 7.0 | 7.2 | 6.8
Yzerman | F | 6.9 | 6.8 | 6.6
Osgood | G | 6.4 | 6.4 | 4.7
Legace | G | 5.8 | 5.7 | 4.8
Kozlov | F | 4.8 | 5.0 | 4.9
Dandenault | D | 4.4 | 4.4 | 3.0
Holmstrom | F | 4.2 | 4.3 | 3.7
Duchesne | D | 4.2 | 4.1 | 2.9
Draper | F | 3.8 | 3.8 | 4.2
Gill | D | 3.8 | 3.8 | 4.2
Verbeek | F | 3.6 | 3.8 | 4.1
Maltby | F | 3.6 | 3.7 | 4.3
Larionov | F | 3.3 | 3.1 | 3.6
Murphy | D | 2.8 | 2.6 | 3.3
Brown | F | 2.7 | 2.7 | 2.8
McCarty | F | 2.4 | 2.6 | 4.1
Ward | D | 2.0 | 2.0 | 4.1
Fischer | D | 2.0 | 1.9 | 2.9
Devereaux | F | 1.3 | 1.4 | 2.1
Chelios | D | 1.3 | 1.3 | 1.9
Gilchrist | F | 0.9 | 0.8 | 2.6
Williams | F | 0.4 | 0.4 | 0.3
Kuznetsov | D | 0.4 | 0.4 | 0.5
Butsayev | F | 0.1 | 0.1 | 0.5

I'm going to try to refine it by using positional adjustments, as I don't see any other way without making assumptions based on knowledge of the players.

 

[wgknestrick]

GVT is about the best current all around metric IMO, but as others have said, it has some room for improvement.

PROS:
-Metric has tangible unit (goals). Not an arbitrary EA Sports like "rating" that means little.
-Adjusts across different years


CONS:
-Player positions should be roughly equal in average GVT, but Goalies run away with highest (and lowest) GVT every year. This is a huge indicator of a problem with the formulas. Defenders should be able to nullify a forward (logic test), but GVT does not indicate that. For every forward with 20+ GVT, I'd expect to see a defender with that high of a value. Chara/Webber can and will shut down Malkin/Crosby/Stamkos etc. Why don't their GVTs equal out?

Example from 2012 season:
Forwards: GVTave 3.81, GVTsd 5.49, GVTmax 33.5, GVTmin -5.8
Defense: GVTave 3.69, GVTsd 4.01, GVTmax 21.5, GVTmin -2.1
Goaltenders: GVTave 3.55, GVTsd 10.56, GVTmax 35, GVTmin -22

Does anyone expect to have that much weight applied to goaltending? It seems that defenders are not being credited enough with regards to how high their goaltender's SV% is (and vice-versa).

GVTave is about equal among positions, but we see the extremes (goaltender GVT) get out of control. This is a BIG indicator for "room for improvement".

-Doesn't break contributions up by EV, PP, and SH. I think this is critical to getting more accurate numbers and future projections.

-Does not take Qual Comp into equations. Calcs lack context without this. This is pretty easy to include to the dGVT part of the calcs.

 

[GuineaPig]

I think GVT correctly weights goalies. They're most definitely (and disproportionately so) the most influential position.

 

[wgknestrick]

Can you explain Mike Smith's move from TBL to PHX and the results?

TOR "seeming" horrible luck with goaltending (no matter who is actually in net)?

BOS always fielding high SV% with their goaltending?

A look into above team's #1 and #2 defenders tells you all you need to know, and to a lessor extend, team D systems. I am just skeptical of the GVT for goaltenders, given the results.



I would logically expect GVT to be roughly equal across all positions. While a goaltender is certainly unique relative to the other positions, I do not put him up at quarterback level with respect to contribution. The team has much more control of what happens in front of him than you think. The goaltender does not touch the puck or influence play as much as the skaters do, yet we think they contribute more to outcome of games? Sure, they are always on the ice, but are they always "in play"?

I just don't see how any goalie could cost his team -22 goals under a "replacement" goal tender and consistently be in the league. If a forward or defender was ever that bad, we'd never hear from him again. I just don't agree with the extremes of the gGVT ends because goaltenders are not consistently the "stars" of the NHL. Goaltenders held 6/10 of the top GVTs last year. Defenders held 0. Forwards 4. I also doubt we are in such an odd era that is blessed with the finest Goaltenders the NHL has ever seen (and that Brian Elliott is one of them?).

The game surely does not revolve around them the way GVT would leave you to believe.

Do you think elite defenders are as about as valuable as an average first line offender or mid tier goalie? GVT would have you believe that, but salaries don't reflect this and teams certainly aren't hoarding offense in favor of top defenders.

 

[mindmasher]

CONS:
-Player positions should be roughly equal in average GVT, but Goalies run away with highest (and lowest) GVT every year. This is a huge indicator of a problem with the formulas. Defenders should be able to nullify a forward (logic test), but GVT does not indicate that. For every forward with 20+ GVT, I'd expect to see a defender with that high of a value. Chara/Webber can and will shut down Malkin/Crosby/Stamkos etc. Why don't their GVTs equal out?

No. I have no idea where you get this idea that a left winger who plays 15 minutes a game is worth the same as a goalie who plays 60 minutes a game. Simply isn't true.

 

[mindmasher]

Do you think elite defenders are as about as valuable as an average first line offender or mid tier goalie? GVT would have you believe that, but salaries don't reflect this and teams certainly aren't hoarding offense in favor of top defenders.

No. A goalie has a much bigger impact on a teams results than any other player, and this is easily demonstrable. Virtually all defensive zone shots are modified by a goalies ability. Defenceman and forwards impact it, but virtually all events are modified to the positive and negative based on the goalies play. Even the biggest minutes #1 defender who can influence shot distance and location is nowhere near the level of influencing EVERY shot result. Even simple stuff like the length of rebound or which corner they direct pucks to impacts a teams bottom defensive line.

 

[Bear of Bad News]

No. A goalie has a much bigger impact on a teams results than any other player, and this is easily demonstrable. Virtually all defensive zone shots are modified by a goalies ability. Defenceman and forwards impact it, but virtually all events are modified to the positive and negative based on the goalies play. Even the biggest minutes #1 defender who can influence shot distance and location is nowhere near the level of influencing EVERY shot result. Even simple stuff like the length of rebound or which corner they direct pucks to impacts a teams bottom defensive line.

I'll play devil's advocate (since I'm widely known as the biggest "goalie homer" on the board).

Suppose that goaltenders have much more influence on the game than either forwards and defensemen (and let's suppose that it isn't even close). That's different than what's being considered here, which is "how much more value does Goaltender X bring to a team than a replacement-level player?".

If we pictured a world where goaltenders had the most influence on the game, but where all goaltenders were identical, then Goaltender X would have no value (because you could go down to the goalie store and get another one just like him).

To summarize: the question here isn't whether or not a goaltender has value. The question is whether or not he has more valuable than a team's next option.

 

[wgknestrick]

No. I have no idea where you get this idea that a left winger who plays 15 minutes a game is worth the same as a goalie who plays 60 minutes a game. Simply isn't true.

I think you are taking my words out of context and applying your own. I'd expect the average forward's GVT to be roughly equal equal to the average goaltender's or Defenders. This is almost the case already with GVT. It's the extremes that I have a problem with. Everything can look unequal on an individual basis.

Also of note: Do you think a goaltender is "In play" for 60min? Sure, they are "on ice", but their ice time is not relevant to a skater's IMO. The puck is not always in the defensive zone where the goaltender has influence on the play. I still fail to see the inherit "godly" advantage that goaltenders have over the skating peasants that play in front of them that GVT tells us.

Hockey's goalies are unlike football's QBs or a baseball pitchers in that the game does not guarantee them touches on the game object due to the rules.

 

[wgknestrick]

GVT tells us that 6 out of the top 10 NHL players are goaltenders. Do you believe this to be true?

0 out of 10 are defenders. Do you believe this to be true?

These types of GVT results are very similar year after year. I am just suggesting that goalies aren't as valuable as GVT leads us to believe and that some of that misplaced GVT "value" belongs to the defenders.

This is why Rask puts up similar SV% to Thomas. The goaltending talent does not technically reside with them, but in Chara and BOS's system. Hell, MAF could probably have a good SV% in Boston.

 

[mindmasher]

I'll play devil's advocate (since I'm widely known as the biggest "goalie homer" on the board).

Suppose that goaltenders have much more influence on the game than either forwards and defensemen (and let's suppose that it isn't even close). That's different than what's being considered here, which is "how much more value does Goaltender X bring to a team than a replacement-level player?".

If we pictured a world where goaltenders had the most influence on the game, but where all goaltenders were identical, then Goaltender X would have no value (because you could go down to the goalie store and get another one just like him).

To summarize: the question here isn't whether or not a goaltender has value. The question is whether or not he has more valuable than a team's next option.

Well Taco, it's funny you bring this up. I think there are a lot of smart people - Dejardins and Dellow for instance - who have argued repeatedly that unless you are getting an elite level goalie, save your money because performance variations amongst middle tier goalies is negligible.

In other words, goaltenders are very important, but tend to be very close in ability. Therefore unless you have a bead on a Tim Thomas type, you should be spending small dollars on a couple of capable NHL starters who simply don't have many teams to play for.

In another way, goaltenders can also be a considerable source of lost points on a team if you deign to start a guy who is performing considerably below replacement level.

So yes, ultimately, goaltenders have little value in that there are a lot of good cheap options. You see some GM's following this principle closely. Detroit, Chicago, Avs, Philly, Phx, etc.

 

[mindmasher]

I think you are taking my words out of context and applying your own. I'd expect the average forward's GVT to be roughly equal equal to the average goaltender's or Defenders. This is almost the case already with GVT. It's the extremes that I have a problem with. Everything can look unequal on an individual basis.

Also of note: Do you think a goaltender is "In play" for 60min? Sure, they are "on ice", but their ice time is not relevant to a skater's IMO. The puck is not always in the defensive zone where the goaltender has influence on the play. I still fail to see the inherit "godly" advantage that goaltenders have over the skating peasants that play in front of them that GVT tells us.

Hockey's goalies are unlike football's QBs or a baseball pitchers in that the game does not guarantee them touches on the game object due to the rules.

In the same sense players aren't -in play- for all of their ice time either. As a former player of both positions (goaltender and Dman) I know this to be true from (for my own experience at least).

 

[mindmasher]

GVT tells us that 6 out of the top 10 NHL players are goaltenders. Do you believe this to be true?

0 out of 10 are defenders. Do you believe this to be true?

These types of GVT results are very similar year after year. I am just suggesting that goalies aren't as valuable as GVT leads us to believe and that some of that misplaced GVT "value" belongs to the defenders.

This is why Rask puts up similar SV% to Thomas. The goaltending talent does not technically reside with them, but in Chara and BOS's system. Hell, MAF could probably have a good SV% in Boston.

A team defensive structure will have some impact on a goalie, this is obvious. But how efficient a goalie is under that system - in my opinion - affects goals allowed at a rate that is much greater than most forwards or defencemen could hope to achieve on either side of the ice.

So yes, in terms of affecting point standings, I do believe that 6 of the top 10 'influencers' are goaltenders.

 

[mindmasher]

This is why Rask puts up similar SV% to Thomas. The goaltending talent does not technically reside with them, but in Chara and BOS's system. Hell, MAF could probably have a good SV% in Boston.

I think Rask puts up a similar % to Thomas because he is a similarly great goaltender.

Don't forget, we can look at goaltenders allowed goals rates based on the shot volume they face, and if you suppose that quality and distance evens out over the whole season for most goaltenders, if they allow considerably less goals than expected - regardless of the D system in front of them - you can make an argument they are skilled goaltenders and are positively influencing results.