2017-18 stats and underlying metrics thread [Mod: updated season]

[Rheged]

Well, yeah, it's probably a consequence of playing Lowry at 3C for the lack of a better option... or is it?

Matthias-Lowry-Armia 44.8GF%, 2.12 GF60 was among the best lines we've seen.

Petan 2015-17 38.5GF% 1.26 GF60
Petan w/o Thorburn, Peluso 2015-17 40.0GF% 1.45 GF60
Petan w/o Thorburn, Peluso, Burmistrov, Stafford 2015-17 45.5GF% 1.83 GF60

Copp 2015-17 60.6GF% 2.12 GF60
Copp w/o Thorburn, Peluso 2015-17 63.3GF% 2.35 GF60
Copp w/o Thorburn, Peluso, Burmistrov, Stafford 2015-17 70.0GF% 2.51 GF60

Maybe try Copp at 3C? It can't get worse than X-Lowry-X.
Petan might have been "thorburnt", but it's not like he's had good results.

In a dream world I'd like to see them try to use Wheeler to carry a line with some of the bubble offensive guys on it. Try something silly like Dano-Petan/Copp-Wheeler for a few games at 5 on 5 and if they can hold their heads above water you've got 3 really good offensive lines to go along with your X-Lowry-Armia defensive line.

I guess the good thing this season, potentially anyways, is that we don't seem to have any 'traditional' 4th liners other than Tanev really.. Maurice may be forced to dress 3 skilled lines by default.

 

[Mortimer Snerd]

In a dream world I'd like to see them try to use Wheeler to carry a line with some of the bubble offensive guys on it. Try something silly like Dano-Petan/Copp-Wheeler for a few games at 5 on 5 and if they can hold their heads above water you've got 3 really good offensive lines to go along with your X-Lowry-Armia defensive line.

I guess the good thing this season, potentially anyways, is that we don't seem to have any 'traditional' 4th liners other than Tanev really.. Maurice may be forced to dress 3 skilled lines by default.

We could then have Perreault-Little-Armia. I think Armia's offense when played with those linemates would surprise more than a few people.

I don't see these as serious suggestions to get the best possible lineup but more like experiments to learn what we really have in some players.

 

[lanky]

PLA would be a fantastic trio at both ends of the ice. I agree that Armia could feast with those linemates.

 

[garret9]

And the teams having their own member do their research can probably be even useful in some ways, but I think it is still more a precaution for them not to loose any information that the other teams might get. Doesn't really prove much to me about the real importance and need for advanced stats.

Finnish coaches don't usually care much for Corsi and stuff like that, and still we have had good success in international hockey, especially compared to our population. Finnish coaches are mostly into estimating the quality of the scoring chances, which in fact I do see having at least some value compared to a completely useless stat like Corsi for example.[mod]

And hockey Canada uses.... wait for it... me!

Most coaches don't care for stats because they are he last to evolve, typically speaking. That's speaking of all most sports and whatnot.

That said, I can tell you full well that the evolution is happening, and everyone here should really just relax and enjoy being able to get a bit of inside look from myself.

 

[grieves]

(This was in the Laine thread but was moved.)

As for advanced stats, I think they are absolutely useful and anyone claiming otherwise needs to get informed. Absolutely everything in this world can be quantified if we would have a way to obtain ALL (and I mean absolutely everything) of the data (which is impossible).

Advanced stats are just sucking in as much data as they can and since we are talking about a "fairly simple" game (which still means there is ridiculous amounts of different things to take into account), advanced stats can absolutely be beneficial in this context.

The problem is the same as with life, there is no way to get ALL of the data, so blindly trusting a model is incorrect, as garrett said. But even the models are getting better and better with time.

The problem is when the model fails to include a key piece or key pieces of data pertaining to a specific case. In this case we can use it to see what an absolute hurricane McDavid is developing into, but we can't use it to compare McDavid with Laine because players reach adulthood at an individual pace. Individuality is poison for predictive models. What we can do is use our eyes and Laine's trainer's comments to conclude that they are not on equal footing when it comes to their development, and therefore conclude that any meaningful statistical comparison between the two should have a huge asterisk beside it. Also, the rink size adjustment and fatigue factor are not included to boot. Playing half a season is not even remotely the same than playing a full season for an 18yo (although Laine is the only one who played a pure 18yo season anyway). It is also difficult to quantify the effects of Laine's mid-season injury (did he get extra rest, was there some sort of emotional trauma etc).

Another thing that could be in Laine's favor is him being goal-heavy whereas McDavid is assist-heavy.

This is a topic I have been interested in but so far the only thing I have to go on is this:

http://hockeyanalytics.com/2006/10/goals-created/

This model seems to value goal scoring a lot more than setting up. It was interesting that a 20 goal player on average clears a bigger paycheck than a 40 assist player so at least within the teams goal scorers seem to be valued very highly. GM's seem to want goals a lot more than assists.

But I have no idea about the validity of this model, so feel free to rip it apart.

 

[Ippenator]

And hockey Canada uses.... wait for it... me!

Most coaches don't care for stats because they are he last to evolve, typically speaking. That's speaking of all most sports and whatnot.

That said, I can tell you full well that the evolution is happening, and everyone here should really just relax and enjoy being able to get a bit of inside look from myself.

So like you write here the coaches know nothing and you know obviously everything and how hockey's future will be? Oh please. Seems that you are a person way into statistics and thinking that the general behaviour of the masses decides everything in the world. Well I believe exactly the opposite way. The world in general and of course sports as a part of it consists of individuals with very different qualities. There might be mass behaviour in some issues and statistical averages can be calculated of practically everything. But those things are absolutely meaningless when we are interacting with individuals and when we have knowledge of the individuals strengths, weaknesses and their potential.

Statistical averages are for people wanting to make guesses in a lazy way instead of gathering more of the essential information on the individual. You can't even make real estimations on the basis of the advanced stats, as they are just some generalizations that in fact almost never apply for individuals in an an accurate way.

I also don't understand why some people have such a need to try to predict so many things. Why do you need to so much try to get info on what could happen in the future? Isn't it enough to have the past and present and give credit to individual's good results and just accept it that none of us can really predict the future. So no need to dig up tons of inaccurate and mostly even irrelevant nitpicking numbers.

I at least enjoy hockey by mostly watching it and don't care much for stats, except the real tangible result based stats that can show us who really had a good season and who didn't. I don't care how much shots and how good shot suppression a player has if the other player scored more goals and points and had a better +/- stat. Then for sure that player had the better season, and no other kind of a stats twisting will change that fact for me.

 

[Constable]

So like you write here the coaches know nothing and you know obviously everything and how hockey's future will be? Oh please. Seems that you are a person way into statistics and thinking that the general behaviour of the masses decides everything in the world. Well I believe exactly the opposite way. The world in general and of course sports as a part of it consists of individuals with very different qualities. There might be mass behaviour in some issues and statistical averages can be calculated of practically everything. But those things are absolutely meaningless when we are interacting with individuals and when we have knowledge of the individuals strengths, weaknesses and their potential.

Statistical averages are for people wanting to make guesses in a lazy way instead of gathering more of the essential information on the individual. You can't even make real estimations on the basis of the advanced stats, as they are just some generalizations that in fact almost never apply for individuals in an an accurate way.

I also don't understand why some people have such a need to try to predict so many things. Why do you need to so much try to get info on what could happen in the future? Isn't it enough to have the past and present and give credit to individual's good results and just accept it that none of us can really predict the future. So no need to dig up tons of inaccurate and mostly even irrelevant nitpicking numbers.

I at least enjoy hockey by mostly watching it and don't care much for stats, except the real tangible result based stats that can show us who really had a good season and who didn't. I don't care how much shots and how good shot suppression a player has if the other player scored more goals and points and had a better +/- stat. Then for sure that player had the better season, and no other kind of a stats twisting will change that fact for me.

(something about +/- and invalid opinions)

 

[Grind]

My only question for you ipp, would be why do you care about who had a good game in the first place?

The reason "most" people care about who had a good game or what happened last game/week/etc is because we assume having a good game means it's likely an indicator that they'll have good games in the future.

I mean, if you don't care about looking at even boxcar stats as an indicator of who's good/bad/played well (and thus will continue to do so in the future)...why look at them at all?

What is the root reason of even paying attention to them if it's not to determine who was good (and will continue to be good?)

 

[Ippenator]

(something about +/- and invalid opinions)

[mod]

Even as +/- stat by itself is also a pretty vulnerable stat, it still tells much more about the positive effect of a player to the real results which are scored and allowed goals. Sure you should not use it by itself as a measure of how good a player is compared to another. Also the but issue here is that the smaller the simple size, the more meaningless +/- stat will be. It for a players whole career it does already have a pretty valid base of showing if the player had a more positive than negative effect on his team's goal differences.

Just think of it. In the long run, does it really even matter how the player managed to get a very positive +/- stat during several seasons, if he just clearly has been on the ice much more so that his team has scored than when they have allowed goals. Sure the teammates have affected things positively also then. But still the more there are played games, the less you can give credit just for the teammates.

Even as this you can't really use the +/- to compare different player's whole careers effectively. But at least you can clearly define if a player has had a clear positive or negative effect on his teams scoring. And even during a full season it does already tell at least something about the players contribution to his team's goal difference. Just with as small sample size as a season you should still look a bit more behind the stats, and get to know how much the player played PP or PK, and what kind of line combinations he was usually playing with.

Shooting and shot supression by themselves tell really absolutely nothing about good end results. For using and interpreting those stats's meaning you will need much more and more detailed knowledge of what different attributes, variables and affecting issues had still an effect on getting real results. With those stats you don't even have any other results than the amount of shot's for both teams. So those are the results that you can predict more accurately with those stats - how much a player might contribute to his team shooting the puck and how much the opponent might shoot when he is on the ice. Useless information by itself, as most of the shots could be bad quality and by that lead into losing the possesion to the opponent. The line could be still very affective with their defending and forechecking, but playing simple mindless and noncreative hockey and achieve great Corsi numbers, because with following Corsi and shot supression and trying to better those stats, the results you get are good amount ir shots and good shot supression. With the right kind of skill and play style you really don't need these kinds of quantities, but you can concentrate on creating quality shots and scoring chances and having the real puck pissession high, but not needing to necessarily have high quantities of shots.

Just think think about it a bit more. It should really make sense when you think about it and understand better how causality works even in hockey.

 

[YWGinYYZ]

A couple articles on the ills of +/-, as this is what Mason the Goalie was referring to (and has been discussed on HFJets a number of times over the last 4+ years):

http://edmontonjournal.com/sports/h...isleading-and-irrelevant-stat-for-nhl-players
https://www.arcticicehockey.com/2014/6/5/5602668/why-plus-minus-is-the-worst-statistic-in-hockey

 

[Ippenator]

My only question for you ipp, would be why do you care about who had a good game in the first place?

The reason "most" people care about who had a good game or what happened last game/week/etc is because we assume having a good game means it's likely an indicator that they'll have good games in the future.

I mean, if you don't care about looking at even boxcar stats as an indicator of who's good/bad/played well (and thus will continue to do so in the future)...why look at them at all?

What is the root reason of even paying attention to them if it's not to determine who was good (and will continue to be good?)

Simple answer. Everything even in hockey is about individuals and I'm very interested in what kind of results different individuals achieve. I can also have personal reasons to like certain players which can make me root for them and hope for them to score well and have good end results.

I don't really care other than for the excitment in the game and some esthetic reasons how they after all achieve their results. But still what only truly matters in the end is do they get the good and relevant end results (goals, assists, points, positive goal difference and wins). If they get consistently those good and relevant end results, they are the real best players. If they get good results in other more irrelevant areas of hockey, but not so good relevant results, they just are not as good players as those who get the results. Sure they might some day even be on the same level as the players achieving great end results. But until they clearly show that they have the needed skills to really achieve the relevant end results. Until they are able to do so, they are just maybe promising, but nothing more than that. Even how much they can shoot the puck at the opponent's goalie.

 

[Grind]

Simple answer. Everything even in hockey is about individuals and I'm very interested in what kind of results different individuals achieve. I can also have personal reasons to like certain players which can make me root for them and hope for them to score well and have good end results.

I don't really care other than for the excitment in the game and some esthetic reasons how they after all achieve their results. But still what only truly matters in the end is do they get the good and relevant end results (goals, assists, points, positive goal difference and wins). If they get consistently those good and relevant end results, they are the real best players. If they get good results in other more irrelevant areas of hockey, but not so good relevant results, they just are not as good players as those who get the results. Sure they might some day even be on the same level as the players achieving great end results. But until they clearly show that they have the needed skills to really achieve the relevant end results. Until they are able to do so, they are just maybe promising, but nothing more than that. Even how much they can shoot the puck at the opponent's goalie.

Ok. But here's the thing. Why do you care who got the good relevant need results? Is it because those players are likely to produce the goood, relevant end results more frequently?

 

[Ippenator]

Ok. But here's the thing. Why do you care who got the good relevant need results? Is it because those players are likely to produce the goood, relevant end results more frequently?

I did already explain the thing. But I will still try to clarify it a bit more. It is the end results only that matter. Even when we have at least somekind of longer scoring history, some deeper knowledge of the player's real potential and abilities and also knowledge of his team's and teammate's strengths and weaknesses, we can't really make any reliable predictions on the player's future good end results. But with the knowledge that I pointed out before, we can at least define if the player is good at scoring and getting good end results at the moment.

With following shooting volumes we only get to know which players are good at shooting the puck a lot. To know if they are really good at scoring goals we simply need to know how many goals they have scored, and preferably with a large enough sample size.

I only need to know which players are good at the moment. And I enjoy the results they achieve with the past and with the present. I would rather do estimations and speculations based on eye test and more detailed knowledge on the individuals than predictions with irrelevant stats.

 

[lomiller1]

And hockey Canada uses.... wait for it... me!

Most coaches don't care for stats because they are he last to evolve, typically speaking. That's speaking of all most sports and whatnot.

That said, I can tell you full well that the evolution is happening, and everyone here should really just relax and enjoy being able to get a bit of inside look from myself.

The thing I find interesting about this is that counting shot attempts should be the most natural thing in the world for coaches. Their go-to when their team is struggling is invariably to shoot more. The only shots they don’t like are the “one and done†shots that allow for easy recover and zone exit, and these have big opportunity costs tied to them that ultimately hurts your overall Corsi.

I wonder if it isn’t just a case of ignoring the obvious because it looks to easy. I think you could make a similar case for scouts ignoring players productivity because it seems too easy.

 

[lomiller1]

So like you write here the coaches know nothing and you know obviously everything and how hockey's future will be? Oh please. Seems that you are a person way into statistics and thinking that the general behaviour of the masses decides everything in the world. Well I believe exactly the opposite way. The world in general and of course sports as a part of it consists of individuals with very different qualities. There might be mass behaviour in some issues and statistical averages can be calculated of practically everything. But those things are absolutely meaningless when we are interacting with individuals and when we have knowledge of the individuals strengths, weaknesses and their potential.

Statistical averages are for people wanting to make guesses in a lazy way instead of gathering more of the essential information on the individual. You can't even make real estimations on the basis of the advanced stats, as they are just some generalizations that in fact almost never apply for individuals in an an accurate way.

I also don't understand why some people have such a need to try to predict so many things. Why do you need to so much try to get info on what could happen in the future? Isn't it enough to have the past and present and give credit to individual's good results and just accept it that none of us can really predict the future. So no need to dig up tons of inaccurate and mostly even irrelevant nitpicking numbers.

I at least enjoy hockey by mostly watching it and don't care much for stats, except the real tangible result based stats that can show us who really had a good season and who didn't. I don't care how much shots and how good shot suppression a player has if the other player scored more goals and points and had a better +/- stat. Then for sure that player had the better season, and no other kind of a stats twisting will change that fact for me.

This is about as wrong as you can get. Properly formulated statistics is the ONLY way you can get meaningful data. This isn’t just a hockey thing it’s an everywhere thing.

 

[garret9]

Not even coaches think +/- is useful.


Almost all the entire hockey management world, whether numbers folks or not, find +/- too inherently flawed to be even useful sometimes.

All models are flawed, but some are useful.
All eye tests are flawed, but some are useful.

+/- is one that is not useful.

If you'ld like to know more:
Why Plus/Minus is the worst statistic in hockey and should be abolished

 

[lomiller1]

I did already explain the thing. But I will still try to clarify it a bit more. It is the end results only that matter. Even when we have at least somekind of longer scoring history, some deeper knowledge of the player's real potential and abilities and also knowledge of his team's and teammate's strengths and weaknesses, we can't really make any reliable predictions on the player's future good end results.

This is contradictory. Shot attempts ARE results in their own right. While they are not exactly the result you ultimately want to know about they are the best indicator of what that result will be.

In hockey speak doing the right things is the best way to get the results you want. If you are doing the right things, but things don’t go your way you keep doing what you are doing. If you are not doing the right things, even if things are going well now, it will come back to bite you.

What the math is telling us is that the best way to measure which players and teams are doing the right things to have success is to look at team shot attempts.

 

[garret9]

This is contradictory. Shot attempts ARE results in their own right. While they are not exactly the result you ultimately want to know about they are the best indicator of what that result will be.

In hockey speak doing the right things is the best way to get the results you want. If you are doing the right things, but things donÂ’t go your way you keep doing what you are doing. If you are not doing the right things, even if things are going well now, it will come back to bite you.

What the math is telling us is that the best way to measure which players and teams are doing the right things to have success is to look at team shot attempts.

Just to comment, shot attempts is not the best way... it's just better than goals. We have stuff better than shot attempts, and we'll eventually have stuff better than that.


+/- < goals for % < Corsi for % < adjusted Corsi % < xGoals < XPM < something in the future that is even better

 

[Ippenator]

This is about as wrong as you can get. Properly formulated statistics is the ONLY way you can get meaningful data. This isn’t just a hockey thing it’s an everywhere thing.

It is, but only if you really want and need to study these averages that have to do with bigger masses and some scientific studies about general volume based scientific issues. Individuals and their success has nothing to do with it. Only things that matter are the individual's and his team's abilities and the different variables that have straightforwardly to do with how effective that individual is in achieving good end results.

I'm sorry to say but you are trying to bash my well based and detailed points with just pointing out practically with saying "no you are wrong, because what I say now is right!" But doing this without giving out any well based and detailed points of where I really went wrong and why. In fact what you pointed out is exactly wrong when discussing individuals and estimating certain individual's success.

 

[Grind]

This is contradictory. Shot attempts ARE results in their own right. While they are not exactly the result you ultimately want to know about they are the best indicator of what that result will be.

In hockey speak doing the right things is the best way to get the results you want. If you are doing the right things, but things don’t go your way you keep doing what you are doing. If you are not doing the right things, even if things are going well now, it will come back to bite you.

What the math is telling us is that the best way to measure which players and teams are doing the right things to have success is to look at team shot attempts.

Exactly what I was building to. Even at the moment, the player influencing the shot attempts is the one who probably IS performing better. It's just taking things like luck out of the equation.

He's not performing better because he's getting more shot attempts and that's what you want. He's getting more shot attempts BECAUSE he's performing better.

We don't care about the attempts, we care about what they indicate/correlate too.

An example would be if I hear the boom of thunder near by I know lightning has struck. The thunder isn't dangerous, it's not burning anything, but it's what I can track to indicate lightning (which is dangerous and setting fires) is striking without seeing it.


Maybe that's not where the disconnect is ip. Maybe you just really don't care about the impacts of players game to game.

But the above seems to be here the primary issue with most who "don't like advanced stats" lies. They misinterpret stats guys as saying what their tracking is what's important and what players should be trying to do, not that they are what's indicating the effect the player is having.

 

[lomiller1]

Just to comment, shot attempts is not the best way... it's just better than goals. We have stuff better than shot attempts, and we'll eventually have stuff better than that.

I’d suspect the better ways you refer to are just additional data on each shot attempt. I still classify this as looking at shot attempts.

 

[Grind]

Just to comment, shot attempts is not the best way... it's just better than goals. We have stuff better than shot attempts, and we'll eventually have stuff better than that.


+/- < goals for % < Corsi for % < adjusted Corsi % < xGoals < XPM < something in the future that is even better

For sure, but I think the primary disconnect carries, which tends to be people getting hung up on corsi/Fenwick/xpm being something other then just goals and that stats guys think these "non goal events" are the end result you want. Their not. We still want the end result of goals. It's iust that what leads to goals creates multiple by products (of which goals are one of) that better predict future goals then just the goals themselves.

Ie: you don't care about xpm because you win a hockey game when you get to 100 xpm, you care about xpm because it's an indicator that your going to score more goals then you allow, which is the only real goal.

 

[lomiller1]

It is, but only if you really want and need to study these averages that have to do with bigger masses and some scientific studies about general volume based scientific issues. Individuals and their success has nothing to do with it. Only things that matter are the individual's and his team's abilities and the different variables that have straightforwardly to do with how effective that individual is in achieving good end results.

Individual success can come for doing things likely to result in success or from luck, but only one of those can be controlled and planned for. For example, my retirement plan is not built around buying lotter tickets and hoping I hit the jackpot. There will invariably be some people that do win the lottery and do better than I will, but I will do far better using sound investment strategies than the vast majority of people gambling on lottery tickets.

The same principle applies in hockey and just about everywhere else. Planning for success and doing the right things to accomplish that plan will lead to a much higher level of success in the long term than trying any old thing and hoping you get lucky.

 

[Ippenator]

I guess it's no use to keep arguing, as this is anyways a pretty much philosophical argument after all. I just am a person who values very much individualism and the differences, strengths and weaknesses that humans can have. All this creates intriguing "stories" and interaction in everything that humans do. Hockey is and even sports are in general are practically individualism used for collective causes in very exciting ways. I just don't see sports benefitting from getting too scientific and "robotized".

As I have said, I'm not really in general interested in advanced stats, but if you guys know if there are any stats about the quality of the shot attempts that NHL players have made for a couple of last seasons, that could be in fact pretty interesting stats for me too. I'm not sure of the predictive value that even those stats would have. But it would in my opinion show at least more about the skills for scoring that each players have.

 

[garret9]

IÂ’d suspect the better ways you refer to are just additional data on each shot attempt. I still classify this as looking at shot attempts.

Yes and no.

GAR introduces things that are not shot attempts, like FOs, Penalties, etc. that impact the game too.

I understand the notion of classifying xGoals or XPM as shot attempts, although personally I prefer shot metrics.