Advanced Hockey Stats Discussion

[MSP4LYFE]

I am a proponent of advanced stats as a way to further my understanding of teams and players, but don't use them to compare player A to player B or to make confident predictions on how a player or team will perform. I think they can be used to identify trends and can be used as a strong predictor for the future, but there are too many factors in play to make concrete predictions and have confidence it will come true.

"Personally I still struggle with the translation of shot attempts (w/ or w/o blocks) to be a good measure of 'possession' "

This was something that I struggled with too since logically to me shot attempts doesn't seem like it should be a proxy for possession. I can't find the article right now, but I read something where someone watched a bunch of games with a stop watch and measured the amount of time teams possessed the puck in the attacking zone. This was then compared to the shot attempt metrics for those games and they lined up showing that those stats did seem to be a perfect proxy for determining which team controlled possession.

Good idea for a thread Shiv, I enjoy discussing this stuff and you can't do it on HF with out all the "haterz" coming out and trying to tell everyone they are wrong without any substance or evidence in their arguments.

I've read the same article Bruyns, and though far from conclusive, it was by far the most promising defense of shot attempts, and how they correlate to possession that I have seen. The mere regression analysis argument, and correlation to wins still irk me, so simple an analysis would not be deemed appropriate in an academic setting, and yet it is taken to be gospel by proponents of advanced stats. If I didn't know any better, I would think that the same proponents protested ice-cream sales on the basis that they correlate strongly with shark attacks.

[MSP4LYFE]

Who's the better player based on WAR shoalzie? Miggy? or Trout?!!?

p.s. sry about the tigers this yr... again

Trout, just like Trout is the better player according to any metric that takes defense and speed into account. Cabrera is the better hitter, but Trout is a comparable hitter, and superior base-runner and defender.

[MSP4LYFE]

I wouldn't argue against you since Bozak seems to be a better fit for the team. I just have an opinion that Grabovski is better at hockey than Bozak and was given a raw deal by Carlyle. I do understand why the decision was made to buy him out and keep Bozak.

Also I'd say that "stats geeks" (stupid term) on average watch WAY more hockey than the vocal crowd who refuses to have an open mind about them.

I'd argue that the average stat geek is equally close minded, if not even moreso.

[Bruyns]

I am a proponent of advanced stats as a way to further my understanding of teams and players, but don't use them to compare player A to player B or to make confident predictions on how a player or team will perform. I think they can be used to identify trends and can be used as a strong predictor for the future, but there are too many factors in play to make concrete predictions and have confidence it will come true.

"Personally I still struggle with the translation of shot attempts (w/ or w/o blocks) to be a good measure of 'possession' "

This was something that I struggled with too since logically to me shot attempts doesn't seem like it should be a proxy for possession. I can't find the article right now, but I read something where someone watched a bunch of games with a stop watch and measured the amount of time teams possessed the puck in the attacking zone. This was then compared to the shot attempt metrics for those games and they lined up showing that those stats did seem to be a perfect proxy for determining which team controlled possession.

Good idea for a thread Shiv, I enjoy discussing this stuff and you can't do it on HF with out all the "haterz" coming out and trying to tell everyone they are wrong without any substance or evidence in their arguments.

I've read the same article Bruyns, and though far from conclusive, it was by far the most promising defense of shot attempts, and how they correlate to possession that I have seen. The mere regression analysis argument, and correlation to wins still irk me, so simple an analysis would not be deemed appropriate in an academic setting, and yet it is taken to be gospel by proponents of advanced stats. If I didn't know any better, I would think that the same proponents protested ice-cream sales on the basis that they correlate strongly with shark attacks.

Agree with the criticism of the staunchest supporters of corsi and the corelation does not equal causation ice cream and sharks example was a nice touch. I have backed off a tad in pledging my full support to things like fenwick close will predict which teams will make the playoffs and other predictions for success based only off shot attempts. I still support the advanced stats movement and think they have a place in the game and will continue to expand and evolve, but I get annoyed with people like Dellow and Burtch who talk like they already know how things will play out and act like stats make everything black & white and conclusions can be drawn. I still respect the work of both bloggers, but in this movement is still in its infancy and if it's going to eventually become more widespread and accepted there needs to be more education rather than a condescending snarkiness to people who don't agree with their beliefs.

[MSP4LYFE]

mathematical correlation with flawed logic is not worth anything.

Just wanted to highlight this point.

[Robin Hood]

mathematical correlation with flawed logic is not worth anything. your theory and measures must be valid before correlation is worth anything. this is the pursuit of understanding and explaining, with large data sets (such as SOG) it's VERY VERY easy to find correlations, and if too many test are ran, even predictive relationships - the truth is without starting out with your own theory and processing it onward, the error far far far exceeds any reasonable confidence guidelines.


I'm not saying corsi/fenwick are useless, i'm saying they are far from perfect when taken too far; logical leaps are happening because they are being accepted as a measure of possession which is an indicator of team being in-control of the game, which is believed to be related to winning the game. it's too many steps removed.

saying the team that has a higher corsi will win more often, I agree is valid, logical, and supported. it has not been shown to cross levels well (to support comparing players). It was likely realized post-hoc, but than repeated over time and found to be consistent.



I agree individual corsi/fenwick numbers become more accurate from 5v5 starts removed - like I said, I read and follow this stuff too- but now coaching use is at least as big a factor.


I personally love quality of competition/line-mates as adding to understanding, however it actually explains very little - because it partially evens out over a season - aside from saying good teams have more good players, are able to play their good players versus other teams good players more often, and still play their (other) good players against other teams less good players, more often. It's support mathematically, and logically - I have no idea which way it was tested first, i accept it as reasonable use of the data, and strong theoretical understanding.

Your "theory" goes against what mathematics suggests, which implies that your theory has a level of bias.

theory is to be tested, saying fenwick is the best measure for possession is a theory which has supposedly since been tested, within the scientific method supported. - the leafs alone showed serious flaws in this theory, the question shouldn't be -> when will the leafs regress back to what our numbers expect, but rather should be -> what is our theory missing.


We should be clear in not only in what you're saying mathematics suggests, but actually in the role of theory. Especially in use for non-hard-science, the front end is extremely important - and you are down playing that here. there are MBA-esq papers and presentation 100000x's a year, which make use of data in an unfounded manner - because their theory is not sound, or they followed 95% of the scientific method, except they data mined without the proper levels of error (and many many more examples). Hell there are so many research journals out there now that there are thousands of poorly written papers that pass peer review at too low a level - but the academic community knows this, if it's not in a B+ level journal in most fields it's ignored and receives no weighting or further expansion. Once again back to the purpose of the research -> to add to understanding.


Apologies from broken reply - obviously at work and writing in-between actual tasks.

1. No one is saying that Corsi/Fenwick are the best models that will EVER be realized ---> not sure where in my post you read that? What I am saying is that you are dismissing it as not a proper mathematical model, when this is not at all a fair characterization.

2. What has been said is that so far, Corsi/Fenwick have provided the best models ---> THUS FAR. It still needs work but the level of confidence is higher than other models.

3. Also, I think you are discounting looking at stats post-hoc severely. If year after year after year stats repeat their patterns, there is most definitely predictive power in the model.

4. The Toronto Maple Leafs are not the rule. They are the exception. In fact, this season has been characterized as a major case study with regards to Corsi to see how the Leafs do (we are not supposed to be this good).

[Bruyns]

I wouldn't argue against you since Bozak seems to be a better fit for the team. I just have an opinion that Grabovski is better at hockey than Bozak and was given a raw deal by Carlyle. I do understand why the decision was made to buy him out and keep Bozak.

Also I'd say that "stats geeks" (stupid term) on average watch WAY more hockey than the vocal crowd who refuses to have an open mind about them.

I'd argue that the average stat geek is equally close minded, if not even moreso.

Care to expand?

Almost all I have come across fully admit the stats should be taken with a grain of salt and still place a huge amount of value on watching the games. Haven't seen many arguements from people saying don't watch games all you need to know are these stats, but I have seen a shit load of people although probably the majority are teenagers on HFboards claim the stats have no value, don't understand them and say ridiculous things like "watch the game" as if people taking time out of their day to discuss hockey on the internet don't watch it on TV.

What I meant by an open mind is taking the time to read and educate one's self on what they mean since I am positive there are plenty of people who have no idea or only a vague understanding of different stats, but still choose to chime into the debate like their opinion is valid when they have no clue what they are talking about. You know what they say though... ignorance is bliss

[MSP4LYFE]

Agree with the criticism of the staunchest supporters of corsi and the corelation does not equal causation ice cream and sharks example was a nice touch. I have backed off a tad in pledging my full support to things like fenwick close will predict which teams will make the playoffs and other predictions for success based only off shot attempts. I still support the advanced stats movement and think they have a place in the game and will continue to expand and evolve, but I get annoyed with people like Dellow and Burtch who talk like they already know how things will play out and act like stats make everything black & white and conclusions can be drawn. I still respect the work of both bloggers, but in this movement is still in its infancy and if it's going to eventually become more widespread and accepted there needs to be more education rather than a condescending snarkiness to people who don't agree with their beliefs.

We are on the same page, I also dislike close-minded people. For me personally, it does not matter what your opinion is, be it traditional, stats-based or a combination of both, the most important thing is that a person is open to new possibilities because at the end of the day no one knows everything, and it has been demonstrated time and time again over the course of history that our everyday beliefs are not as concrete and foundational as we like to believe. It should therefore come as no surprise that those beliefs that we hold true to our hearts in hockey may also be based on some false premises.

Using this discussion as an example, I can already tell that I can have an intelligent and rational debate with you and Nick, whereas I don't think I can have that with Shiv.

[Robin Hood]

mathematical correlation with flawed logic is not worth anything.

Just wanted to highlight this point.

Where is the flawed logic? Corsi/Fenwick are not "flawed". They may not have the highest level of correlation with what happens. But there is no way the logic is flawed. If you can produce mathematics to suggest how the logic is flawed, I'd be happy to hear it. But this is the kind of argument the "dinosaurs" hurl towards the statisticians when these arguments start. I have yet to read a mathematically sound post that discounts the logic of fenwick and corsi.

[Nick]

it's always that interesting balance. Does the better physical hockey player understand the game better than those who do not?
does some distance from the on-ice game give a better perspective at what actually work (emotions and norms removed)?
someone who played a little and studied a lot a good balance?
Played a lot and studied a little ?


Back to the best player not making the best coach/gm argument. Understanding the immeasurable aspects of a team is vital in those roles (dressing room, traveling life, physical cost, party life, etc) - but what's actually the proper balance of experienced/knowing/ignorance to deliver the best on ice team? In actual practice I believe we all see coach makes poor choices because they favor their own belief/urge to be right *Wilson*, or maybe an owner dips in on a personal decision (ahem Schneider) but I know there is a role for analytics to be used, however it cannot be an either/or situation - but rather the first to integrate it properly.




Wow got way off-topic there.


what are some aspects of a game/team/player that we 'know' helps a team win?

I love the role of the distracter in helping star players perform. Getting opposing stars off of their game, getting opposing attention of of their line-mate, and typically the ability to increase the 'temperature' of the game.

Burrows, Downie, Hartnell, Kunitz, Ablekater, etc - it's not an entirely rare skill, but is hard to find chemistry with the star (considering there is almost always a big skill gap) and the tenure in the role. I believe if measured correctly we could find a link to stars production.

[Nick]

Where is the flawed logic? Corsi/Fenwick are not "flawed". They may not have the highest level of correlation with what happens. But there is no way the logic is flawed. If you can produce mathematics to suggest how the logic is flawed, I'd be happy to hear it. But this is the kind of argument the "dinosaurs" hurl towards the statisticians when these arguments start. I have yet to read a mathematically sound post that discounts the logic of fenwick and corsi.

there is a logical flaw that corsi/fenwick are a measure of a team in control of the game.

logical and mathematically sound can be mutually exclusive. shark attacks and ice cream sales as an obvious example.

[Robin Hood]

what are some aspects of a game/team/player that we 'know' helps a team win?

I love the role of the distracter in helping star players perform. Getting opposing stars off of their game, getting opposing attention of of their line-mate, and typically the ability to increase the 'temperature' of the game.

Burrows, Downie, Hartnell, Kunitz, Ablekater, etc - it's not an entirely rare skill, but is hard to find chemistry with the star (considering there is almost always a big skill gap) and the tenure in the role. I believe if measured correctly we could find a link to stars production.

I have a big issue with the underlined statement.

Before I start - Kareem just took a shot at me above. Let's stop that. Moving on:

Here is my issue:

What we "know" is from how we watch the game. A good example is a fight. Conventional wisdom says:

1. A fight can turn momentum
2. Fight when you're down, and pump up your team
3. Fight when you're up, and you might help the other team

Now I have NO idea whether or not this is right, I'm just using it as an example here.

To evaluate this conventional wisdom, stats can play a huge role. Using Corsi as an indicator, for example: Does a team's SOG differential/60 minutes increase after a fight? Does the SOG differential/60 increase more when it the game is close?

What we "know" is not CLOSE to being verified. We do not know how much value a fight has. A lot of what you mentioned is subjective. Does a guy like Burrows throwing a hit help you more than a guy that spends most of his shift in the offensive zone?

[Bruyns]

Agree with the criticism of the staunchest supporters of corsi and the corelation does not equal causation ice cream and sharks example was a nice touch. I have backed off a tad in pledging my full support to things like fenwick close will predict which teams will make the playoffs and other predictions for success based only off shot attempts. I still support the advanced stats movement and think they have a place in the game and will continue to expand and evolve, but I get annoyed with people like Dellow and Burtch who talk like they already know how things will play out and act like stats make everything black & white and conclusions can be drawn. I still respect the work of both bloggers, but in this movement is still in its infancy and if it's going to eventually become more widespread and accepted there needs to be more education rather than a condescending snarkiness to people who don't agree with their beliefs.

We are on the same page, I also dislike close-minded people. For me personally, it does not matter what your opinion is, be it traditional, stats-based or a combination of both, the most important thing is that a person is open to new possibilities because at the end of the day no one knows everything, and it has been demonstrated time and time again over the course of history that our everyday beliefs are not as concrete and foundational as we like to believe. It should therefore come as no surprise that those beliefs that we hold true to our hearts in hockey may also be based on some false premises.

Using this discussion as an example, I can already tell that I can have an intelligent and rational debate with you and Nick, whereas I don't think I can have that with Shiv.

Burn!

First paragraph was well put, I have no problem admitting when I am wrong and people that struggle with this can come across as foolish. No problems with Shiv's stance either though as he is passionate in his beliefs and has no current evidence to dissuade him from the opinion he holds.

[Robin Hood]

Where is the flawed logic? Corsi/Fenwick are not "flawed". They may not have the highest level of correlation with what happens. But there is no way the logic is flawed. If you can produce mathematics to suggest how the logic is flawed, I'd be happy to hear it. But this is the kind of argument the "dinosaurs" hurl towards the statisticians when these arguments start. I have yet to read a mathematically sound post that discounts the logic of fenwick and corsi.

there is a logical flaw that corsi/fenwick are a measure of a team in control of the game.

logical and mathematically sound can be mutually exclusive. shark attacks and ice cream sales as an obvious example.

Nick you keep saying there is a logical flaw. But I don't have a objective reason from you for why a mathematical model with a lot of proof points behind it is flawed?

Now, I agree that they are not perfect. (Don't have the actual stat on this) but let's say Corsi can predict a Team's control of the game 60% of the time (could be higher than that). That is still a "model" that can be used.

To use it as gospel is wrong. But to use is as a metric along with other data makes a lot of sense.

[Nick]

what are some aspects of a game/team/player that we 'know' helps a team win?

I love the role of the distracter in helping star players perform. Getting opposing stars off of their game, getting opposing attention of of their line-mate, and typically the ability to increase the 'temperature' of the game.

Burrows, Downie, Hartnell, Kunitz, Ablekater, etc - it's not an entirely rare skill, but is hard to find chemistry with the star (considering there is almost always a big skill gap) and the tenure in the role. I believe if measured correctly we could find a link to stars production.

What we "know" is from how we watch the game. A good example is a fight. Conventional wisdom says:

1. A fight can turn momentum
2. Fight when you're down, and pump up your team
3. Fight when you're up, and you might help the other team

What we "know" is not CLOSE to being verified. We do not know how much value a fight has. A lot of what you mentioned is subjective. Does a guy like Burrows throwing a hit help you more than a guy that spends most of his shift in the offensive zone?

... reading my comments wrong it seems.

What we 'know' is in quotations because it's a commonly held belief that we do not actually have evidence that it's worthwhile. the question is how is it measured, and what are the expected benefits that we could potentially measure ?

My measure was not in regards to a fight either -> and like you I suspect have read the 2012-2013 fight outcomes on corsi... but I disagree (once again) with the front end theory that an enforcer fight it meant to have it's net impact in a single game - after the fight. In fact that screams of mis-casting the role with the purpose of proving it useless.

[Nick]

Any statistician worth their lunch can set up numbers to reach a desired outcome - back to my original point on the importance of front-end theory needing to be logically founded with proper understanding of it's scope.

[Robin Hood]

what are some aspects of a game/team/player that we 'know' helps a team win?

I love the role of the distracter in helping star players perform. Getting opposing stars off of their game, getting opposing attention of of their line-mate, and typically the ability to increase the 'temperature' of the game.

Burrows, Downie, Hartnell, Kunitz, Ablekater, etc - it's not an entirely rare skill, but is hard to find chemistry with the star (considering there is almost always a big skill gap) and the tenure in the role. I believe if measured correctly we could find a link to stars production.

What we "know" is from how we watch the game. A good example is a fight. Conventional wisdom says:

1. A fight can turn momentum
2. Fight when you're down, and pump up your team
3. Fight when you're up, and you might help the other team

What we "know" is not CLOSE to being verified. We do not know how much value a fight has. A lot of what you mentioned is subjective. Does a guy like Burrows throwing a hit help you more than a guy that spends most of his shift in the offensive zone?

... reading my comments wrong it seems.

What we 'know' is in quotations because it's a commonly held belief that we do not actually have evidence that it's worthwhile. the question is how is it measured, and what are the expected benefits that we could potentially measure ?

My measure was not in regards to a fight either -> and like you I suspect have read the 2012-2013 fight outcomes on corsi... but I disagree (once again) with the front end theory that an enforcer fight it meant to have it's net impact in a single game - after the fight. In fact that screams of mis-casting the role with the purpose of proving it useless.

I think we're finding a point of agreement which has been lost in the shuffle.

You think Corsi/Fenwick don't account for "enough" to be considered legitimate models since they "miss" things that can be very relevant to the outcome a game outside of what Corsi/Fenwick would predict.

I agree with that sentiment, but think that Corsi/Fenwick do provide models with some predictive power at a level of confidence (e.g. 50-60%, not sure if accurate).

Definitely as this movement goes forward though, models to incorporate more "events" and "variables" will come into play. What Corsi and Fenwick have done though is create a great model to build on. If using Corsi + adding other variables eventually leads to a 70% confidence level, for example, that would be a great thing.

[Nick]

Nick you keep saying there is a logical flaw. But I don't have a objective reason from you for why a mathematical model with a lot of proof points behind it is flawed?

Now, I agree that they are not perfect. (Don't have the actual stat on this) but let's say Corsi can predict a Team's control of the game 60% of the time (could be higher than that). That is still a "model" that can be used.

To use it as gospel is wrong. But to use is as a metric along with other data makes a lot of sense.

I do not need to show you a mathematical example of where the logic is flawed (that's itself a failing of any sounds theory? the scientist should be trying to prove their theory incorrect).

I'm not arguing that Corsi is a reasonable measure of possession of the puck - I'm saying that it is not one in the same. People take that acceptable relationship of cosri predicting possession, calling it in-control of the game, and than saying (example) boston wasn't in control of the finals even in the games they won - is so full of crucial logical gaps that I constantly feel like the discussion is off-mark.

If Boston plans on not having the puck in 'dangerous' ice -> rather than trying to generate shots on opposing goal, who is controlling the game? Because Corsi would say they are not in control of the game. So there is a logical failing in corsi being an indication of possession (because they won't have the goal of taking shots), nor the often common step to control of the game.


Perhaps we're saying the same thing from different side. I believe Corsi has a use, it's just not on saying who is in control of the game.

[Robin Hood]

Nick you keep saying there is a logical flaw. But I don't have a objective reason from you for why a mathematical model with a lot of proof points behind it is flawed?

Now, I agree that they are not perfect. (Don't have the actual stat on this) but let's say Corsi can predict a Team's control of the game 60% of the time (could be higher than that). That is still a "model" that can be used.

To use it as gospel is wrong. But to use is as a metric along with other data makes a lot of sense.

I do not need to show you a mathematical example of where the logic is flawed (that's itself a failing of logic?).

I'm not arguing that Corsi is a reasonable measure of possession of the puck - I'm saying that it is not one in the same. People take that acceptable relationship of cosri predicting possession, calling it in-control of the game, and than saying (example) boston wasn't in control of the finals even in the games they won - is so full of crucial logical gaps that I constantly feel like the discussion is off-mark.

If Boston plans on not having the puck in 'dangerous' ice -> rather than trying to generate shots on opposing goal, who is controlling the game? Because Corsi would say they are not in control of the game. So there is a logical failing in corsi being an indication of possession (because they won't have the goal of taking shots), nor the often common step to control of the game.

Perhaps we're saying the same thing from different side. I believe Corsi has a use, it's just not on saying who is in control of the game.

We're saying the same thing lol. It's pretty funny actually.

I, for example, do not think the Leafs will "regress" back to the mean, which is what Corsi suggests - because it doesn't cover everything.

But I also think the Leafs are the exception and not the rule.

[Nick]

being an exception to a theory is proving a shortfall in a theory. Often where the most progress is made: it's support for the null-hypothesis that corsi/PDO is not a measure for expected outcomes. Although -> in a grand-population sense, it's a useful measure, it just cannot cross levels.


I should add that PDO may be applicable across multi-year comparison, but has zero use on a team or player basis.


I'm more interested in reading the post-hoc analysis of others, seeing how it's applied, and trying to grow our own little tidbits - that would also get us more access to what Mrtle (w/e) is seeing/hearing.