Eklund Rumor: Leafs in on Cam Fowler

[Number1RedWingsFan52]

I'd love Fowler on the leafs but because this is from Eklund it's almost certain we aren't actually in on Fowler..

Exactly considering that Eklund is the only mentioning Fowler to the Leafs, None of the other insiders are reporting it so take it with a grain of salt it's Eklund. And i'm not suggesting it wont happen just sceptical is all.

 

[mytduxfan*]

There is just soooo much wrong with this post? +/- as evidence of anything? Gardiner only played 17 minutes a games? Gardiner is 'awful' at anaytics? Especially relative to teammates? Literally all of these claims are factually inaccurate. Most are true of Cam Fowler oddly enough, but Jake Gardiner? Nah

... and Trouba. Kids a bust.

 

[mytduxfan*]

Statistically valid findings, but scientifically assbackwards process. There's a reason important research is conducted with experimental control and clinical trials that trump the raw epidemiological data.

Taking a clearly shown fundamental relationship, introducing a myriad of uncontrolled confounding variables, then failing to model an adjustment for the fundamental relationship doesn't invalidate the relationship, it means that the adjustment model isn't up to snuff.

Well played sir.

Unfortunately, not everyone looks at these advanced stats with the same scientific scrutiny as you or I. People need to understand where these numbers come from before they start spewing them out as if they're the gospel truth. Some people on here seem to even struggle with the basic concept of "context" or "player usage"... well, until you use the exact same backward logic to slate their players, then they seem to have a stream of plausible reasons as to why the numbers of said players aren't as good as they should. Funnily enough, they're the exact same plausible reasons you listed earlier when they were slating the crap out of your player because of his poor possession stats. However, at that time, they were "just excuses".

P.S. Does anyone know why these advanced statistics don't come with statistical error? Statistical error is a good way to judge the robustness of data. However, these numbers always seem to be devoid of such a simple value.

 

[Randy Randerson]

Well played sir.

Unfortunately, not everyone looks at these advanced stats with the same scientific scrutiny as you or I. People need to understand where these numbers come from before they start spewing them out as if they're the gospel truth. Some people on here seem to even struggle with the basic concept of "context" or "player usage"... well, until you use the exact same backward logic to slate their players, then they seem to have a stream of plausible reasons as to why the numbers of said players aren't as good as they should. Funnily enough, they're the exact same plausible reasons you listed earlier when they were slating the crap out of your player because of his poor possession stats. However, at that time, they were "just excuses".

P.S. Does anyone know why these advanced statistics don't come with statistical error? Statistical error is a good way to judge the robustness of data. However, these numbers always seem to be devoid of such a simple value.

I think most of them were developed by statistically inclined Hockey people, rather than Statisticians, who knew what they wanted to measure and took the shortest most logical way to get there without running them through the validation rigors that would be in place for something facing a peer review scrutiny, and adding those rigors makes them more difficult to understand for those who would use them/fans

Also, the cost of controlled variable lab environment would be large. Apparently U of Waterloo is doing some of that kind of thing now, as I found out in this thread: http://hfboards.mandatory.com/showthread.php?t=2115861
So the developers of what we have now probably didn't have much of a choice but to use live hockey to develop the metrics

my 2 cents

 

[4thline]

I think most of them were developed by statistically inclined Hockey people, rather than Statisticians, who knew what they wanted to measure and took the shortest most logical way to get there without running them through the validation rigors that would be in place for something facing a peer review scrutiny, and adding those rigors makes them more difficult to understand for those who would use them/fans

Also, the cost of controlled variable lab environment would be large. Apparently U of Waterloo is doing some of that kind of thing now, as I found out in this thread: http://hfboards.mandatory.com/showthread.php?t=2115861
So the developers of what we have now probably didn't have much of a choice but to use live hockey to develop the metrics

my 2 cents

Just wanted to say that the uWaterloo stuff isn't controlled variable testing as far as I know, in fact far from it. Mainly tech and code development using the arena facilities, Varsity teams, and intramural players. All live play.

 

[Randy Randerson]

Just wanted to say that the uWaterloo stuff isn't controlled variable testing as far as I know, in fact far from it. Mainly tech and code development using the arena facilities, Varsity teams, and intramural players. All live play.

Ah I see - well hopefully a gateway to more money put in to the development of all kinds of things that will help us understand the games better. The development of tracking tech that can be used in all situations without inhibiting gameplay is a massive leap forward IMO, whether its wearable or doesn't require that like the system which is linked in that thread

 

[4thline]

Well played sir.

Unfortunately, not everyone looks at these advanced stats with the same scientific scrutiny as you or I. People need to understand where these numbers come from before they start spewing them out as if they're the gospel truth. Some people on here seem to even struggle with the basic concept of "context" or "player usage"... well, until you use the exact same backward logic to slate their players, then they seem to have a stream of plausible reasons as to why the numbers of said players aren't as good as they should. Funnily enough, they're the exact same plausible reasons you listed earlier when they were slating the crap out of your player because of his poor possession stats. However, at that time, they were "just excuses".

P.S. Does anyone know why these advanced statistics don't come with statistical error? Statistical error is a good way to judge the robustness of data. However, these numbers always seem to be devoid of such a simple value.

I don't think that's quite fair, most work I've read the statistical application is perfect (error and correlation coefficients etc. are presented in the development work, not the fan applications). My only beef is with the conclusions surrounding the various "adj" and "rel" factors, and to a lesser extend sh/sv % influencers. The player themselves, QoC, QoT, ZS, score effect, random chance, all interacting, and more importantly all changing. Impossible to model the individual impacts of each on a macro level without considering the others, and there's just never going to be the data to support developing that model. That many of the relationships get smoothed on the aggregate is common sense, in a complex competitive system inefficiencies are going to be corrected before they have a chance to be statistically significant. Player gets better, gets better linemates against better opponents, team gets worse, starts in his end more. No single variable "adj" is going to capture that. By limiting your analysis of contextual factors to what shows up in the ****** macro data you're cutting the usefulness of the good controlled data off at the knees.

 

[mytduxfan*]

Ah I see - well hopefully a gateway to more money put in to the development of all kinds of things that will help us understand the games better. The development of tracking tech that can be used in all situations without inhibiting gameplay is a massive leap forward IMO, whether its wearable or doesn't require that like the system which is linked in that thread

I think this is the general progression here. However, IMO, it is the models that need to be improved upon, not necessarily the method of collecting data. Still, as 4thliner said, a comprehensive model that considers all the factors of ones play i.e. usage, QoT, QoC, ZS, etc. and the interrelationships of those factors is where it gets very, very complicated. However, for me, until you do have that model, referencing shot suppression/possession stats is a fairly redundant process.

I don't think that's quite fair, most work I've read the statistical application is perfect (error and correlation coefficients etc. are presented in the development work, not the fan applications). My only beef is with the conclusions surrounding the various "adj" and "rel" factors, and to a lesser extend sh/sv % influencers. The player themselves, QoC, QoT, ZS, score effect, random chance, all interacting, and more importantly all changing. Impossible to model the individual impacts of each on a macro level without considering the others, and there's just never going to be the data to support developing that model. That many of the relationships get smoothed on the aggregate is common sense, in a complex competitive system inefficiencies are going to be corrected before they have a chance to be statistically significant. Player gets better, gets better linemates against better opponents, team gets worse, starts in his end more. No single variable "adj" is going to capture that. By limiting your analysis of contextual factors to what shows up in the ****** macro data you're cutting the usefulness of the good controlled data off at the knees.

I agree. However, it's the method or models themselves, or, at the very least, what they represent that I tend to question. For example, the model behind AdjSV% uses cutoff points to weight the difficulty of shots which are arbitrary at best.

 

[Machinehead]

That needs to be done to validate those conclusions, because as you said the effects are huge. It's pretty lazy thinking to average times when teams are trying to generate shots with those when they aren't, then say that there's no effect based on zone starts. There may not BE a difference in results, but it's pretty shoddy work to just assume that there isn't. That's my main issue with advanced stats in hockey - an assumption is made without necessarily checking all the facts, and it becomes dogma.

And you think we haven't examined score effects on Fowler? The only assumption being made is that the stats people are assuming anything.

 

[Sojourn]

And you think we haven't examined score effects on Fowler? The only assumption being made is that the stats people are assuming anything.

Considering the statistics don't cover everything, that's a pretty safe assumption. There are a lot of assumptions made in these statistics. Possession, for one. It's a possession statistic that assumes possession. It doesn't actually keep track of possession. It keeps track of shots.

I do enjoy the "we" aspect of this, when the we is simply you and your agenda against Fowler.

 

[Machinehead]

Considering the statistics don't cover everything, that's a pretty safe assumption. There are a lot of assumptions made in these statistics. Possession, for one. It's a possession statistic that assumes possession. It doesn't actually keep track of possession. It keeps track of shots.

I do enjoy the "we" aspect of this, when the we is simply you and your agenda against Fowler.

Find me one advanced stats writer who thinks Fowler is a good player.

 

[Sojourn]

Find me one advanced stats writer who thinks Fowler is a good player.

That's not a defense. That's deflection. And completely ridiculous, by the way.

You live and die with these statistics. A statistician who doesn't recognize the limitations of the numbers, or is unwilling to accept that the data is incomplete is a poor one.

 

[dracom]

Find me one advanced stats writer who thinks Fowler is a good player.

Find me one respected hockey analysts that thinks Fowler is a bad hockey player.

FYI, I'll trust the analysts and those actually working in the league over "fancy stats writers" any day.

 

[Sojourn]

Find me one respected hockey analysts that thinks Fowler is a bad hockey player.

FYI, I'll trust the analysts and those actually working in the league over "fancy stats writers" any day.

It's easier to let the numbers make your decisions for you. How comforting must it be to know that you no longer have to do any work to decide if a player is good or bad? Just let a formula decide for you. Hockey isn't more complicated than that, right?

 

[Machinehead]

That's not a defense. That's deflection. And completely ridiculous, by the way.

You live and die with these statistics. A statistician who doesn't recognize the limitations of the numbers, or is unwilling to accept that the data is incomplete is a poor one.

And again we're back to the myth that Fowler hasn't been examined and probed to the ends of the Earth and there's no context here.

 

[Sojourn]

And again we're back to the myth that Fowler hasn't been examined and probed to the ends of the Earth and there's no context here.

Since you don't have access to the statistics that are capable of doing so, I don't think it's a myth at all.

 

[Machinehead]

Since you don't have access to the statistics that are capable of doing so, I don't think it's a myth at all.

And we're back to the mystery statistics that nobody's ever seen that say Fowler is great, which the Ducks have access to, even though they don't have an analytics department.

 

[Sojourn]

And we're back to the mystery statistics that nobody's ever seen that say Fowler is great, which the Ducks have access to, even though they don't have an analytics department.

As opposed to the assumption that teams would invest a lot of money into an analytics department that would consist of them just having a site with Corsi numbers bookmarked?

I don't think you should be trying to suggest someone else is being ridiculous, if you're going to come out with statements like that.

Edit: The rest of that nonsense I never said.

 

[Ducks DVM]

And you think we haven't examined score effects on Fowler? The only assumption being made is that the stats people are assuming anything.

I'm not talking about Fowler at all.

I'm talking about the glaring holes you can drive a truck through in the methodology used on that "settled" topic you brought up. You should be intellectually honest enough to admit I have a point if you want to be taken seriously about the topic.

Edit - and the bolded is ridiculous. Many of the blogs put out have the line "we may assume" in them, most importantly many of the first ones put out on a topic, that are then referenced as "proof" by people who build upon the work. There's an egregious amount of non-rigorous work done in this field. I'm not a statistician, but I read a lot of medical journals and can certainly spot where a study design is poor and where people skip steps and then fuzz things by using statistics loosely.

 

[WhatTheDuck]

And we're back to the mystery statistics that nobody's ever seen that say Fowler is great, which the Ducks have access to, even though they don't have an analytics department.

http://www.sportsnet.ca/hockey/nhl/defencemen-best-exiting-zone/

 

[lindholmie]

http://www.sportsnet.ca/hockey/nhl/defencemen-best-exiting-zone/

Do you have a link to the one from the playoffs?

 

[JetsHomer]

http://www.sportsnet.ca/hockey/nhl/defencemen-best-exiting-zone/

Care to show how these stats were collected/what the sample size is?

Oh that's right, you can't, because that is an absolutely garbage article

 

[Ciao]

A lot of people use statistics like a drunk uses a lamp post: for support, not illumination.

Was it Samuel Clemens who said to the effect: "There are lies; damn lies; and then there are statistics!"

If you already have an open mind on a topic and you want to use statistics to help understand some aspect of what you're studying they can be very useful.

If you're having an argument over something in respect of which you've already made up your mind -- well, go ahead and knock yourself out. You will find some to support whatever you want them to say.

I've been watching hockey long enough that I don't need any statistics to tell me when I see a player I like.

I haven't watched Fowler enough to know anything about him.

Jake Gardiner has been one of my favourites since he joined the Leafs. I really enjoy watching him play. He skates so very well and he is learning how to be stronger on his feet and to play a good defensive game. He's always improving and he's a better player year by year. I think Gardiner is a keeper.

 

[4thline]

I'm not talking about Fowler at all.

I'm talking about the glaring holes you can drive a truck through in the methodology used on that "settled" topic you brought up. You should be intellectually honest enough to admit I have a point if you want to be taken seriously about the topic.

Edit - and the bolded is ridiculous. Many of the blogs put out have the line "we may assume" in them, most importantly many of the first ones put out on a topic, that are then referenced as "proof" by people who build upon the work. There's an egregious amount of non-rigorous work done in this field. I'm not a statistician, but I read a lot of medical journals and can certainly spot where a study design is poor and where people skip steps and then fuzz things by using statistics loosely.

That's a good comparison. The analogue of a lot of the more "in depth" corsi research would be trying to identify fundamental causal relationships to human health (measured by bp and heartrate = corsi) with a 5 question survey where each variable is dependent on the others .

 

[Machinehead]

http://www.sportsnet.ca/hockey/nhl/defencemen-best-exiting-zone/

Obviously that's an aspect of the game he excels at, but what does that mean if he's not using it to prevent shots or chances?

He's a good skater too, but it's not helping his end result any.