Maple Leafs hire Kyle Dubas as AGM. Poulin/Loiselle gone.

[Tinordi]

*EDIT* Upon further reflection, I now wonder how useful it is to be both a competent observer of the sport and a trained statistician. I am now starting to think that certain benefits would follow from an entirely blind system of data processing. How much does one's own expectations and intuitions actually interfere with statistical results?

That's the rub and explains why there's such an acrimonious debate over using statistics in hockey among alot of fans. Stats try to remove bias from the interpreter. All of the best analysts I read try their best to let the numbers talk to them before their eyes. This is anathema to many observers like Simmons and alot of fans who want to construct narratives about teams, players, systems, coaches to weave an understanding of the game.

I would tell you though that no analyst would be hired to a team if he wasn't first a numbers interpreter and second a story teller. Anyone with an iota of understanding of the value of statistics wouldn't make that first fundamental mistake.

[edslunch]

Originally Posted by Textcritic

Well yes. But the degree is merely a reflection of one's successful comprehension and application of skills derived from the years of specialised training.

So would I, but I'm not at all convinced that the bloggers I am most familiar with have enough of a clear understanding of how to do advanced statistics to make them altogether useful. I imagine that I could read a book or take a basic introductory course and produce the same sorts of analyses. But I am also fairly confident that this limited "training" would no doubt fail at various points in my amateur attempts to measure, apply, and interpret data.

*EDIT* Upon further reflection, I now wonder how useful it is to be both a competent observer of the sport and a trained statistician. I am now starting to think that certain benefits would follow from an entirely blind system of data processing. How much does one's own expectations and intuitions actually interfere with statistical results?

In my industry we pair a domain expert with a 'data scientist' to come up with new insights and correlations. The data scientist understands stats to the nth degree and has a toolkit that goes lightyears beyond comparing shots for and against, the domain expert provides input on which factors are important to consider and helps apply the findings back in the real world. It's a good mix.

Edit: I should add that my industry is bound by the laws of physics and chemistry, so that no matter how surprising a correlation may be if it exists it exists. That doesn't apply so cleanly when humans are involved.

[Textcritic]

Originally Posted by Tinordi

...I would tell you though that no analyst would be hired to a team if he wasn't first a numbers interpreter and second a story teller. Anyone with an iota of understanding of the value of statistics wouldn't make that first fundamental mistake.

I'm not sure I understand what you are trying to say: Are you suggesting that every person hired by NHL teams to analyse and interpret statistics is "first a numbers interpreter and second a story teller" because he is employed by a NHL team? That doesn't seem right in light of its insufferable circularity.

[Textcritic]

Originally Posted by edslunch

In my industry we pair a domain expert with a 'data scientist' to come up with new insights and correlations. The data scientist understands stats to the nth degree and has a toolkit that goes lightyears beyond comparing shots for and against, the domain expert provides input on which factors are important to consider and helps apply the findings back in the real world. It's a good mix.

This is exactly what I'm getting at.

Originally Posted by edslunch

Edit: I should add that my industry is bound by the laws of physics and chemistry, so that no matter how surprising a correlation may be if it exists it exists. That doesn't apply so cleanly when humans are involved.

And here is where I think the so-called stats revolution has substantial room to grow. How do we take measurements of on ice events? What sorts of data do we collect? How do we weigh various factors? etc. As near as I can tell, the systems invented and presently employed are still very much in an emerging phase of development, and have not yet adequately addressed these many troubling nuances.

[Tinordi]

Originally Posted by Textcritic

I'm not sure I understand what you are trying to say: Are you suggesting that every person hired by NHL teams to analyse and interpret statistics is "first a numbers interpreter and second a story teller" because he is employed by a NHL team? That doesn't seem right in light of its insufferable circularity.

It's not a hard point to understand. If you're good enough to be hired by an NHL team, you're a good enough analyst to not fall victim to the very basic criticisms of using statistics that you raise.

[Textcritic]

Originally Posted by Tinordi

It's not a hard point to understand. If you're good enough to be hired by an NHL team, you're a good enough analyst to not fall victim to the very basic criticisms of using statistics that you raise.

I think that in itself is a rather presumptuous conclusion that merely posits that NHL teams will all do their due diligence. Except that we know from history that this is often not the case. I'm not prepared to give Brendan Shanahan or Craig MacTavish and Dallas Eakins so much benefit of doubt, and to merely assume that they know what they are doing.

[dustygoon]

Originally Posted by Tinordi

What makes someone a "statistician" a degree? I'll take the guys who both understand stats and are applying them to hockey.

Umm...you do that. And I will take the PhD academic with math/physics/economics background who is smart enough to apply his trade almost anywhere.

[bubbsy]

i applaud the move by the leafs, not because i'm a hard believer of advanced stats and their impact on hockey decisions as of today.

i like it. as the team with very deep pockets has to be creative when looking for an edge when they can't simply buy up all the talent in a cap league. This seems to be throwing money at an area that could really impact the on ice product in the short term (if the coaches use this to determine how they are executing game plans and line decisions) and/or long term (when GM using the data as a tool to determine player decisions in drafting , trade, or free agency.

i don't think any proffessional organization suffers from having too much data on hand. Investing in the group that is mining the data and creating cross reference views of the data that could lead to meaningful conclusions (working in conjunction with existing/traditional means used today), seems like a reasonable investment that i'm pretty sure all 30 teams will have done if not already then within a year or so.

[Ashasx]

Originally Posted by dustygoon

Umm...you do that. And I will take the PhD academic with math/physics/economics background who is smart enough to apply his trade almost anywhere.

Advanced stats right now in hockey are so basic that we call them advanced stats because they are simply a step above TOI, blocked shots, etc. They are more "advanced" than the basics, but they really aren't difficult to grasp.

You don't need a PhD to understand it or find value in these numbers. And I'd argue that if you do need a PhD to understand them, they have no value in hockey related decisions.

I'd far rather have somebody like Kent Wilson hired by the Flames than some mathematician who thinks he knows hockey.

[dustygoon]

Originally Posted by Ashasx

Advanced stats right now in hockey are so basic that we call them advanced stats because they are simply a step above TOI, blocked shots, etc. They are more "advanced" than the basics, but they really aren't difficult to grasp.

You don't need a PhD to understand it or find value in these numbers. And I'd argue that if you do need a PhD to understand them, they have no value in hockey related decisions.

I'd far rather have somebody like Kent Wilson hired by the Flames than some mathematician who thinks he knows hockey.

I know what current "advanced" stats entail. And cool. You grab Kent. Nice guy. But you can have a lot of academic horsepower for a relatively cheap cost...."knowing hockey" will not take long for the right guy or girl. As Text says, a fresh non-hockey perspective is upside bonus.

And the text book on hockey advanced stats is only going to grow as more technologies to track become available. In a few years we will laugh at how simple corsi was.

We can argue this all day long, but the results of hiring real academics to apply their knowledge to new complex situations are established.

[Chill Cosby]

Originally Posted by Ashasx

Advanced stats right now in hockey are so basic that we call them advanced stats because they are simply a step above TOI, blocked shots, etc. They are more "advanced" than the basics, but they really aren't difficult to grasp.

You don't need a PhD to understand it or find value in these numbers. And I'd argue that if you do need a PhD to understand them, they have no value in hockey related decisions.

I'd far rather have somebody like Kent Wilson hired by the Flames than some mathematician who thinks he knows hockey.

I don't think you're considering what these people should be hired for.

You don't even need to be Kent Wilson to understand advanced stats. The Flames could hire YOU if all they wanted was to understand advanced stats.

The benefit to hiring a professional, someone with a PhD perhaps, is that you're getting the understanding of what advanced stats are now, and the ability to expand that information in to far more intricate and exacting details. They're going to better know what numbers are valuable, what aren't, what numbers tell what truths, and what don't. It takes bloggers years to perfect and play with advanced stats, or to even decide what some of them mean. Someone with the proper education is going to be able to do far more with numbers than a blogger or a journalist.

Not to say these are all bad hires, but to say there aren't better hires than the blogger crowd is a bit silly.

[Textcritic]

Originally Posted by Ashasx

Advanced stats right now in hockey are so basic that we call them advanced stats because they are simply a step above TOI, blocked shots, etc. They are more "advanced" than the basics, but they really aren't difficult to grasp.

But part of what I was getting at is that this may be precisely the problem with advanced statistics. As near as I can tell, they theoretically depend upon methodological principles and models derived from the academic and scientific disciplines of statistics, but have been developed by amateurs, who may not necessarily understand enough about statistical analysis to make the best use of them. Their present usefulness is much debated, and the current results have been mixed. I would counter that this is possibly an outcome of the fact that the systems themselves are flawed in such ways that they were not correctly developed in the first place. In any event, I think it would be worthwhile to discover how a trained statistician would approach the same matter.

Originally Posted by Ashasx

...You don't need a PhD to understand it or find value in these numbers. And I'd argue that if you do need a PhD to understand them, they have no value in hockey related decisions.

What kind of an argument is that?

Originally Posted by Ashasx

I'd far rather have somebody like Kent Wilson hired by the Flames than some mathematician who thinks he knows hockey.

For me this would depend entirely upon the job for which such a person was hired. If that job is restricted to exotic data collection and processing, then I would prefer to leave that to a mathematician.

[CliffFletcher]

I guess the question is if the NHL has enough raw data to take advantage of the expertise the PhD level statistician brings to the table.

[bubbsy]

Originally Posted by CliffFletcher

I guess the question is if the NHL has enough raw data to take advantage of the expertise the PhD level statistician brings to the table.

i thinkn the biggest thing to make advanced stats successful is exactly what new raw data are they going to track. even the corsi, etc, advanced stats are using some pretty high level data (zone start/finish, toi, shots), but what if they started capturing board battle winning %, passing %, avg on ice skating distance or speed per game, faceoff % against right/left handers or against players taller, gameplay penalties vs post-whistle penalties, i mean the possibilities are endless.

as i've stated before, i find that the hardest thing about hockey is just how much of a team sport it is (more so than basketball as an example. Baseball is a stat man's dream, since the variables to an outcome are more controlled (ie. pitcher vs batter). The outcomes being tracked in hockey are a function of the 10 players on the ice, so how do you create the necessary mathematical models to be able to extract data for an individual's contribution and not have it skewed by their teammates, or the opposition quality. the number of variables we're talking about make it seem more chaotic in nature than say a baseball players batting tendency.

back to the point in question, i think advanced stats are so elementary in their evolution today, that i think both skillsets being discussed (math stats phd, vs hockey data guy) are likely needed to make this a meaningful exercise.

[Textcritic]

Originally Posted by CliffFletcher

I guess the question is if the NHL has enough raw data to take advantage of the expertise the PhD level statistician brings to the table.

I agree, but I also think there is more to this. It is one thing to properly collect and analyse data, but I would also expect that with professional training one also acquires requisite skills and expertise to cut through so much of the "noise" that is produced in these complicated sorts of models, and to accurately identify the right trends and averages that translate into meaningful information. In the end it is not just as simple as plotting coordinates on a chart or a graph.

[Flash Walken]

Originally Posted by Textcritic

I agree, but I also think there is more to this. It is one thing to properly collect and analyse data, but I would also expect that with professional training one also acquires requisite skills and expertise to cut through so much of the "noise" that is produced in these complicated sorts of models, and to accurately identify the right trends and averages that translate into meaningful information. In the end it is not just as simple as plotting coordinates on a chart or a graph.

I agree, and I think that's where having competent hockey people involved really matters:

http://forum.calgarypuck.com/showthr...cs#post3586050

Originally Posted by Flash Walken

The point where analytics meets area expertise is where excellence is formed.

Finding that balance is the difference between successful and unsuccessful organizations. Making sure your analytics are relevant to the goals of your team is the imperative, as no matter how much or how little date you gather, if in the end it is irrelevant towards contributing to the decision making process, it's been a waste of time. A big part of that, and where team's struggle with their implementation of statistics and predictive analytics is likely in incorporating their existing sources of expertise with their data sets; interpreting the statistics.

Coaches and GMs presented with a massive volume of information are naturally going to be intimidated by it. The good teams boil down this information to specific areas that the organization has identified as priorities, and avoid being mired in the numbers.

It's why extensive video libraries are now so important. You need to see more than the numbers, or you'll fall victim to Money Ball.

[JayP]

Originally Posted by CliffFletcher

I guess the question is if the NHL has enough raw data to take advantage of the expertise the PhD level statistician brings to the table.

I think once the player movement tracking systems are installed around the league, the sky is the limit and what is considered advanced stats now will be considered kid's play. The way the NBA analyzes that type of data is incredible.

[Flash Walken]

Originally Posted by JayP

I think once the player movement tracking systems are installed around the league, the sky is the limit and what is considered advanced stats now will be considered kid's play. The way the NBA analyzes that type of data is incredible.

And when it's introduced in Juniors, it should make the draft even more competitive.

[Rutuu]

Originally Posted by Textcritic

So would I, but I'm not at all convinced that the bloggers I am most familiar with have enough of a clear understanding of how to do advanced statistics to make them altogether useful. I imagine that I could read a book or take a basic introductory course and produce the same sorts of analyses. But I am also fairly confident that this limited "training" would no doubt fail at various points in my amateur attempts to measure, apply, and interpret data.

David Staples is the worst offender out there...

1. Super fan boy...check
2. Creating his own "data set" based on how he perceives scoring chances...check
3. Only watches Oiler games for his data set and not Det vs Fla on a monday night...check
4. Has a twitter handle and blog...check

[Street Pharmacist]

Originally Posted by JayP

I think once the player movement tracking systems are installed around the league, the sky is the limit and what is considered advanced stats now will be considered kid's play. The way the NBA analyzes that type of data is incredible.

Being completely unknowledgeable about the NBA, I'm curious if it's actually changed anything?