I totally agree, and said so recently in other threads. To NHL teams that are on the forefront of analytics, stats like corsi and corsi related stats are archaic.
- MJL

Not really. There are more important metrics, but shot-based stats still offers useful, if not complete, information.

1) Analytics, pretty much by definition, is the process of using verified data to test hypotheses and make educated conclusions. Doesn't sound too awful, does it?

2) People assume that the stats we see is the extent of the data available to teams. I would propose that a good amount of time, money and effort has been invested by teams into more in-depth proprietary analyses that far surpass that which is available in the public sphere...
- lumlums

Unfortunately our ability to collect and analyze data is far from perfect and this is just as true for the proprietary data provided to NHL teams.

Also, inferring the best plan of action from analytic results is often an extremely subjective process. Pure analytics can be objective (to some extent), but in deciding which of these analytics should form our opinions/judgement and how we are going to choose to weigh them is an entirely subjective exercise.

For example, take Corsi, Fenwick and +/-. You can say with a reasonable degree of accuracy that these are objective data driven analytics. But then when you combine these 3 analytics to shape a narrative as to who is a good/bad player that is now entirely subjective because you have now decided the criteria that deems a player good or bad. Unfortunately, nobody is the objective authority on which model is the best model for evaluating performance.

In saying all of that, Tanner is not correct, there are plenty of deficiencies (mostly human) in applied statistics, and without a sanity check on the results (e.g., some experienced professionals) pure analytic decision making (not even really a thing) can go very wrong. Thus firing your entire scouting staff and looking purely at stats is not advisable.

And he's right. To paraphrase.
- Feds91Stammer

No he's not. At least not about the draft part.

Not really. There are more important metrics, but shot-based stats still offers useful, if not complete, information.
- lumlums

For the fans, it offers some useful information. NHL teams are way past it. I think they use some of the science of it, but their data collection methods are far better, which allows them to come to better conclusions.

For the fans, it offers some useful information. NHL teams are way past it. I think they use some of the science of it, but their data collection methods are far better, which allows them to come to better conclusions.
- MJL

This is 100% wrong. If anything, the public sphere is miles ahead of the majority of the league. Most teams use analytics to get the answers they want, or they have analytics departments that they fire or ignore if they aren't yes men.

There is still exponentially way more group-think/ eye-test confirmation bias garbage decision making in the NHL than there is teams that actually make use of the data available to them.

The idea that teams have cracked some code and the public stats guys (I'm just a writer, here I refer to the people actually doing the work) are clueless morons is as laughable as the idea that eye-test and stats analysis are things that should have an equal weight.

Unfortunately our ability to collect and analyze data is far from perfect and this is just as true for the proprietary data provided to NHL teams.

Also, inferring the best plan of action from analytic results is often an extremely subjective process. Pure analytics can be objective (to some extent), but in deciding which of these analytics should form our opinions/judgement and how we are going to choose to weigh them is an entirely subjective exercise.

For example, take Corsi, Fenwick and +/-. You can say with a reasonable degree of accuracy that these are objective data driven analytics. But then when you combine these 3 analytics to shape a narrative as to who is a good/bad player that is now entirely subjective because you have now decided the criteria that deems a player good or bad. Unfortunately, nobody is the objective authority on which model is the best model for evaluating performance.

In saying all of that, Tanner is not correct, there are plenty of deficiencies (mostly human) in applied statistics, and without a sanity check on the results (e.g., some experienced professionals) pure analytic decision making (not even really a thing) can go very wrong. Thus firing your entire scouting staff and looking purely at stats is not advisable.
- RonPielep

True to a point, though now there are better ways of collecting data, such as proper player tracking via cameras.

Your point about the decision makers is a good one - the best to hope for are good hockey people who believe not only in the stats, but also in their limitations.

Unfortunately our ability to collect and analyze data is far from perfect and this is just as true for the proprietary data provided to NHL teams.

Also, inferring the best plan of action from analytic results is often an extremely subjective process. Pure analytics can be objective (to some extent), but in deciding which of these analytics should form our opinions/judgement and how we are going to choose to weigh them is an entirely subjective exercise.

For example, take Corsi, Fenwick and +/-. You can say with a reasonable degree of accuracy that these are objective data driven analytics. But then when you combine these 3 analytics to shape a narrative as to who is a good/bad player that is now entirely subjective because you have now decided the criteria that deems a player good or bad. Unfortunately, nobody is the objective authority on which model is the best model for evaluating performance.

In saying all of that, Tanner is not correct, there are plenty of deficiencies (mostly human) in applied statistics, and without a sanity check on the results (e.g., some experienced professionals) pure analytic decision making (not even really a thing) can go very wrong. Thus firing your entire scouting staff and looking purely at stats is not advisable.
- RonPielep

Very well said. Only thing I'll add is that NHL teams have far better methods of gathering data than any of the stats based websites that we as fans have access to.

Thank you James, and I wish you success at MSNBC.
- MJL

The idea that MSNBC and FOX are equally bias is another fiction we should do away with. MSNBC is slanted to the left. Fox is outright lies and I would say is 99% responsible for the current political climate.

No he's not. At least not about the draft part.
- sbroads24

This is because there doesn't seem to be available data for minor leagues and even if there was, it'd be hard to compare across leagues. Obviously teams can do there own work in this regard and I bet it's an exploitable area.

Another thing: In the minors, players have a much wider range of skill levels, making it easier to tell who is good at what. When you watch NHL hockey, most of the players are so close in skill level scouting becomes much more of a crapshoot.

I read once that the only difference between a second line player a fourth line player was that the second line player got the opportunity and got lucky enough during that opportunity to score a few and get a longer look. It's very similar to a sociological theory about the accumulation of opportunity.

Anyways, currently, when I say fire your scouts, I mean your NHL scouts.

The idea that MSNBC and FOX are equally bias is another fiction we should do away with. MSNBC is slanted to the left. Fox is outright lies and I would say is 99% responsible for the current political climate.
- James_Tanner

Have you watched CNN lately? They make fox look neutral

This is 100% wrong. If anything, the public sphere is miles ahead of the majority of the league. Most teams use analytics to get the answers they want, or they have analytics departments that they fire or ignore if they aren't yes men.

There is still exponentially way more group-think/ eye-test confirmation bias garbage decision making in the NHL than there is teams that actually make use of the data available to them.

The idea that teams have cracked some code and the public stats guys (I'm just a writer, here I refer to the people actually doing the work) are clueless morons is as laughable as the idea that eye-test and stats analysis are things that should have an equal weight.
- James_Tanner

No, it's 100% right James. It's amazing that you make such claims to being an analytics guy, but yet, don't have the basic understanding of how NHL teams have advanced past the point of the basic stats we as fans have access to. I think you're basically afraid to admit that, because you'd then have to reassess your entire method of looking at players and at teams, and admit to the flaws that exist. You have this ridiculous belief that the NHL hierarchy if full a buffoons, unwilling to use the new and better methods of analytics for making decisions, that you have mastered. I know the Tanner playbook, you'll reference mistakes made by NHL teams as evidence, as if analytics never resulted in a bad decision being made. Yet, we read about the ones you make all the time.
The guy you love to trumpet, John Chayka talks about spending up to 25 hours breaking down tape, of a game that was just played. What do you think he's doing? Why doesn't just wait 10 minutes and go on any of the numerous analytics websites where the numbers are so readily available? If you read articles of GM's such as Stan Bowman, on the subject of analytics, he's reluctant to talk about what they do, and how, in terms of analytics. I guess he's just unwilling to talk a bout basic shot differential stats like Corsi. Bowman labels their stats propietary. I guess maybe he doesn't realize that other NHL teams have RelCoris stats also.

This is 100% wrong. If anything, the public sphere is miles ahead of the majority of the league. Most teams use analytics to get the answers they want, or they have analytics departments that they fire or ignore if they aren't yes men.

There is still exponentially way more group-think/ eye-test confirmation bias garbage decision making in the NHL than there is teams that actually make use of the data available to them.

The idea that teams have cracked some code and the public stats guys (I'm just a writer, here I refer to the people actually doing the work) are clueless morons is as laughable as the idea that eye-test and stats analysis are things that should have an equal weight.
- James_Tanner

Are you referring to the oilers and dellow with him not being a yes man?
I've heard that being floated around like its a fact. Does media and writers not remember them screaming to fire him back in February when he went on a rant on Twitter about #bellletstalk.
I can't say I know for sure why he was not resigned but this could be a larger factor than people assuming he wasn't a yes man

No, it's 100% right James. It's amazing that you make such claims to being an analytics guy, but yet, don't have the basic understanding of how NHL teams have advanced past the point of the basic stats we as fans have access to. I think you're basically afraid to admit that, because you'd then have to reassess your entire method of looking at players and at teams, and admit to the flaws that exist. You have this ridiculous belief that the NHL hierarchy if full a buffoons, unwilling to use the new and better methods of analytics for making decisions, that you have mastered. I know the Tanner playbook, you'll reference mistakes made by NHL teams as evidence, as if analytics never resulted in a bad decision being made. Yet, we read about the ones you make all the time.
The guy you love to trumpet, John Chayka talks about spending up to 25 hours breaking down tape, of a game that was just played. What do you think he's doing? Why doesn't just wait 10 minutes and go on any of the numerous analytics websites where the numbers are so readily available? If you read articles of GM's such as Stan Bowman, on the subject of analytics, he's reluctant to talk about what they do, and how, in terms of analytics. I guess he's just unwilling to talk a bout basic shot differential stats like Corsi. Bowman labels their stats propietary. I guess maybe he doesn't realize that other NHL teams have RelCoris stats also.
- MJL

I agree with some of this. It's not like teams are making trades based on players HERO charts

This is 100% wrong. If anything, the public sphere is miles ahead of the majority of the league. Most teams use analytics to get the answers they want, or they have analytics departments that they fire or ignore if they aren't yes men.

There is still exponentially way more group-think/ eye-test confirmation bias garbage decision making in the NHL than there is teams that actually make use of the data available to them.

The idea that teams have cracked some code and the public stats guys (I'm just a writer, here I refer to the people actually doing the work) are clueless morons is as laughable as the idea that eye-test and stats analysis are things that should have an equal weight.
- James_Tanner

The advantage that the teams have (those who use stats well anyway) over the public stats guys is the ability to derive data that's not commonly available to the public. The public stats guys do a solid job, but with few exceptions (such as Corey Snajzder), they are limited by the data that is available, and have to trust that the data obtained is accurate...

The idea that MSNBC and FOX are equally bias is another fiction we should do away with. MSNBC is slanted to the left. Fox is outright lies and I would say is 99% responsible for the current political climate.
- James_Tanner

You may have missed a massive email dump a few weeks ago. Just sayin.

The advantage that the teams have (those who use stats well anyway) over the public stats guys is the ability to derive data that's not commonly available to the public. The public stats guys do a solid job, but with few exceptions (such as Corey Snajzder), they are limited by the data that is available, and have to trust that the data obtained is accurate...
- lumlums

No doubt. Teams have the ability using video and software to break games down shift by shift, player by player. Instead of taking team shot data and applying it to every player that is on the ice, they can actually apply data to individuals player, and cut down a lot of the noise and false data that is involved with the stats that we the fans have access to. Roger Neilson did something similar back in the 70's when he used video to create a more accurate +/- stat. The limitation for NHL teams is time, and not being able to do it with every team, and every game played. In time, data collection will change, and technology will evolve to greatly improve analytics.

The advantage that the teams have (those who use stats well anyway) over the public stats guys is the ability to derive data that's not commonly available to the public. The public stats guys do a solid job, but with few exceptions (such as Corey Snajzder), they are limited by the data that is available, and have to trust that the data obtained is accurate...
- lumlums

One thing that people don't realize is that unless the data is like crazily out of whack - which I think people would notice - then even big improvements would only have minor consequences. You're still talking probabilities and percentages and the problems (as they are called) with data collection I would think make very little actual impact.

You are right that teams have better or proprietary models that are theirs alone, and that is one of the problems with improving analytics, as people at the forefront of the movement aren't exactly getting together to compare notes.

However, I think both the problems with public data and team proprietary information are minor problems for people legitimately interested in the topic, while a major problem is that guys like MJL discredit the whole industry by saying "public stats r dum."

You may have missed a massive email dump a few weeks ago. Just sayin.
- Mr.Bobby

Nah that's just classic liberal James. I'm still waiting on your response to the kadri drai argument James. It's been like 3 blogs now and your still not answering me

No doubt. Teams have the ability using video and software to break games down shift by shift, player by player. Instead of taking team shot data and applying it to every player that is on the ice, they can actually apply data to individuals player, and cut down a lot of the noise and false data that is involved with the stats that we the fans have access to. The limitation for NHL teams is time, and not being able to do it with every team, and every game played. In time, data collection will change, and technology will evolve to greatly improve analytics.
- MJL

I don't think you know this, but if you break down tape for ten hours, that isn't "the eye test" that is analytics. You are using data, observation and the scientific method to compare players and make an analysis.

When people (correctly I think I now have to say) mention the "eye test" derisively, they are talking about people watching games in real time.

Are you referring to the oilers and dellow with him not being a yes man?
I've heard that being floated around like its a fact. Does media and writers not remember them screaming to fire him back in February when he went on a rant on Twitter about #bellletstalk.
I can't say I know for sure why he was not resigned but this could be a larger factor than people assuming he wasn't a yes man
- kessellover69

I am not referring to him. I am referring to the fact that every team has some form of analytic department and most of those teams continue to make moves based on their 70 year old GMs feelings.

Nah that's just classic liberal James. I'm still waiting on your response to the kadri drai argument James. It's been like 3 blogs now and your still not answering me
- kessellover69

I can't read every comment, so if I don't answer you, shoot me a PM. I think it's clear by now I'll answer anything and that I don't care in the least if I am wrong or look stupid doing so.

One thing that people don't realize is that unless the data is like crazily out of whack - which I think people would notice - then even big improvements would only have minor consequences. You're still talking probabilities and percentages and the problems (as they are called) with data collection I would think make very little actual impact.

You are right that teams have better or proprietary models that are theirs alone, and that is one of the problems with improving analytics, as people at the forefront of the movement aren't exactly getting together to compare notes.

However, I think both the problems with public data and team proprietary information are minor problems for people legitimately interested in the topic, while a major problem is that guys like MJL discredit the whole industry by saying "public stats r dum."
- James_Tanner

I haven't discredited the whole industry. Again, this goes back to the fact that you just can't handle that the fantasy utopia you have of analytics, that are available to us, are flawed. Which is why you make statements that a team that fires it's entire scouting staff and just uses analytics would have a huge advantage over other teams. That is exhibit A of that. I use analytics all the time, to help me see what happens out there. From there, analysis is needed. There is a lot of false data in corsi, and any corsi related stat. Corsi in and of itself is a sound science. The data collection is the issue. Analytics is a part of the equation, not the entire equation. I'm smart enough to realize that.

One thing that people don't realize is that unless the data is like crazily out of whack - which I think people would notice - then even big improvements would only have minor consequences. You're still talking probabilities and percentages and the problems (as they are called) with data collection I would think make very little actual impact.

You are right that teams have better or proprietary models that are theirs alone, and that is one of the problems with improving analytics, as people at the forefront of the movement aren't exactly getting together to compare notes.

However, I think both the problems with public data and team proprietary information are minor problems for people legitimately interested in the topic, while a major problem is that guys like MJL discredit the whole industry by saying "public stats r dum."
- James_Tanner

You do a much better job of discrediting the entire industry every time you use stats inappropriately to support some of your more ridiculous hyberbolic click-bait statements.

If I had to choose between the confirmation bias of the eye-test and your relentlessly irresponsible use of applied statistics I would choose the former 10/10 times.

Luckily a middle ground exists. You know, people who are capable of expressing doubt and admitting potential flaws in their statistical analysis.

I can't read every comment, so if I don't answer you, shoot me a PM. I think it's clear by now I'll answer anything and that I don't care in the least if I am wrong or look stupid doing so.
- James_Tanner

Fair enough. I'm just referring to your comment in a blog that draisaitl would be lucky to develop into kadri. Looking at the regular box car I see that in draisaitl first full season hit 51 point and kadri in his career has never exceeded 50 points. I don't k ow a ton about advanced stats but I glance at a hero chart and see draisaitl' swaps better. So I'm just curious as to how you came to that conclusion.

I don't think you know this, but if you break down tape for ten hours, that isn't "the eye test" that is analytics. You are using data, observation and the scientific method to compare players and make an analysis.

When people (correctly I think I now have to say) mention the "eye test" derisively, they are talking about people watching games in real time.
- James_Tanner

I do know this, and I'm not talking about breaking down tape for an eye test, I'm talking about breaking it down to collect data to compile analytics.

Even John Chayka knows that you just can't rely on analytics. Maybe you should listen to him.

http://www.usatoday.com/s...-first-3-months/87757732/