Fuzzy Analytics

[saltming]

So I want to start by saying I'm a complete analytics layman. I understand what they mean and do use them to support or challenge my eye test of players. This article is quite good and pretty much how I view and use analytics.

http://blueseatblogs.com/2016/07/24/putting-together-hockey-systems-stats-tools-talent-evaluation-primer/

I would recommend it to anyone who is also an analytics layman.
As I was reading I came across some interesting things like this that got me thinking there is obviously some analytics we are not privy too:

This is one area where analytics in hockey can get derailed. If everything is about Corsi, for example, that won't be the aspect that is undervalued. We're not at that stage yet, but understanding that it's about market inefficiencies means that should the day come when puck possession stats are overvalued metrics, teams can find a new edge with whatever is the inefficiency of that particular time.

This says to me that the true important analytics are looking for things other than what the market has already identified.

Also when dubas said this:

The parallels with the Leafs were obvious – and it was hard to not read into his comments when he discussed the transformative effect a coaching change had with the Greyhounds.

“What I learned was the extreme value of the buy-in throughout the organization,†Dubas said as he presented a graph of the Greyhounds’ dramatic improvement in possession data after hiring Sheldon Keefe as coach.

“If you have a coach that isn’t really buying into it and you’re bringing in personnel that are analytics-friendly and you have a coach that isn’t, it’s not going to work, because they have to work together. They can’t work separately.â€

It made me think that Dubas already has a formula that extends past what we (or I) know as the norm for advanced analytics. He talks about how the total buy in has to be there for it to work, GM, coach, players etc.

So in light of the Martin and Polak signings with the subsequent HF grumblings of signing these players, I submit that there is a fuzzy analytic that says these are the guys we need because the formula we have designed needs X+Y÷Z to be a winning team.

Do any of you that are well versed in analytics know if these analytical formulas exist or have any theories or evidence that will either support or challenge my thoughts?

Click to read more...

 

[Duke Silver]

Just because the Leafs are analytics-inclined, does not mean all decisions are made using analytics.

Martin brings a physical element to the forward ranks that is lacking on this team aside from Komarov. Especially important now that we have young rookies coming into the league. They need to feel safe.

Polak brings a physical element to the back end that is lacking on this team as well. Additionally, he sounds like a good soldier. Someone who sets an example in the gym and has leadership qualities. A glue guy.

My guess? Babcock thinks we're too soft.

We can't just ice a team of 6 puck-moving defencemen and 12 possession-driving forwards. Teams need balance. Sometimes you forego the analytics piece to address a need that is undervalued or immeasurable by metrics.

 

[saltming]

Just because the Leafs are analytics-inclined, does not mean all decisions are made using analytics.

Martin brings a physical element to the forward ranks that is lacking on this team aside from Komarov. Especially important now that we have young rookies coming into the league. They need to feel safe.

Polak brings a physical element to the back end that is lacking on this team as well. Additionally, he sounds like a good soldier. Someone who sets an example in the gym and has leadership qualities. A glue guy.

My guess? Babcock thinks we're too soft.

We can't just ice a team of 6 puck-moving defencemen and 12 possession-driving forwards. Teams need balance. Sometimes you forego the analytics piece to address a need that is undervalued or immeasurable by metrics.

So you don't think there might be a measurable quality that was identified as needed nd we call it toughness?
The red wings never had a classical tough guy but were/are very successful.
Perhaps the formula calls for X amount of hits from your forwards and X amout of hits from your defense?
I don't know and am just asking to see if any of the HF poster that are well versed in analytics have any thoughts on this.

 

[Duke Silver]

The fact you think there's some sort of magical formula underscores your green-ness with analytics. There are so many different metrics and so many contextual variables to account for. Analytics help inform decisions, that's it and that's all.

Red Wings never had a classical tough guy but they did have some jam. In recent years, gritty guys like Abdelkader, Ericsson, Kronwall, Helm, Miller, Glendening, Cleary all provided some physicality. They weren't a bruising team, nor did they fight hardly ever, but they didn't get pushed around.

 

[Mr Hockey*]

Babcock says, there are scouts/stat players and then there are coaches players.

 

[Nithoniniel]

I think the Dubas quotes are interesting. I agree with them with how I see analytics develop. Already, and perhaps more so in the coming years, we'll see how standardized, generic metrics such as possession impact will become an integral part of how players are evaluated.

That will lead to a point where teams will attempt to find exceptions to the general rules, players that perhaps impact quality more profoundly than quantity and thus would be undervalued in a market that relies on quantity as a measurement.

That doesn't mean that possession metrics will lose their meaning, just that regardless of how complete measurement metrics and systems we come up with, teams will seek an advantage by finding those that are not evaluated accurately.

I don't believe in a formula for success. I do believe there's a formula for team-level improvement though, and I think that goes through the market inefficiencies that Dubas mention.

 

[Randy Randerson]

That's a neat article, summarizes well what scouts/hockey ops people/writers have had to do to adapt - I would imagine that most of the current people in those positions who have more than 5 years of tenure are from Hockey backgrounds rather than stats backgrounds, so its interesting to see that they're all making an effort to adapt...and that they feel they all need to adapt, because advanced stats work if applied properly to analyze, identify weak links, and build better teams.

I really like the use of these stats, as they're eliminating many of the arguments for "intangibles" that kept guys who weren't NHL hockey players in the game, which in my opinion is making the game better to watch. I think its amazing how the bottom 2 lines of most NHL teams have changed in the last 5 years, less size, less grit, and can actually play hockey - to the point where there is now so much depth in the league that guys like Mike Santorelli are going to Europe because there's so much depth talent, and guys who are exclusively enforcers are basically extinct. Many of us had a "WTF" moment when the leafs signed Matt Martin...but even he can play, its not like it used to be with the Jay Rosehill's and Fraser McLaren's. Also interesting to see that big guys with grit who use it well are still valuable, guys who can close gaps quickly and create turnovers, or use their size to keep pucks and cycle.

I'm looking forward to the next generation of Analytics - what we currently use focuses on Correlation rather than Causality, because it's all derived from the result data of what happened when Player A was on the ice and often contrasted against what happened when they weren't. I believe within the next 5-10 years we'll see technology implemented that will actually track player actions (movements, ability to create gaps, close gaps, outcomes as a result of each player's affect on a play etc) that will help move us towards statistics that focus on Causality. Should get to be an interesting time to be an analytical fan.


The end of the article that focused on the visualization tools that are currently available was also interesting (HERO charts, WOWYS) - this part is really important to make the stats relate-able for the average person. I've got a finance degree, did a lot of regression, bell curving, R & R squared value stuff in school, and I still much prefer a visual representation of the results rather than looking at raw outputs - as people we're not computers and were designed to learn visually, it's great that these tools are being built and provided for free to the public

 

[saltming]

The fact you think there's some sort of magical formula underscores your green-ness with analytics. There are so many different metrics and so many contextual variables to account for. Analytics help inform decisions, that's it and that's all.

Red Wings never had a classical tough guy but they did have some jam. In recent years, gritty guys like Abdelkader, Ericsson, Kronwall, Helm, Miller, Glendening, Cleary all provided some physicality. They weren't a bruising team, nor did they fight hardly ever, but they didn't get pushed around.

Admittedly green.

So when dubas is tlkng about team systems and a direct correlation to analytics you don't think there is a formula of some sort?

Also I'm not thinking that analytics are the be all and end all, I do think they area tool, even in this aspect. I'm questioning if there is another level to them that we the general public are not aware of.

Dubas said, it's about finding that next level, the untapped niche that no other team is using.

 

[saltming]

I think the Dubas quotes are interesting. I agree with them with how I see analytics develop. Already, and perhaps more so in the coming years, we'll see how standardized, generic metrics such as possession impact will become an integral part of how players are evaluated.

That will lead to a point where teams will attempt to find exceptions to the general rules, players that perhaps impact quality more profoundly than quantity and thus would be undervalued in a market that relies on quantity as a measurement.

That doesn't mean that possession metrics will lose their meaning, just that regardless of how complete measurement metrics and systems we come up with, teams will seek an advantage by finding those that are not evaluated accurately.

I don't believe in a formula for success. I do believe there's a formula for team-level improvement though, and I think that goes through the market inefficiencies that Dubas mention.

Thank you.
That sheds more light on it for me.
I didn't think there was a formula for success but rather one for team building which is all inclusive.

 

[Joey Hoser]

Babcock says, there are scouts/stat players and then there are coaches players.

Yeah like Byron Froese.

 

[Mad Brills*]

Martin's shot suppression numbers are interesting though.

 

[indigobuffalo]

Just because the Leafs are analytics-inclined, does not mean all decisions are made using analytics.

Martin brings a physical element to the forward ranks that is lacking on this team aside from Komarov. Especially important now that we have young rookies coming into the league. They need to feel safe.

Polak brings a physical element to the back end that is lacking on this team as well. Additionally, he sounds like a good soldier. Someone who sets an example in the gym and has leadership qualities. A glue guy.

My guess? Babcock thinks we're too soft.

We can't just ice a team of 6 puck-moving defencemen and 12 possession-driving forwards. Teams need balance. Sometimes you forego the analytics piece to address a need that is undervalued or immeasurable by metrics.

Toughness is relatively meaningless on its own. Martin and Komarov are successful players because the hit counts come with other positive metrics like turnover ratios, and Corsi metrics.

Otherwise we'd still have Colton Orr. The fact of the matter is that there are driving factors beyond simply adding toughness. Martin was specifically targeted.

 

[Mad Brills*]

matt martin is a decent nhl player.

Colton orr was not.

 

[saltming]

That's a neat article, summarizes well what scouts/hockey ops people/writers have had to do to adapt - I would imagine that most of the current people in those positions who have more than 5 years of tenure are from Hockey backgrounds rather than stats backgrounds, so its interesting to see that they're all making an effort to adapt...and that they feel they all need to adapt, because advanced stats work if applied properly to analyze, identify weak links, and build better teams.

I really like the use of these stats, as they're eliminating many of the arguments for "intangibles" that kept guys who weren't NHL hockey players in the game, which in my opinion is making the game better to watch. I think its amazing how the bottom 2 lines of most NHL teams have changed in the last 5 years, less size, less grit, and can actually play hockey - to the point where there is now so much depth in the league that guys like Mike Santorelli are going to Europe because there's so much depth talent, and guys who are exclusively enforcers are basically extinct. Many of us had a "WTF" moment when the leafs signed Matt Martin...but even he can play, its not like it used to be with the Jay Rosehill's and Fraser McLaren's. Also interesting to see that big guys with grit who use it well are still valuable, guys who can close gaps quickly and create turnovers, or use their size to keep pucks and cycle.

I'm looking forward to the next generation of Analytics - what we currently use focuses on Correlation rather than Causality, because it's all derived from the result data of what happened when Player A was on the ice and often contrasted against what happened when they weren't. I believe within the next 5-10 years we'll see technology implemented that will actually track player actions (movements, ability to create gaps, close gaps, outcomes as a result of each player's affect on a play etc) that will help move us towards statistics that focus on Causality. Should get to be an interesting time to be an analytical fan.


The end of the article that focused on the visualization tools that are currently available was also interesting (HERO charts, WOWYS) - this part is really important to make the stats relate-able for the average person. I've got a finance degree, did a lot of regression, bell curving, R & R squared value stuff in school, and I still much prefer a visual representation of the results rather than looking at raw outputs - as people we're not computers and were designed to learn visually, it's great that these tools are being built and provided for free to the public

Very interesting about the causality versus correlation, but don't you think they do this already.
They coach and preach things like gap control and the like ad nauseum.

 

[saltming]

Yeah like Byron Froese.

Part of the tank equation?

 

[Mr Hockey*]

Part of the tank equation?

There was a discussion about where does the "GOAT" fit in on the Leafs.

 

[indigobuffalo]

As for OP, there are LOTS of metrics that we aren't privy to.

Not just specific ones like statistical data from biometrics systems (that Dubas already discussed, which can track player movement and even muscular contraction rates and determine if a player needs time in the pressbox to avoid injury); but also from heavy analysis of the data we do have, and forming coherent links that can produce results.

Not to mention having MORE ACCURATE data than what is available to us, because our data is taken from game results and that can have inaccuracies arena to arena.

A team like Toronto can employ entire crews for each arena and track data much more accurately and with greater detail, and get better, more accurate information as a result.

And there are so many metrics that track so many different things that even the "masters" can find false positives or false negatives will relative frequency.

Corsi and everything else are really just fancy stats that help fans get euphoric whenever SIWOTI (someone is wrong on the Internet)!

A great example of how most everyone ignored a really important stat that the Leafs jumped on was at the draft.

A metric was developed to better predict D+1 and D+2 players and the Leafs identified several key players that they targeted throughout the draft. Shortly thereafter, articles explaining the metric blow up hockey Twitter.

And yes, as your article pointed out, the teams looking to gain an edge will always be looking for the area where other teams aren't, and getting all the assets they can before other teams catch on.

Our current areas: overage draft picks and Russian players (many teams fear the "Russian factor" and stay clear)

 

[saltming]

There was a discussion about where does the "GOAT" fit in on the Leafs.

I don't know. He is part of the old regime but he is also an analytics sweetheart.

 

[indigobuffalo]

matt martin is a decent nhl player.

Colton orr was not.

Exactly. But you can explain why using analytics like turnover rates and Corsi to understand why.

Colton laid hits but usually arrived late after a player had had enough time to make a play.

And so then he's out of position making a meaningless hit.

Komarov is fast and attacks hard, and causes a lot of panicked plays and turnovers ensue, leading to better possession and scoring chances. And he has the hands to finish which Colton didn't.

Martin isn't as skilled offensively but he is just as fast and hits even harder, leading to just as many turnovers.

 

[Randy Randerson]

Very interesting about the causality versus correlation, but don't you think they do this already.
They coach and preach things like gap control and the like ad nauseum.

They don't have the ability to measure them right now - all the stats that are kept are the results of plays (shots, shot attempts, blocked shots, possession time, scoring, etc), and all the stats are derivative of those results. I agree that they are already coached and tracked by way of eye test by scouts, but they don't keep data on gap control (mostly because you would have to fill about half the stadium with stat keepers to have any chance at doing it effectively)

I think before too long you'll see setups in the arenas that allow for serious motion-capture of all players, and the stats will be modified to find correlation between different movements/spacings/player sizes/combinations of those/much more and the outcomes of the play like shots/blocks/shot attempts/possession/scoring, and move towards more determinant stats rather than correlation stats

The tech to do this already exists, just needs to be applied, I think you'll start to see that players start to wear equipment that allow for movement tracking and that arenas get equipped with a medium to record it all...this is probably still a while away, but I think that's where it will go:

 

[saltming]

As for OP, there are LOTS of metrics that we aren't privy to.

Not just specific ones like statistical data from biometrics systems (that Dubas already discussed, which can track player movement and even muscular contraction rates and determine if a player needs time in the pressbox to avoid injury); but also from heavy analysis of the data we do have, and forming coherent links that can produce results.

Not to mention having MORE ACCURATE data than what is available to us, because our data is taken from game results and that can have inaccuracies arena to arena.

A team like Toronto can employ entire crews for each arena and track data much more accurately and with greater detail, and get better, more accurate information as a result.

And there are so many metrics that track so many different things that even the "masters" can find false positives or false negatives will relative frequency.

Corsi and everything else are really just fancy stats that help fans get euphoric whenever SIWOTI (someone is wrong on the Internet)!

A great example of how most everyone ignored a really important stat that the Leafs jumped on was at the draft.

A metric was developed to better predict D+1 and D+2 players and the Leafs identified several key players that they targeted throughout the draft. Shortly thereafter, articles explaining the metric blow up hockey Twitter.

And yes, as your article pointed out, the teams looking to gain an edge will always be looking for the area where other teams aren't, and getting all the assets they can before other teams catch on.

Our current areas: overage draft picks and Russian players (many teams fear the "Russian factor" and stay clear)

Thanks! I completely overlooked the draft metric.
Add that on the pile.

 

[Mr Hockey*]

Babcock says, there are scouts/stat players and then there are coaches players.

Yeah like Byron Froese.

Part of the tank equation?

There was a discussion about where does the "GOAT" fit in on the Leafs.

I don't know. He is part of the old regime but he is also an analytics sweetheart.

lol ...

 

[saltming]

They don't have the ability to measure them right now - all the stats that are kept are the results of plays (shots, shot attempts, blocked shots, possession time, scoring, etc), and all the stats are derivative of those results. I agree that they are already coached and tracked by way of eye test by scouts, but they don't keep data on gap control (mostly because you would have to fill about half the stadium with stat keepers to have any chance at doing it effectively)

I think before too long you'll see setups in the arenas that allow for serious motion-capture of all players, and the stats will be modified to find correlation between different movements/spacings/player sizes/combinations of those/much more and the outcomes of the play like shots/blocks/shot attempts/possession/scoring, and move towards more determinant stats rather than correlation stats

The tech to do this already exists, just needs to be applied, I think you'll start to see that players start to wear equipment that allow for movement tracking and that arenas get equipped with a medium to record it all...this is probably still a while away, but I think that's where it will go:

OIC
That then is a very interesting idea. I've worked with the motion capture device on video, very neat!
That whole idea puts a very futuristic picturein my head.i like it!

 

[Mr Hockey*]

As for OP, there are LOTS of metrics that we aren't privy to.

Not just specific ones like statistical data from biometrics systems (that Dubas already discussed, which can track player movement and even muscular contraction rates and determine if a player needs time in the pressbox to avoid injury); but also from heavy analysis of the data we do have, and forming coherent links that can produce results.

Not to mention having MORE ACCURATE data than what is available to us, because our data is taken from game results and that can have inaccuracies arena to arena.

A team like Toronto can employ entire crews for each arena and track data much more accurately and with greater detail, and get better, more accurate information as a result.

And there are so many metrics that track so many different things that even the "masters" can find false positives or false negatives will relative frequency.

Corsi and everything else are really just fancy stats that help fans get euphoric whenever SIWOTI (someone is wrong on the Internet)!

A great example of how most everyone ignored a really important stat that the Leafs jumped on was at the draft.

A metric was developed to better predict D+1 and D+2 players and the Leafs identified several key players that they targeted throughout the draft. Shortly thereafter, articles explaining the metric blow up hockey Twitter.

And yes, as your article pointed out, the teams looking to gain an edge will always be looking for the area where other teams aren't, and getting all the assets they can before other teams catch on.

Our current areas: overage draft picks and Russian players (many teams fear the "Russian factor" and stay clear)

Those metrics showed that Adam Brooks was worthy of being a late first round pick in the 2016 draft.

 

[Randy Randerson]

OIC
That then is a very interesting idea. I've worked with the motion capture device on video, very neat!
That whole idea puts a very futuristic picturein my head.i like it!

The stuff that we have created as human beings for military, space exploration, and lots of scientific applications is honestly amazing. The stuff that we're talking about here has existed for decades - 1980's apache helicopters had those landscape-mapping domes with very accurate targeting using motion capture/prediction, and EA has been using those motion capture setups to make the player movements look realistic in games since the early 2000's at least. It seems futuristic but really just applying stuff we've had for a long time

Would also make the stats kept much more reliable than the human being-tracked ones now too