Around the NHL: 2019-20 Pt. 3

[Jim Bob]

Benning denies report Canucks are exploring Boeser trade

 

[Jim Bob]

Flames' Gaudreau not off to ideal start as playoff proving grounds loom - Sportsnet.ca

 

[Ehran]

I know I shouldn't find this as funny as I do, but with the NHL so heck-bent on "returning to play" the idea that one of their selected venues is flooding just hits me hilariously.

Rogers Place suffers water damage 2 weeks before NHL hub city hosting - TSN.ca

 

[Fearnot]

I know I shouldn't find this as funny as I do, but with the NHL so heck-bent on "returning to play" the idea that one of their selected venues is flooding just hits me hilariously.

Rogers Place suffers water damage 2 weeks before NHL hub city hosting - TSN.ca

That isn't how it's always looked?

 

[joshjull]

Probably a good idea but no fun.

The players should be pretty easy to hear without fans.

 

[Baccus]

Probably a good idea but no fun.

The players should be pretty easy to hear without fans.

The NHL is amazingly adept at ruining all things.

 

[joshjull]

 

[Fezzy126]

Their cap situation is really interesting

 

[SundherDome]

And why would he beheading up negotiations?
(I did that on purpose)

 

[joshjull]

And why would he beheading up negotiations?
(I did that on purpose)

well played.

 

[jc17]

Not going to lie, part of me likes that people point at Arizona as proof that analytics don't mean anything. If they had suddenly gotten good there would have been a much larger push around the league, and less competitive advantage in having them along with being that much further behind without them.

 

[SundherDome]

Not going to lie, part of me likes that people point at Arizona as proof that analytics don't mean anything. If they had suddenly gotten good there would have been a much larger push around the league, and less competitive advantage in having them along with being that much further behind without them.

It seems like they abandoned analytics over the last two years

 

[vcv]

Not going to lie, part of me likes that people point at Arizona as proof that analytics don't mean anything. If they had suddenly gotten good there would have been a much larger push around the league, and less competitive advantage in having them along with being that much further behind without them.

I mean.. that's not really how it works. Analytics are extremely valuable. But for them to be valuable:
1. You have to have the right data.
2. You have to have an acute understanding of that data.
3. You have to know how to translate that data into meaningful inferences.
4. You have to know how to validate your inferences.
5. You have to know how and when to make decisions based on all of that.

It's actually really hard to do right if you don't have competent people and full buy-in.

 

[Gras]

 

[jc17]

I mean.. that's not really how it works. Analytics are extremely valuable. But for them to be valuable:
1. You have to have the right data.
2. You have to have an acute understanding of that data.
3. You have to know how to translate that data into meaningful inferences.
4. You have to know how to validate your inferences.
5. You have to know how and when to make decisions based on all of that.

It's actually really hard to do right if you don't have competent people and full buy-in.

I dont know about arizona, but I think 3 is a sticking point often. Whether its twitter or even stuff during game broadcasts, there are a lot of numbers/visualizations that get thrown around that look like they should "good" metrics, but not a ton of evidence to back it up. For the flack corsi got, I remember a paper being written on how well it reflected possession in terms of o-zone ice time. Now it seems like many metrics pop-up with out an in-depth look at the insights they give

 

[SackTastic]

I dont know about arizona, but I think 3 is a sticking point often. Whether its twitter or even stuff during game broadcasts, there are a lot of numbers/visualizations that get thrown around that look like they should "good" metrics, but not a ton of evidence to back it up. For the flack corsi got, I remember a paper being written on how well it reflected possession in terms of o-zone ice time. Now it seems like many metrics pop-up with out an in-depth look at the insights they give

Creating a stat is easy. Proving that stat actually reflects something is hard.

The baseball stat revolution happened because nerds actually scienced things ; hypothesis , experiment, analyze, test, rinse, repeat. Too much of the hockey analytic community today only does part of the work ; hypothesis, experiment, twitter. This has lead to some flat out bad 'stats', and many people who riff off an unproven stat to create their own version of an unproven stat. Yay..?

 

[vcv]

I dont know about arizona, but I think 3 is a sticking point often. Whether its twitter or even stuff during game broadcasts, there are a lot of numbers/visualizations that get thrown around that look like they should "good" metrics, but not a ton of evidence to back it up. For the flack corsi got, I remember a paper being written on how well it reflected possession in terms of o-zone ice time. Now it seems like many metrics pop-up with out an in-depth look at the insights they give

Right... so it's not the analytics that are the problem. It's the people.

 

[Doug Prishpreed]

Right... so it's not the analytics that are the problem. It's the people.

Nope, the problem is the analytics/data. If it's so easy to make incorrect inferences (and it is so easy), is it really all that valuable?

There's no one person in the NHL who has even mastered it, from what I can tell. Whenever it seems like one management team or another has figured it out, their advantage never seems to last very long.

 

[jc17]

Right... so it's not the analytics that are the problem. It's the people.

You misinterpreted my first comment. I'm saying I like that analytics aren't universally accepted, or "proven" in hockey yet, because it keeps the window open for a team (perhaps one located in Buffalo) to gain a competitive advantage using them.

 

[jc17]

Nope, the problem is the analytics/data. If it's so easy to make incorrect inferences (and it is so easy), is it really all that valuable?

There's no one person in the NHL who has even mastered it, from what I can tell. Whenever it seems like one management team or another has figured it out, their advantage never seems to last very long.

Its also easy to make the wrong inference using the eye-test. Doesn't mean we'll fire all the scouts. (Well maybe WE will, but most teams wont)

No one has figured out coaching. We see "good" coaches fail, and coaches with poor histories do fine.

No one has figured out coaching, scouting, or analytics, but they each can provide value. If anything was ever "figured out" in sports, the team that did it would never lose

 

[vcv]

Nope, the problem is the analytics/data. If it's so easy to make incorrect inferences (and it is so easy), is it really all that valuable?

So you're argument is that if something is complex, then it's not that valuable? Uh ok. Good luck with that.

You misinterpreted my first comment. I'm saying I like that analytics aren't universally accepted, or "proven" in hockey yet, because it keeps the window open for a team (perhaps one located in Buffalo) to gain a competitive advantage using them.

Ok, that makes more sense. That didn't come through at all in your first comment though: as proof that analytics don't mean anything. , even with the sentence you followed up with.

Anyway, the "data revolution" (i don't know if there is a standard/accepted term) is still so so young. There's a lot of maturing to do. HOWEVER, using data should still be using the scientific method (like Sack said), and if people are using data without following best or standard practices, then that's not a flaw in analytics/data, that's a flaw in that person's education.

No one has figured out coaching, scouting, or analytics, but they each can provide value. If anything was ever "figured out" in sports, the team that did it would never lose

I can't agree with this. There are so many other variables in sports. Development,intangibles, randomization, politics, and so on. And even if a team does figure out how to do one thing really really well, it's only a matter of time before they are copied and lose their competitive advantage.

 

[joshjull]

Nope, the problem is the analytics/data. If it's so easy to make incorrect inferences (and it is so easy), is it really all that valuable?

There's no one person in the NHL who has even mastered it, from what I can tell. Whenever it seems like one management team or another has figured it out, their advantage never seems to last very long.

Its not that the numbers/stats are bad. Its Just that they can only get so much out of them. Hockey is the toughest sport to get clean numbers in. There are too many factors or noise to filter out that can impact individual and team performance. Which reduces the impact using those numbers have relative to other sports. Its why that initial positive bump you mentioned only lasts so long.

 

[vcv]

Its not that the numbers/stats are bad. Its Just that they can only get so much out of them. Hockey is the toughest sport to get clean numbers in. There are too many factors or noise to filter out that can impact individual and team performance. Which reduces the impact of using those numbers have relative to other sports. Its why that initial positive bump you mentioned only lasts so long.

There's a common term that can describe this. One that my team and I love to use.

GIGO.

And you're dead on about Hockey being the toughest sport. I think it's also generally always been behind the other big 3 sports in terms of advancements in just about anything, including stats.

The league and teams are working on stuff, but it's pretty slow moving w/o the huge $$$.

 

[brian_griffin]

Doesn’t baseball have a great advantage for data vs any “in-motion” sport in that nearly all its events are “static”? I don’t think Discrete vs continuous is the right term, what is?

 

[dotcommunism]

Doesn’t baseball have a great advantage for data vs any “in-motion” sport in that nearly all its events are “static”? I don’t think Discrete vs continuous is the right term, what is?

Well, plays in baseball certainly are discrete, especially compared to sports like hockey and basketball. Of course, football plays are discrete events too. A big thing with baseball is that a lot of plays have very few participants in them which reduces how many variables you have to take into account. But even with baseball statistics, when you start getting into anything involving measuring defense, you're in muddy waters. A lot of defensive statistics range from ambiguously useful to outright useless. Outside of how much more difficult it is to objectively measure defense, a lot of defensive statistics, UZR for example, require multiple seasons worth of data to be reliable. Most hitting stats, on the other hand, are reliable within a few hundred plate appearances.

Aside from the difficulty of measuring everything that happens in a hockey rink at all once, the other big limiting factor as far as hockey analytics go is that hockey is a low event game. Your average NBA game, another sport with almost as many moving parts as hockey (although basketball players can't change on the fly), had around 80 field goals made per game. The average NHL game this season had about 6 goals scored per game. Having a higher amount of scoring events makes it easier to put together reliable stats.