Who are the analytics darlings that are terrible?

[SDS]

Since doubters routinely call analytics into question, who does the analytics crowd say they are in the top 20%, but really are are in the bottom 20% of the league? (and vice versa)

Are the errors bars equal in each case? Who are we afraid that the nerds say 👍 but the doubters say 👎.

[Thorny]

Jeff Skinner, - @dudacek

[SDS]

1 hour ago, Thorny said:

Jeff Skinner, - @dudacek

Is he considered a top 20% guy in analytics?

[LGR4GM]

1 minute ago, SDS said:

Is he considered a top 20% guy in analytics?

No. He's a volume scoring who outshot his sh% during his contract year and the Sabres were too stupid to notice. 

[Randall Flagg]

Skinner was like top 5 in even strength goal scoring over like a 5 year span before that contract and I explicitly remember his excellent metrics when we traded for him 

[thewookie1]

Matt Irwin was a darling only 3 or so years ago.

[Thorny]

16 minutes ago, SDS said:

Is he considered a top 20% guy in analytics?

Maybe not top 20. He's had some really good numbers, though, as recently as last season. And even this season he was among the elite last I looked at drawing penalties - which leads to goals. I was thinking he fits as an example of a player who has had better underlying numbers than production

Edited July 10, 2021 by Thorny

[LGR4GM]

2 minutes ago, Thorny said:

Maybe not top 20. He's had some really good numbers, though, as recently as last season. And even this season he was among the elite last I looked at drawing penalties - which leads to goals. I was thinking he fits as an example of a player who has had better underlying numbers than production

He does, his xgf I remember being good and he was still generating decent HD shots

[Eleven]

2 hours ago, LGR4GM said:

No. He's a volume scoring who outshot his sh% during his contract year and the Sabres were too stupid to notice. 

Krueger is involved though.

[pi2000]

cody eakin

[BagBoy]

Sorry if I'm twisting the original intent of this thread, but I am just as interested in how the NHL employees who are getting paid to crunch analytics data rank.  Who is doing a great job and why?  Who are the laggards and where have they failed?  I'd like to have a feel for who's who as long as Uncle Terry is still doing the hiring.

[rakish]

40 minutes ago, BagBoy said:

Sorry if I'm twisting the original intent of this thread, but I am just as interested in how the NHL employees who are getting paid to crunch analytics data rank.  Who is doing a great job and why?  Who are the laggards and where have they failed?  I'd like to have a feel for who's who as long as Uncle Terry is still doing the hiring.

I did a video about this in the 2021 draft thread. My argument is that you value a teams ability to do analytics by examining their draft picks to see how they deal with age bias. I looked at 4 teams, to summarize, LA is good, Carolina does OK, Buffalo not so much, Pittsburgh doesn't recognize there's a problem that they need to adjust for.

[Huckleberry]

21 minutes ago, rakish said:

I did a video about this in the 2021 draft thread. My argument is that you value a teams ability to do analytics by examining their draft picks to see how they deal with age bias. I looked at 4 teams, to summarize, LA is good, Carolina does OK, Buffalo not so much, Pittsburgh doesn't recognize there's a problem that they need to adjust for.

So lets basically turn this assumption into Quinn.

My eye test of him says he will be a middle six at least, maybe 1RW (hoping for a 50 a point mark stone type). 
But buffalo analytics track record has him as a 4RW ?

How far back do we go, because I really always thought our drafting got better in the Botteril years.

[rakish]

1 minute ago, Huckleberry said:

So lets basically turn this assumption into Quinn.

My eye test of him says he will be a middle six at least, maybe 1RW (hoping for a 50 a point mark stone type). 
But buffalo analytics track record has him as a 4RW ?

How far back do we go, because I really always thought our drafting got better in the Botteril years.

Liger believes the drafting is better, I'm not as convinced. But you're getting away from the conversation. We're talking about methods of valuing analytics, and who is doing it well. My argument is that a teams ability to correct for age bias demonstrates their sophistication in analytics.

How do you know Buffalo analytics track record has him as a 4RW? And what does that mean?

[LGR4GM]

3 minutes ago, rakish said:

Liger believes the drafting is better, I'm not as convinced. But you're getting away from the conversation. We're talking about methods of valuing analytics, and who is doing it well. My argument is that a teams ability to correct for age bias demonstrates their sophistication in analytics.

How do you know Buffalo analytics track record has him as a 4RW? And what does that mean?

The 2nd component for adjusting age is how old you are for your draft class. I worry about Quinn because of it. 

Idk if 2020 was better because they only had 2 picks that really matter. 2019 and 2017 I think it was showed promise.

[Huckleberry]

45 minutes ago, rakish said:

Liger believes the drafting is better, I'm not as convinced. But you're getting away from the conversation. We're talking about methods of valuing analytics, and who is doing it well. My argument is that a teams ability to correct for age bias demonstrates their sophistication in analytics.

How do you know Buffalo analytics track record has him as a 4RW? And what does that mean?

I don't know, just meant to say if our track record is bad, and we don't know what to do, he might turn out to be a 4RW ? 

[Marvin]

<math_stats_geek>

There really are multiple questions for analytics supporters and skeptics to look at which this thread touches upon:

1. Prerequisite questions

A. What data are being collected and quantified?

B. What are the data you collected supposed to be telling you?

C. How did you demonstrate that the data you collected actually tell you what you say it does?

D. How did you demonstrate that the data you collected actually matter to winning or losing?  Championships?  (Based on my research, "team built for regular season but not the playoffs" is statistically demonstrable.)

E. Can you find your blind spots, biases, etc. and how do you account for them?  For instance, how do you recognise good players on bad teams even though they might have mediocre analytics?  For players who do badly, is there a reasons to expect them to improve?  Are there times when a bad player could have good numbers?

2. Then, once you answer those questions, you can get around to the main point of the thread.

A. Who are the analytics darlings who are bad and how do you recognise them?

B. Who are the analytics busts who are good and how do you recognise them?

</math_stats_geek>

I argue that questions 2.A. and 2.B. are largely subjective unless you are quantifying something most people are not accounting for.  Moreover, I think that it is fair to argue more than numbers.  The statistics which the NHL has kept over the years don't fairly evaluate players, which is why we need ways to make adjustments.  Dave "Cementhead" Semenko scores 20 goals on Gretzky's wing.  For years, Boston's young D had the good luck to be paired with Bobby Orr and then Brad Park.  Even though we have more numbers, we still must recognise their limitations.

I am a numbers guy and am guided maybe a little to much by the analytics.  Even with all that bias, I myself argue, "when all other things are equal, a little more will beats a little more skill."  And I believe that whole-heartedly.  (It is one of the reasons I actually feel pretty good about GMKA and HCDG.)  Of course, if I knew how to quantify "will versus skill," I would be a multi-millionaire.

Edited July 11, 2021 by Marvin, Sabres Fan

[triumph_communes]

Scandella was one of the most egregious examples

[JoeSchmoe]

Here's an honest question for the diehard analytics guys out there.

Ask yourself what the 5 most important analytic quantities are. CF%? XGF%? Etc?

Then... How did Moulson and Okposo fare for those measurement points with NYI compared to Buffalo? With an accurate measurement point that correctly assesses a player, there should be no appreciable drop off considering both players came to Buffalo in their primes.

I think a lot of analytics need to be taken with a grain of salt and looked at with greater context. Its a tool... but it still takes human reasoning to interpret the numbers and make non-linear decisions on what to look at in a player vs what not to consider.

2 minutes ago, JoeSchmoe said:

 

 

Edited July 11, 2021 by JoeSchmoe

[pi2000]

42 minutes ago, JoeSchmoe said:

Here's an honest question for the diehard analytics guys out there.

Ask yourself what the 5 most important analytic quantities are. CF%? XGF%? Etc?

Then... How did Moulson and Okposo fare for those measurement points with NYI compared to Buffalo? With an accurate measurement point that correctly assesses a player, there should be no appreciable drop off considering both players came to Buffalo in their primes.

I think a lot of analytics need to be taken with a grain of salt and looked at with greater context. Its a tool... but it still takes human reasoning to interpret the numbers and make non-linear decisions on what to look at in a player vs what not to consider.

 

TRpm tho 🙃

[LGR4GM]

7 hours ago, JoeSchmoe said:

Here's an honest question for the diehard analytics guys out there.

Ask yourself what the 5 most important analytic quantities are. CF%? XGF%? Etc?

Then... How did Moulson and Okposo fare for those measurement points with NYI compared to Buffalo? With an accurate measurement point that correctly assesses a player, there should be no appreciable drop off considering both players came to Buffalo in their primes.

I think a lot of analytics need to be taken with a grain of salt and looked at with greater context. Its a tool... but it still takes human reasoning to interpret the numbers and make non-linear decisions on what to look at in a player vs what not to consider.

 

This isn't accurate. We know that linemates, systems, teams impact metrics. A players metrics fall on a bell curve, they aren't static and context matters. Analytics are an important tool just like watching the games. It's how you don't do dumb stuff like trade anything for Frolic.

[rakish]

2 hours ago, LGR4GM said:

This isn't accurate. We know that linemates, systems, teams impact metrics. A players metrics fall on a bell curve, they aren't static and context matters. Analytics are an important tool just like watching the games. It's how you don't do dumb stuff like trade anything for Frolic.

What's a bit spooky is that you made this comment while I was making the video, which addresses the same issues.

[Brawndo]

Colin Miller in Vegas, then He became a Sabre

[Marvin]

6 hours ago, Brawndo said:

Colin Miller in Vegas, then He became a Sabre

Consider this the Rasmus Ristolainen argument: one of the top problems with analytics is accounting for usage.  Analytics, as we understand it, presumes that players are performing optimally at the function they have been given.  Clearly, that does not always happen.  For all we know, Colin Miller will revert to form with the proper usage.

[Thorny]

On 7/11/2021 at 12:21 AM, JoeSchmoe said:

Here's an honest question for the diehard analytics guys out there.

Ask yourself what the 5 most important analytic quantities are. CF%? XGF%? Etc?

Then... How did Moulson and Okposo fare for those measurement points with NYI compared to Buffalo? With an accurate measurement point that correctly assesses a player, there should be no appreciable drop off considering both players came to Buffalo in their primes.

I think a lot of analytics need to be taken with a grain of salt and looked at with greater context. Its a tool... but it still takes human reasoning to interpret the numbers and make non-linear decisions on what to look at in a player vs what not to consider.

 

Also weren't they both pretty good here at first? The aging curve would have something to say about an accurate comparison 

18 hours ago, Marvin, Sabres Fan said:

Consider this the Rasmus Ristolainen argument: one of the top problems with analytics is accounting for usage.  Analytics, as we understand it, presumes that players are performing optimally at the function they have been given.  Clearly, that does not always happen.  For all we know, Colin Miller will revert to form with the proper usage.

See also: Jeff Skinner