Microchips Being Inserted In Pucks Soon

[Tondas]

Any way this chip can determine if a puck was deflected above the crossbar?

[Taro T]

43 minutes ago, Tondas said:

Any way this chip can determine if a puck was deflected above the crossbar?

Probably.  But the equipment necessary to get the to that level of precision would be more expensive than the league would be willing to spend.

Same answer for whether they can depend if the puck has, in its entirety, crossed the goal line.  Though that would be more doable.  You're looking at measuring the location within a much smaller space.

[LGR4GM]

12 hours ago, erickompositör72 said:

So much of this is really more a result of specific coaching decisions/philosophy. Analytics cannot be thought of as some sort of tool to optimize a team in a vacuum. There are way too many variables. We rely on coaches because they see nuances in players and utilize them with their visions.

In order for analytics to really become a helpful tool, we know we need large sample sizes, and specifics points of comparison. It seems like many analytic-happy posters here would prefer this whole season would just be another experiment for comparing different scenarios they fancy, when the reality is, Krueger is trusting his instincts/experience to do what he thinks will help us win.

Analytics are useless in small sample sizes, so they won't be relevant to this team until they've played much longer under Krueger, with poignant comparisons of specific situations/usages.

I mean, that just isn't true at all. We have 50 games with Krueger. Every single player was on the team last year so we can compare back to that. In your scenario you are isolating Krueger as the main variable. That's fine but that isn't the entire picture. Ignoring the past because they haven't played under Krueger is just a bad way to do anything and sounds like the company line the Sabres keep using. "well so and so is new so let's see how player x is under them before we do anything" 

 

To the italics. This team isn't winning. I like Krueger actually but some of his decisions have been obvious mistakes. I'll give him credit for Risto, he stopped making Risto do things he was bad at. I'll take away something from him because he saddled Skinner with Sobotka. Offense died every single time it touched Sobotka. Skinner didn't score because Sobotka was on his line but because his shooting percentage was up and he was creating well with Johansson. Those are all analytic observations. Just like last year, after 10 games it was painfully obvious that Tage was not ready at all. After 20 games you couldn't deny it. After 40 games it became a joke. Now Krueger was not here for that and thankfully Tage started in Rochester but you don't need a full season. 30-40 games will give you a good feel for a team. We have had more than that and we can see what this team is.

This team lacks depth scoring. They need a #2 center (don't need analytics to tell you this but you could have used analytics to know in about 2.5 seconds that trading ROR for Sobtka and Berglund was inherently stupid). The defense is actually fairly decent at goal suppression. The forwards do not go to certain areas of the ice regularly in which most NHL goals are scored, analytics helps us see that. 40+ games is not a small sample size. 

No one is arguing for experimenting. They are arguing for data driven decision making. You are arguing for experimenting. Krueger trusting his instincts is experimenting. Looking at the numbers to know that Skinner needs help and that Lazar won't keep scoring at his current rate is just using the data to get ahead. 

Edited January 28, 2020 by LGR4GM

[shrader]

13 hours ago, Crosschecking said:

I hated that glowing puck. It was such a distraction from the rest of the game.

But the fighting robots after each goal?  Now those were awesome.  That was the only thing that made it worthwhile to watch an all star game.

11 hours ago, Taro T said:

Probably.  But the equipment necessary to get the to that level of precision would be more expensive than the league would be willing to spend.

Same answer for whether they can depend if the puck has, in its entirety, crossed the goal line.  Though that would be more doable.  You're looking at measuring the location within a much smaller space.

Even with that, there's always going to be some level of gray area when it comes to using this stuff for location.  Depending on where the chip is, a puck that enters the net flat vs. one on edge, they have to reach a different point to guarantee the whole puck is over the line.  Then there's the pesky little issue of determining when the whistle blows or was intended to be blown.  That and if they ever did come up with any system that could do this stuff, I wouldn't trust the new england's of the world where it mysteriously doesn't work correctly at a key moment.  I'm fine with keeping the human element for these things.

[SwampD]

I would love to see a study of team specific use of analytics and how it relates to winning.

My guess is that whoever the analyst was who did the report, would burn it before anyone else saw it.

?

[Randall Flagg]

19 minutes ago, SwampD said:

I would love to see a study of team specific use of analytics and how it relates to winning.

My guess is that whoever the analyst was who did the report, would burn it before anyone else saw it.

?

I'm gonna look for that list that compiles each team's "analytics" employees. Then you can just skim it and compare the size of the department to the team's recent success

[Randall Flagg]

Found it:

Of course, this list gives us zero information about the nature of the work these guys do, and how their GMs and coaches apply it 

Edited January 28, 2020 by Randall Flagg

[darksabre]

I gotta be honest, Flagg, I don't see anything interesting there. N/A just won a Cup with the Blues lol

[Randall Flagg]

Just now, darksabre said:

I gotta be honest, Flagg, I don't see anything interesting there. N/A just won a Cup with the Blues lol

I agree that it's completely useless. But it's closer to Swamp's idea than anything else we could ever have access to, and was jogged into memory when I read his post lol

The insufferable parts of hockey analytics twitter regularly post this updated chart with the importance of a normal person keeping track of their bills in a spreadsheet or something it's kinda silly 

[shrader]

15 minutes ago, darksabre said:

I gotta be honest, Flagg, I don't see anything interesting there. N/A just won a Cup with the Blues lol

I'm interested in digging up more info on this Pending Hire character.

[Randall Flagg]

On 1/27/2020 at 9:06 AM, LGR4GM said:

Mitts being up to start the season, Sobotka on line 2, not getting Skinner a center or putting him with Eichel, not moving Reinhart to the line with Skinner, starting Hutton as much as they did while things started to go south, not seeing the issues with the pp, defensive pairs, using Bogosian at all, and how they draft. 

That's just off the top of my head but all of these things were decisions that data said was bad. Mitts was clearly not ready and eye test data or analytics showed that. Sobotka was a black hole on that 2nd line and we might have gotten lucky that he was injured. Skinner isn't going to produce on his own and needs a set up guy. Speaking of setup guys one of the best on the team is Reinhart who honestly doesn't need Eichel to produce according to the data but could help Skinner produce. Hutton was bad for longer than he should have been before they pulled the plug there but this can be given a pass. The PP should be run below the goal line, anyone in 2020 who things you should try to get some point shots tipped is out of their minds. There are mountains of evidence that certainly players work better in pairs than others but the merry go round goes on. Bogo has god awful metrics and yet we still get to watch him suck each night. I have explained my view on the draft in multiple places. 

 

On 1/27/2020 at 12:32 PM, LGR4GM said:

I don't want to wade through the numbers. Specifically though Sobotka was bad, very bad offensively with Skinner and Johannson. He was the offensive blackhole on that line. 

It isn't about faulty analytics it is about not using them.

I think your last sentence in conjunction with the first post paints a false dichotomy, as far as what analytics are and what information they provide. 

Where do the analytics say that Mitts should have started in the AHL? There are no analytics on what potential offseason development can do to a player his age, and as far as I'm aware, if you were to list the top 13 forwards on our roster in early October by the previous season's analytics, he'd be in the top 12 and would thus make the team. Now, of course I agree that he should not have been in the NHL this year or last year, with the benefit of hindsight (though I've always been more lukewarm on him and on his presence in the NHL tthan the average Sabre fan) but there isn't some set of analytics out there that presents the binary decision to you, and gives you the rock solid answer, about Mitts' situation or literally any other hockey decision that you highlight. 

Another example would be Carter Hutton - I don't think their decision to not waive Hutton has anything to do with "not using analytics." First of all, nobody needs analytics to give a full analysis of Carter Hutton's season, and I'm unaware of goalie analytics that are significantly more useful/telling than traditional measures of goaltending, relative to some advanced stats we have for skaters. But Carter has been bad enough that the goals allowed column in his list of starts, by itself, is enough for a non-hockey fan to tell you that he shouldn't be playing. Which obviously means that there are other reasons Carter still gets used. On the head coaching level, they are clear - Carter generally only plays now when we have back to backs, and he's balancing the fact that starters traditionally play worse in a back to back second game than even the numbers Hutton has this year, and so it's worth taking the risk to start Hutton and hope you get a game like @Toronto a few weeks back. General workload concerns for Ullmark will give his backup starts no matter who the backup is, and this has nothing to do with an organization's investments in and abilities with analytics. And on the GM level, I'm sure Jason is well aware that Hutton is trash, and while I'd love for him to waive Hutton and bring up someone else like Hamburglar or something, it's naive to assume that we have a clear answer available and it's really that easy to fix and will definitely work, while doing no damage to contracts, cap, AHL team, player good will etc, and the only thing stopping us from doing this is that we don't use analytics. Everyone and their mother knows Hutton sucks and they are making the decision to keep the organization goaltending ladder stable as is for the benefit of UPL, Johansson, and Ullmark, while hoping that Hutton can rebound to where he was in the first month or so of the season. I disagree with this decision, but I also disagree with the idea that it's an example of faulty analytics use.

What on earth do the analytics purport to claim about draft choices? As far as I know no relevant hockey advanced stat is used in the leagues these kids are coming out of, and they'd be far less useful when the goal is to figure out which of these kids will be great NHLers in 5 years than it is to describe already established players in stable and well-known NHL environemnts now, which advanced stats are still incredibly vague and mediocre at doing anyway, even if they do beat out other traditional stats

I'd never have brought Sobotka back this season, but he didn't hurt Skinner and Johansson's production, and nobody we've put with them since has done a better job at making that line do its thing, even if they've improved its underlying metrics. Jeff and Marcus are a +2 with Vlad (if you don't like the way I presented that info, then I'll give the same information this way - they have a 57 GF% with Vlad) and are -6, or 12.5 GF% without him. Now, it's true that he was an expected goal anchor for them, and their metrics got better as soon as he got hurt, and I even believe that long term there is no way that Vlad in a top six role is sustainable, because he also did well in that same position for us through November 2018, before the cliff that came after that for him and for us during his shifts, but Ralph's decision to keep putting out a line that keeps winning its matchups is not an objectively wrong decision, and it is not the result of an inescapable binary pro vs. anti analytics decision tree. These stats in general are so touchy and take so long to settle into their peak (sometimes borderline useless) predictive values that getting mad about lineup choices on this basis has basically no merit, and the idea that they should simply go with the best advanced stats lineup every night and would end up with a good record is fallacious, because it would in turn fundamentally change roles and usage to the point that we can't even trust the initial analytics anymore.

On top of that, which analytics should we use? Because 90-53-17 had worse Corsi and expected goals, but better shots and high danger chances and actual goals than 90-53-anyone else. So which one is the right decision for Thursday night versus Arizona? 

The Sabres don't suck because of an analytics problem, they suck because they are incapable, relative to other teams in the league, of finding and/or developing NHL talent, and building it into a cohesive team that can come out victorious over other NHL teams about 54% of the time over the course of a ***** ton of games. Not because they refuse to look at a player's expected goals percentage (I guarantee you they see all these stats and more that we've never dreamed of every time they make any roster decision). They just do too many things wrong or badly, compared to the amount that even good organizations will inevitably goof up. Detailing everything both good and bad about our organization would far exceed the scope of hockey analytics in depth and complexity and none of us would be capable of doing it alone

This post got a lot bigger than I intended it and is no longer directly a response to Liger, just points being made in general that outline why I get so sick and tired of the discussion of hockey analytics from all angles. There are obv ways to use analytics that will give you more insight to things you didn't get to see yourself than trad stats. I'd love it if the hockey stats fan community stuck to using them that way, rather than doing what they actually do. Which is use a RAPM chart as if the first and second and third standard deviations grow directly proportionally with the player's standing/competence in that stat, and scrape NHL data for individual games, post the charts on twitter, and insinuate their own superior hockey knowledge because of this, despite the extent of their written conclusions being "x was pretty good" and "y did a bad job tonight." (coughcoughChadcoughsorrybudbutit'sinsufferable)

I mean really, the reason we prefer corsi to +/- is literally entirely because of an increase in sample size, telling me the expected goal split of the second period of the game and asserting it tells me whether the Sabres were good or bad in that period literally blows up the entire point of the move in the direction of better stats in the first place

 

Edited January 28, 2020 by Randall Flagg

[LGR4GM]

Analytics are not looking at 1 number and comparing it to the next. They are looking at all the numbers holistically and combining that with experience and anecdotal evidence to reach the correct conclusion. The Sabres do not do that regularly. 

Also I fully disagree on your statement about analytics and drafting. Just because we aren't using xGF and Corsi for draft picks doesn't mean we shouldn't and doesn't mean there are not comparisons and numbers we can use. Here is a good example: https://canucksarmy.com/2017/03/30/casey-mittelstadt-and-the-canucks/ 

To often we consider analytics to be the fancy stats that we use a lot. That isn't what analytics is. It can use them and be based around them but analyzing the numbers and drawing correct conclusions is also part of the gig. 

"However, this is precisely why you combine stats and eye test to come to the best conclusion possible." This line from that Canucks Army page I found to be pertinent. Sadly Canucks does not do what they did a few years ago. I think one of their primary data guys was hired elsewhere. 2018 profiles compared to 2019 are different (https://canucksarmy.com/2018/06/15/canucksarmys-2018-nhl-draft-profiles-1-rasmus-dahlin/ v. https://canucksarmy.com/2019/06/18/2019-draft-countdown-no-7-dylan-cozens/ )

[Radar]

19 hours ago, Zamboni said:

Um .... what?

Unwatchable

[Weave]

4 hours ago, Randall Flagg said:

 

I think your last sentence in conjunction with the first post paints a false dichotomy, as far as what analytics are and what information they provide. 

Where do the analytics say that Mitts should have started .......

 

That was quite the novel.  But I enjoyed reading it.  It said some things I couldn't quite put words to.

Not to further derail the subject but, I think your points fit well with what I think Ralph's primary motivation was for putting Sobotka on the 2nd line, having a vet on the line that knows where he should be, has been under pressure situations, and most importantly had predictability to the details of his game.  In Ralph's words, a grown man.  That 2nd line was not a second scoring line as much as it was a line capable of playing under pressure.  And the results of that line with and without Sobotka appear to bear that out right now.

Of course Sobotka is not a 2nd line player.  That he's been the best choice so far speaks to the roster makeup far more than it speaks whether the coach uses metrics to make his decisions.  I expect Frolik to be an improvement while providing the same sort of game Ralph was looking for out of Sobotka.

[Curt]

4 hours ago, Randall Flagg said:

What on earth do the analytics purport to claim about draft choices? As far as I know no relevant hockey advanced stat is used in the leagues these kids are coming out of, and they'd be far less useful when the goal is to figure out which of these kids will be great NHLers in 5 years than it is to describe already established players in stable and well-known NHL environemnts now, which advanced stats are still incredibly vague and mediocre at doing anyway, even if they do beat out other traditional stats

I agree with a lot of what you said, but I want to knit pick a little.  People, and perhaps teams, do track things like passes into danger areas, shots from danger areas, controlled zone entries, and a bunch of other stuff for prospects.  You want to find players who are already successfully doing some of the things that you value highly from your NHL players.

It’s not that any one statistic is going to guarantee NHL success, but stuff like this is useful information to consider alongside traditional scouting.  What Buffalo in particular looks at, I could only guess.

[Taro T]

40 minutes ago, Weave said:

That was quite the novel.  But I enjoyed reading it.  It said some things I couldn't quite put words to.

Not to further derail the subject but, I think your points fit well with what I think Ralph's primary motivation was for putting Sobotka on the 2nd line, having a vet on the line that knows where he should be, has been under pressure situations, and most importantly had predictability to the details of his game.  In Ralph's words, a grown man.  That 2nd line was not a second scoring line as much as it was a line capable of playing under pressure.  And the results of that line with and without Sobotka appear to bear that out right now.

Of course Sobotka is not a 2nd line player.  That he's been the best choice so far speaks to the roster makeup far more than it speaks whether the coach uses metrics to make his decisions.  I expect Frolik to be an improvement while providing the same sort of game Ralph was looking for out of Sobotka.

Except the line had already fallen off prior to Sobotka breaking.  And IF he really was the straw that stirred that drink, they'd be bringing him back next year.  He has as much of a chance of being brought back as an UFA as Pominville had.  Though we'll likely have people around here calling for him being re-signed in the off-season; just like we did w/ Jason.  (Though mercifully we'll only have a couple of posters suggesting that rather than as many as we got looking for the return of Pomms.)

Edited January 28, 2020 by Taro T

[Weave]

24 minutes ago, Taro T said:

Except the line had already fallen off prior to Sobotka breaking.  And IF he really was the straw that stirred that drink, they'd be bringing him back next year.  He has as much of a chance of being brought back as an UFA as Pominville had.  Though we'll likely have people around here calling for him being re-signed in the off-season; just like we did w/ Jason.  (Though mercifully we'll only have a couple of posters suggesting that rather than as many as we got looking for the return of Pomms.)

Noone is suggesting there isn't room for improvement.  I'm pretty sure Frolik is going to be improvement.  And noone is suggesting he was some magical line chemistry elixir. that we'll be pining for next season.   It's nothing more than, he was able to play a role that the line was put together to play better than the accumulated ***** that was put on that line afterwards.  Even after the fall off that was apparent.

[Taro T]

6 minutes ago, Weave said:

Noone is suggesting there isn't room for improvement.  I'm pretty sure Frolik is going to be improvement.  And noone is suggesting he was some magical line chemistry elixir. that we'll be pining for next season.   It's nothing more than, he was able to play a role that the line was put together to play better than the accumulated ***** that was put on that line afterwards. Even after the fall off that was apparent.

Agree that the line should perform well with Frolik.  But we already have posts here (not saying you were necessarily doing it yourself) pining for Sobotka to be back on that line. 

And do not catch what your getting at with the bolded.

[Weave]

5 minutes ago, Taro T said:

Agree that the line should perform well with Frolik.  But we already have posts here (not saying you were necessarily doing it yourself) pining for Sobotka to be back on that line. 

And do not catch what your getting at with the bolded.

I'm pretty sure the pining for Sobotka is relative to what's been on that line up until Frolik.  If you are reading more into it than Sobotka was the best of a bad lot then  I think you are reading too much into it.

The bolded was my attempt to say even after the falloff Sobotka was better than what followed him.

[Taro T]

21 minutes ago, Weave said:

I'm pretty sure the pining for Sobotka is relative to what's been on that line up until Frolik.  If you are reading more into it than Sobotka was the best of a bad lot then  I think you are reading too much into it.

The bolded was my attempt to say even after the falloff Sobotka was better than what followed him.

Thanks for the clarification.  But, while that line worked better when all 3 were healthy than after Johansson was playing hurt, that dropoff had FAR more to do with MoJo's injury than Sobotka's absence.

[Randall Flagg]

On 1/28/2020 at 6:09 PM, Weave said:

That was quite the novel.  But I enjoyed reading it.  It said some things I couldn't quite put words to.

Not to further derail the subject but, I think your points fit well with what I think Ralph's primary motivation was for putting Sobotka on the 2nd line, having a vet on the line that knows where he should be, has been under pressure situations, and most importantly had predictability to the details of his game.  In Ralph's words, a grown man.  That 2nd line was not a second scoring line as much as it was a line capable of playing under pressure.  And the results of that line with and without Sobotka appear to bear that out right now.

Of course Sobotka is not a 2nd line player.  That he's been the best choice so far speaks to the roster makeup far more than it speaks whether the coach uses metrics to make his decisions.  I expect Frolik to be an improvement while providing the same sort of game Ralph was looking for out of Sobotka.

I like the thought. I have to say, though, Frolik was among the least impressive things about my first in-person experience with the team this season. But in principle it makes sense.

Edited January 30, 2020 by Randall Flagg

[Randall Flagg]

On 1/28/2020 at 2:08 PM, LGR4GM said:

Analytics are not looking at 1 number and comparing it to the next. They are looking at all the numbers holistically and combining that with experience and anecdotal evidence to reach the correct conclusion. The Sabres do not do that regularly. 

Also I fully disagree on your statement about analytics and drafting. Just because we aren't using xGF and Corsi for draft picks doesn't mean we shouldn't and doesn't mean there are not comparisons and numbers we can use. Here is a good example: https://canucksarmy.com/2017/03/30/casey-mittelstadt-and-the-canucks/ 

To often we consider analytics to be the fancy stats that we use a lot. That isn't what analytics is. It can use them and be based around them but analyzing the numbers and drawing correct conclusions is also part of the gig. 

"However, this is precisely why you combine stats and eye test to come to the best conclusion possible." This line from that Canucks Army page I found to be pertinent. Sadly Canucks does not do what they did a few years ago. I think one of their primary data guys was hired elsewhere. 2018 profiles compared to 2019 are different (https://canucksarmy.com/2018/06/15/canucksarmys-2018-nhl-draft-profiles-1-rasmus-dahlin/ v. https://canucksarmy.com/2019/06/18/2019-draft-countdown-no-7-dylan-cozens/ )

Gonna need you to tell me what the Sabres do instead. I can't just take your word here - what do they do, what should they do more of, and what should they do less of? 

As an applied example, lead me down the analytics path that says Ralph was wrong to put together the most productive and best non-Eichel line we've seen since ROR was here? Where did they step off this path, and why is this path correct? Is it unique? Would all organizations that are good, and thus use analytics, have chosen this same path, the same way? 

Edited January 30, 2020 by Randall Flagg

[Marvin]

On 1/28/2020 at 9:08 AM, shrader said:

But the fighting robots after each goal?  Now those were awesome.  That was the only thing that made it worthwhile to watch an all star game.

Even with that, there's always going to be some level of gray area when it comes to using this stuff for location.  Depending on where the chip is, a puck that enters the net flat vs. one on edge, they have to reach a different point to guarantee the whole puck is over the line.  Then there's the pesky little issue of determining when the whistle blows or was intended to be blown.  That and if they ever did come up with any system that could do this stuff, I wouldn't trust the new england's of the world where it mysteriously doesn't work correctly at a key moment.  I'm fine with keeping the human element for these things.

I consulted for someone who was working on the goal determination and had the patent for doing this.  I think his kids sold it to the NHL when he died a few years ago.  It would not surprise me if this is an outgrowth of that work.

You can, in theory, work out precisely where the puck is if it completely crosses over the goal-line -- even if it is on-edge, angled, etc.

However, the metal in the goal is ferromagnetic, so unless you have very good proximate sensors, they distort the data slightly, so you have the grey area you mentioned.  Also, the machinery in the ice has electricity which generates magnetic fields when it varies in current, so that also affects the sensors in the puck.

Thus, if the puck crosses the line at a funny angle for a fraction of a second, you probably can't do anything.  Where it would be useful would be when the puck clearly crosses the goal line, but is not visible.  That information could be transmitted within the televised signal to Toronto and to both teams' representatives at the game.

[Weave]

55 minutes ago, E4 ... Ke2 said:

I consulted for someone who was working on the goal determination and had the patent for doing this.  I think his kids sold it to the NHL when he died a few years ago.  It would not surprise me if this is an outgrowth of that work.

You can, in theory, work out precisely where the puck is if it completely crosses over the goal-line -- even if it is on-edge, angled, etc.

However, the metal in the goal is ferromagnetic, so unless you have very good proximate sensors, they distort the data slightly, so you have the grey area you mentioned.  Also, the machinery in the ice has electricity which generates magnetic fields when it varies in current, so that also affects the sensors in the puck.

Thus, if the puck crosses the line at a funny angle for a fraction of a second, you probably can't do anything.  Where it would be useful would be when the puck clearly crosses the goal line, but is not visible.  That information could be transmitted within the televised signal to Toronto and to both teams' representatives at the game.

Vision systems today are stupid good.  There is no reason why it couldn't be supplemented with a good vision system.  You wouldn't want to rely on line of sight for everything, but between the two I bet you'd have better results than the officials and video replay.