2014 Sabres Training Camp

[Taro T]

Brad Boyes would be an interesting case study in analytics IMHO.

 

He's a guy that plays veerry well when teamed w/ top level talent - the top level guys definitely seem to raise his game substantially. But he is mediocre / sub-mediocre when teamed w/ average guys.

 

The q for the analytics guys is, how much is his game upgrade worth relative to the expected downgrade in the game of the studs he plays w/ downgraded by playing w/ him especially when compared to playing him w/ average guys?

 

The q for the 'eyeball' guys is, does playing a guy like Boyes w/ your top players simply look too wrong even if they're being successful?

 

If Boyes hits too close to home, having never really played w/ the Sabres' 'top line,' then substitute Rob Brown who was very good playing w/ Mario and very sub/average playing w/ most anybody else.

[I am Defecting]

You know, I believe in the analytics movement but I love this argument. Hope you don't mind if I use it in future conversations.

 

Please do. As always, I enjoy reading the conversations here. This is a unique place. :wub:

[shrader]

The mistake is thinking that one of these stats generalize to a single moment. Take one of the most simple stats out there. A good goalie doesn't let 93% of the puck into the net on every shot. These things are used to measure or predict performance over the long term. Why wouldn't you want to explore something like this if it gives you even a slightly better chance of acquiring the right player?

 

Also, just because a team utilizes them doesn't mean a fan has to pay any attentuon to that while watching a game.

[K-9]

Unfortunately, we see advanced stats used to try to prove that a mediocre player is actually a good one way more often than we see it showing us a player that we already knew was good, was actually good.

 

I agree with this. I can't think of a player that wasn't considered good become good because of his favorable analytics. If a player is good, his stats will usually back that up over the long term. In the short term and in the absence of those established analytics, I'll trust a scout's and coach's ability to assess his potential based upon their criteria which usually includes their opinion that a kid can play or not based upon what they see on the ice.

 

GO SABRES!!!

[...]

The article doesn't say, or imply, that Nolan is outright rejecting stats. The article is based on a handful of comments made by Nolan in what context, or in response to what statement or question, none of which are revealed. The article is written by Paul Hamilton.

 

For all we know Nolan could have qualified the comments with something like "yeah, I look at the information the stats guys give me...", or "such and such likes to go over the stats and he tells me this or that about the players...".

 

The hand-wringing is ridiculous. Nolan has never been known to get into detail when talking to reporters. For all we know, thanks to the statistician(s) the Sabres have on staff, they may not want to tip their hand about what they're doing.

 

Regardless, even if Nolan himself prefers to rely on unscientific data, he has a staff he consults, and more than likely at least one of the staff provides the data in a summarized, efficiently comprehendible manner that Nolan would use to help make coaching decisions, not unlike the subjective observations he makes. To take Paul Hamilton's article and deduce Nolan must not use stats in some manner at all is highly dramatic.

Edited September 23, 2014 by sizzlemeister

[SwampD]

Internalism vs Externalism. It's two different ways of constructing knowledge (epistemology). The internalist would argue that a capable mind can best determine what is true. The externalist would argue that the advanced stats would best tell us what is true. The internalist would argue that advanced stats are not capable of the nuanced and subtle decisions that the human brain is capable of. The externalist would argue that the human brain is imperfect and subject to delusions, which would be overcome by giving priority to advanced stats.

 

The best chess playing computer can analyze the possible decisions better than the best chess playing human. The millions of variations that one might encounter in a chess game are best solved by a computer, basically using advanced stats to determine the best possible move in every possible situation. Still, pawn to e4 is always pawn to e4. Ovechkin on a regular shift is sometimes happy, sometimes sad, sometimes motivated, sometimes not. There is no telling through advanced stats exactly what you'll get out of a hockey player in a given situation. Hockey is much more complex.

 

If you punch all of the raw data available into the computer that you can possibly think of, and do what the stats tell you every time, I believe that a smart unconventional coach will beat you every time. What coaching decisions will be made using advanced stats, that couldn't also be made through keen observation? Sorry, I'm rambling, but we're a long ways away from the primacy of advanced stats, because there is always, currently, a human interpreting the stats, which is no better and no different than a coach with some knowledge interpreting what he sees with his own eyes, at training camp, for example.

This is a great post.

[Johnny DangerFace]

 

 

What if a player is signed for excelling with certain specific usage, and then the coach puts him in a situation to fail? I'm thinking a player puts up big numbers getting a ton of power play time and matches against weak opponents (the defensive equivalent of Hodgson, for example), then the coach on the new team takes that away and makes him kill penalties and play against the Crosbys of the world. Good player turns into crappy player pretty quick. The entire organization needs to be on the same page with this stuff.

 

Then that sounds like a communication problem. GMs and scouts need to find a way for stats to help them and I think they have a strong place in the front office. However, a GM should also acquire people that he knows fits in the coaches style and will be used right. That's on both the coach and GM to do.

 

But again, I don't fully see why the coach should invest much into stats for a lot of reasons. One, that's not sonething he has ever done. He has become an nhl coach for a reason, he needs to stick to it. The last thing I want is a coach not doing what he knows best because of stats. Also, stats are very powerful, but also very confusing. It takes time and skill to use them property. With stats it's all about context and usually other stats. If you don't properly understand how to use it correctly, it won't do you any good. Having an old school coach try to adapt to it could he a disaster.

 

So IMO, a front office should use stats, as much as possible. I'm new school when it comes to it. And a GM needs to communicate with a coach the strengths and weaknesses if a player and why he was brought in. However, as far as nolan being aware if any stats, applying stats, keeping up with stars, etc. I don't think it's required.

 

Again I could be missing something and it going over my head, but I don't think this is an issue in the slightest

[wjag]

This quote from Nolan has the analytics crowd in a twitter frenzy

 

 

 

 

Link to the story

 

http://www.wgr550.co...ytics-/19956968

 

Smart phones, calculators, computers, don't get me started. I rely on my good old land line, abacus and eyes to get the job done. I put one marble in my right pocket for a face off won and one in my left pocket for a face off lost. It's fool proof. We all know that the only thing computers are good for are playing games and ordering pizza.

[Eleven]

:lol:

Burgers and a little Kraft Dinner is your go to meal without much in the cupboard, but when the GMTM is buying filet mignon and finely aged Stanton cheese, we might need a new sous chef.

 

Whoa, look who's buying the brand name mac n cheese!

 

Seriously, though, I agree with you and TW. I don't think Ted Nolan will be the guy behind the bench when this team turns the corner. I like him, and I never thought he should have been fired the first time around, but this team is going to need someone with a system.

[Campy]

The article doesn't say, or imply, that Nolan is outright rejecting stats. The article is based on a handful of comments made by Nolan in what context, or in response to what statement or question, none of which are revealed. <snip>

 

I can't say I disagree with this part of his remarks: "In modern day hockey there's a lot of emphasis put on analytics and how much possession time, but you can never underestimate the human aspect of the game. Hockey's not all that difficult, there's a science to it, but it's not rocket science."

[SwampD]

Whoa, look who's buying the brand name mac n cheese!

 

Seriously, though, I agree with you and TW. I don't think Ted Nolan will be the guy behind the bench when this team turns the corner. I like him, and I never thought he should have been fired the first time around, but this team is going to need someone with a system.

I just threw up in my mouth a little.

[nfreeman]

Whoa, look who's buying the brand name mac n cheese!

 

I made baked mac'n'cheese for the first time a few weeks ago as a side dish for ribs. With enough cheese on top, it browns and forms the crispy edge. Freaking fantastic.

[darksabre]

 

 

I made baked mac'n'cheese for the first time a few weeks ago as a side dish for ribs. With enough cheese on top, it browns and forms the crispy edge. Freaking fantastic.

 

Baked Mac and Cheese is the only way.

[Wraith]

I think the notion that a coach doesn't have to understand analytics at all is a fallacy. Yes, a big part of the current work in most sports is player evaluation and thus handled by the General Manager, not the coach. But an underrated aspect of sports analysis is evaluating tactics.

 

Since the sports data revolution started with baseball, there is a great example about how analytics have not infiltrated the managers the way they have the front office. This has a profound effect on day-to-day tactics. Managers will, almost every day, refuse to use their best relief pitchers in the highest leverage (most important) game situations because it's not a "save situation." Team's regularly blow leads in the sixth, seventh, or eighth inning with an inferior relief pitcher on the mound while their closer, inarguably the best reliever on the team, watches because it's not a save situation yet. It's infuriating. It's especially infuriating because baseball, with it's rich history and incredibly large sample size, has a system that can tell you, in real time, when the high leverage situations are! Having your best relief pitcher enter the game at the start of the ninth inning with nobody on base is almost never the best use of his skills. But the current generation of managers have grown up being told that is how you use a bullpen and so it remains.

 

Now to hockey. One tactical situation that is starting to get noticed by the stat heads in the dump-in versus the carry-in entry into the offensive zone. Willfully turning the puck over to the opponent at a very high rate using the dump and chase method is incredibly inefficient compared to the carry-in.This article sums it up nicely:

 

http://www.si.com/nh...-chase-strategy

 

Can any of you image a traditionalist like Nolan, who is all about the effort, would ever advocate for a lower rate of dump and chases?

Edited September 23, 2014 by Wraith

[That Aud Smell]

Can any of you image a traditionalist like Nolan, who is all about the effort, would ever advocate for a lower rate of dump and chases?

 

If his eyes, heart, and soul tell him that it makes better sense to carry the puck in, then that is what his team will do. :angel:

[MattPie]

If you punch all of the raw data available into the computer that you can possibly think of, and do what the stats tell you every time, I believe that a smart unconventional coach will beat you every time. What coaching decisions will be made using advanced stats, that couldn't also be made through keen observation? Sorry, I'm rambling, but we're a long ways away from the primacy of advanced stats, because there is always, currently, a human interpreting the stats, which is no better and no different than a coach with some knowledge interpreting what he sees with his own eyes, at training camp, for example.

 

Even the best human is an imperfect observer. We let our personal biases take over, and once a bias starts to take hold we start emphasizing things we observe that fit into that bias. A great example is refs. Buy most measures refs are pretty unbiased, yet every fan seems to think their team is unfairly targeted. Bruins fans will rail about how the refs pick on them. Sure, we call them homers and crazy, but we're no different. I had a long conversation with a Flyers fan friend at one point about how the Campbell-Umberger hit was "dirty".

 

On a team scale, people are always going to "like" a set of people in a group more than others in the group. And once you like someone, it's hard to objectively observe them because you want to give them the benefit of the doubt when something negative happens.

[That Aud Smell]

Even the best human is an imperfect observer. We let our personal biases take over, and once a bias starts to take hold we start emphasizing things we observe that fit into that bias. A great example is refs. Buy most measures refs are pretty unbiased, yet every fan seems to think their team is unfairly targeted. Bruins fans will rail about how the refs pick on them. Sure, we call them homers and crazy, but we're no different. I had a long conversation with a Flyers fan friend at one point about how the Campbell-Umberger hit was "dirty".

 

On a team scale, people are always going to "like" a set of people in a group more than others in the group. And once you like someone, it's hard to objectively observe them because you want to give them the benefit of the doubt when something negative happens.

 

Word.

 

People who haven't dabbled a bit in cognitive bias theory would do well to do so -- that includes the coach.

 

The idea that someone in Nolan's position would lack the open-mindedness and humility to say, "I trust what I see, I believe in what I see, but I am open to seeing how all of that stacks up against the data" is disappointing.

[LTS]

Humanity is on this kick to explain everything and in order to do so they need to build a model. Ultimately, all the data scientists in the world will finally model the universe. Until then, everything will be imperfect.

 

A model is only as good as its data. The data is influenced by various factors and those factors are determined by the data scientist. The sum collection of data may be incomplete, may be misinterpreted, may be subject to bias. What is possession? One would imagine it is the amount of time a team remains in control, so how is that measured? Does a team lose possession when puck glances off an opponent stick during a pass? Why not? Is that a failing point of your model? What if I did count that and made it a category of "grade A possession / grade B possession".

 

Ultimately you can get bogged down in it and end up basically determining what a good coach can see because he's with these players day in and day out. His entire basis of experience is more than anything any data scientist has put into a model. Think about it. He sees the game a certain way and up to this point a data model has never included or tracked that factor. Coach has 20 years of experience seeing this factor day in and day out. The model either has to be manually updated to include this factor from every game over that same time period or else the data scientist has to say from here on this is important but it lacks the context.

 

I'm not a stats genius but I do firmly believe that relying solely (soully? :)) on data is also a mistake. Until the model can factor all aspects of life it will be fallible.

 

Yuri had a great point. I've watched guys playing hockey who can barely skate continually frustrate the hell out of a great skater who normally dominates. Why? Because the great skater has been taught that other great players will attempt to play a certain way. The less skilled player doesn't know this and he's doing whatever he thinks is good.

 

I think you can use stats to compare players over the long term. I think Ted Nolan knows which players play better defense, score more, take better faceoffs, etc. He's not consulting his analytics book during the game to see what instructions to bark out to Ennis while he enters a 1 on 1 with Subban.

[That Aud Smell]

Humanity is on this kick to explain everything and in order to do so they need to build a model. Ultimately, all the data scientists in the world will finally model the universe. Until then, everything will be imperfect.

 

Great post. I also really liked what Yuri had to say - fascinating stuff.

 

My point remains more or less the same.

 

I do firmly believe that relying solely (soully? :)) on data is also a mistake.

Precisely no one is arguing that #fancystats should determine what a GM or coach should do in a given situation.

 

Of course #fancystats are limited and imperfect. So too are a coach's perceptions. Maybe the latter is more valuable than the former -- in fact, that is probably so. But the former appear to have real value; to swipe them aside as a meaningless redundancy** is foolish.

 

**I'm not saying you are.

Edited September 23, 2014 by That Aud Smell

[MattPie]

Humanity is on this kick to explain everything and in order to do so they need to build a model. Ultimately, all the data scientists in the world will finally model the universe. Until then, everything will be imperfect.

 

A model is only as good as its data. The data is influenced by various factors and those factors are determined by the data scientist. The sum collection of data may be incomplete, may be misinterpreted, may be subject to bias. What is possession? One would imagine it is the amount of time a team remains in control, so how is that measured? Does a team lose possession when puck glances off an opponent stick during a pass? Why not? Is that a failing point of your model? What if I did count that and made it a category of "grade A possession / grade B possession".

 

Ultimately you can get bogged down in it and end up basically determining what a good coach can see because he's with these players day in and day out. His entire basis of experience is more than anything any data scientist has put into a model. Think about it. He sees the game a certain way and up to this point a data model has never included or tracked that factor. Coach has 20 years of experience seeing this factor day in and day out. The model either has to be manually updated to include this factor from every game over that same time period or else the data scientist has to say from here on this is important but it lacks the context.

 

I'm not a stats genius but I do firmly believe that relying solely (soully? :)) on data is also a mistake. Until the model can factor all aspects of life it will be fallible.

 

Yuri had a great point. I've watched guys playing hockey who can barely skate continually frustrate the hell out of a great skater who normally dominates. Why? Because the great skater has been taught that other great players will attempt to play a certain way. The less skilled player doesn't know this and he's doing whatever he thinks is good.

 

I think you can use stats to compare players over the long term. I think Ted Nolan knows which players play better defense, score more, take better faceoffs, etc. He's not consulting his analytics book during the game to see what instructions to bark out to Ennis while he enters a 1 on 1 with Subban.

 

There's a lot in there, but quick notes:

A: I don't think anyone would argue that the stats are perfect or that they'll always describe everything that happens in hockey. Baseball has it easier there since there's a much smaller tree of things that can happen on each play.

B: I don't think anyone is saying to rely solely (or cod-ly) on data and not watch the players. Yes, the model can be failable, but so can observation.

C: Playing unconventionally can work (seeL Hasek, Dominic), but I'm not sure if that works as well for coaching. There's a lot of smart people out there analyzing tape and any "unconventional" coaching probably has a lot more pattern to it than you'd think, just not the pattern that the other coaches are used to seeing.

D: As for analytics during the game, you're just being silly. No one is really suggesting that.

 

I'm sure there was a coach somewhere that said, "eh, watching film is imperfect, you can't see the entire playing field and you don't get the feel of the team at the time. I'm not going to use it and just rely on what I saw during the game." That sounds completely ridiculous now, and Nolan's quote at face value is in the same league.

Edited September 23, 2014 by MattPie

[SwampD]

Why does a coach need analytics? I understand why a GM might need them, but why a coach? I see nothing wrong with Nolan's view.

 

I look at it like I look at music. There are musicians who know every single note in every mode that with work over every chord. Then there are the musicians that just pick up their ax and play by feel and don't know what they are playing. They both tend to lack something when they swing to far to their respective ends of the spectrum. I tend to think that the best musicians are the ones who learn all the theory, but when it comes time, they just play.

 

I will say, though, that if I had to choose only one, give me the guy who just plays.

[That Aud Smell]

I look at it like I look at music. There are musicians who know every single note in every mode that with work over every chord. Then there are the musicians that just pick up their ax and play by feel and don't know what they are playing. They both tend to lack something when they swing to far to their respective ends of the spectrum. I tend to think that the best musicians are the ones who learn all the theory, but when it comes time, they just play.

 

I appreciate the analogy, but I don't think it's an apt comparison. The object and intended result of playing music for others is to inspire a subjective appreciation of an artistic expression; the object and intended result of playing hockey is to outscore an opponent.

 

In addition to which, if a soulful player who is a self-taught/-learned virtuoso were struggling with a particular change or whatever (I'm not a musician), and someone suggested that a technical explanation/breakdown of the progression would help him get there, and he was all "dude, no way, you're harshin' my mellow, lay off with the BIG DATA", I'd think he was passing up an opportunity to learn and grow.

 

And, again, it ain't a matter of choosing one or the other, or even exalting analytics above first-hand observations.

[Ghost of Dwight Drane]

Why does a coach need analytics? I understand why a GM might need them, but why a coach? I see nothing wrong with Nolan's view.

 

I look at it like I look at music. There are musicians who know every single note in every mode that with work over every chord. Then there are the musicians that just pick up their ax and play by feel and don't know what they are playing. They both tend to lack something when they swing to far to their respective ends of the spectrum. I tend to think that the best musicians are the ones who learn all the theory, but when it comes time, they just play.

 

I will say, though, that if I had to choose only one, give me the guy who just plays.

 

Amen. I think too many people who think analytics are some great force are people who have office jobs, or like goofing around on computers. Music is one of the true comparative fields. You can either swing, or you can't. Nolan swings like freaking Basie, and he doesn't have time to coddle a bunch of Asian string players from Juilliard to try and get them to trade 4's at the 4th of July concert at the bandshell. Give him the 50 year old with the .18 BAC and a junior high education and he'll be just fine.

 

I also love the stats article linked by Wraith. All this hubbub about how dinosaur-like dump and chase is, with the detail and breakdown, then in half a sentence it excuses the times it doesn't work to "bad defense and weak goaltending". Hey Sparky....maybe the reason the defense and goalie looks bad is because while your puss-laiden guys are wracking up possesion time on a carry entry, the times it goes bad leads to a guy getting destroyed, and the remaining partner caught flat footed. Surpise-surprise....Lindy has his own entry stats he follows.

 

I find it amusing the flack Nolan got for his comments. By definition, if you are so bright and feel that he is left behind by not using analytics because everyone uses them, aren't you proving that he is actually correct because everyone else is valuing assets on the same data, thus overvaluing players on a certain set of criteria?

 

Jimminy-Christmas....you had the modern day inventor of this movement on the coaching staff of the Sabres the past decade plus....and look at what it left you with? The biggest collection of underachieving pussbags this side of Paris, and a historic crash and burn.

[darksabre]

Amen. I think too many people who think analytics are some great force are people who have office jobs, or like goofing around on computers. Music is one of the true comparative fields. You can either swing, or you can't. Nolan swings like freaking Basie, and he doesn't have time to coddle a bunch of Asian string players from Juilliard to try and get them to trade 4's at the 4th of July concert at the bandshell. Give him the 50 year old with the .18 BAC and a junior high education and he'll be just fine.

 

I also love the stats article linked by Wraith. All this hubbub about how dinosaur-like dump and chase is, with the detail and breakdown, then in half a sentence it excuses the times it doesn't work to "bad defense and weak goaltending". Hey Sparky....maybe the reason the defense and goalie looks bad is because while your puss-laiden guys are wracking up possesion time on a carry entry, the times it goes bad leads to a guy getting destroyed, and the remaining partner caught flat footed. Surpise-surprise....Lindy has his own entry stats he follows.

 

I find it amusing the flack Nolan got for his comments. By definition, if you are so bright and feel that he is left behind by not using analytics because everyone uses them, aren't you proving that he is actually correct because everyone else is valuing assets on the same data, thus overvaluing players on a certain set of criteria?

 

Jimminy-Christmas....you had the modern day inventor of this movement on the coaching staff of the Sabres the past decade plus....and look at what it left you with? The biggest collection of underachieving pussbags this side of Paris, and a historic crash and burn.

 

Amen Ghost. Amen.

 

[spndnchz]

Blandisi, Cornel, Joly, Karabacek, Rodewald, MacKinnon, Walsh, Martin, and Brassard return to their clubs today.