The Athletic: The 2020 NHL Poll - Players have their say

[The Yen Man]

Originally Posted by Textcritic

I don't know about Marchand—he has a pretty storied history of cheap shots. But I like where you are going with this: a "dirty" player should be one who flouts the rules in an effort to gain an advantage, and I sure don't see Tkachuk doing that. He is a relentless agitator, which is quite different in my books.

Agreed, I think players are confusing dirty with aggravating player to play against. Other than that elbow to Doughty (which I think he's learned from), Chucky hasn't done anything else dirty. He just makes sure to get under people's skin, but there's nothing malicious about his actions. He's not running guys from behind or taking out their knees. He just does the cheap after the whistle convenient facewash, or the "whoops, you pushed me so I'll accidentally fall on your goalie while dragging you down with me" move.

[Calgary4LIfe]

Originally Posted by CorsiHockeyLeague

This just demonstrates that you don't understand what you're talking about.

HDSC isn't subjective at all. It's just shots from a particular area of the ice, where the highest proportion of goals are scored from. Nothing remotely subjective about it.

PDO isn't "garbage". It's just... math. NHL save percentage plus shooting percentage always equals an even 1.000. If your PDO is above 1.000, you're either saving goals at an above average rate, or scoring at an above average rate on your shots, or both. Nothing remotely controversial or subjective about that.

No, they weren't. You're deluding yourself. There's a reason they got absolutely crushed by the Ducks that season - they were a bad hockey team, the second worst team in the playoffs that year. They won a bunch of games by being the "comeback kids", and if you think that's a sustainable way to win hockey games, you're just... wrong. Comparing the Hartley Flames to the way they are now is totally night and day.

PDO IS garbage, sorry. Yeah, I understand where the math part comes from, it is just a completely stupid rationale. It is using simple math to justify something that it shouldn't. No, it doesn't always equal 1.000. It SHOULD equal how good you are at both. For instance, a good team SHOULD be over 1.000, no? A bad team will be under 1.000, no? Just because your PDO >1.000 doesn't mean you should see regression to the 'mean', and vice-versa.


Better teams SHOULD have solid goaltending and defence, correct? Better teams should also have better offensive players as well, correct? Why should they equal 1.000 and we should think of them as 'lucky' and that they are due for a regression? That makes zero sense. How about we think of them as a GOOD team. Sure, regression will eventually hit that team, but probably due to injuries, poor outings in which they are tired, or even in the long-term as teams start ageing and replacement players are of a lower quality.


So yes, PDO to me is a completely garbage stat created to explain away.... luck? Never bought it, never will.


Also, I am not 'against' metrics, and you really don't have to talk down to me here. Please go and see any actuary and he will look at CORSI and will tell you that the confidence interval is WAY too low to trust. Most risk-based industries will fire you on the spot with confidence numbers that low. This isn't to say that CORSI and the rest of the advanced metrics that we are using right now is useless - on the contrary, it still does help to explain a lot. The issue I have (and I am going to guess many people have) is that they are NOT infallible enough that you can really trust them to the point where if I question them, you get a stern talking to.


No, these metrics that you are passionately defending are not infallible, and there are always outliers - that's the basis of why they are not trustworthy enough - their confidence interval is too low.


Part of it has to do with making too many inferences. Ok, so can we agree that part of your name is kind of garbage? CORSI - what is CORSI? What I mean is what is the measurement describing? In the end, it is attempting to describe possession, correct? By using total shots + missed shots + blocked shots, at even strength 5on5, minus the opposing teams total shots + missed shots + blocked shots. In this way, you can infer possession, right? Do I have this right?


Well, now you have to count on the people tallying up these numbers have it right. Some arenas are notorious for poor counting stats - just shots on goal. Less of them to count, but they often get it wrong. "Nah, that was just a dump-in". So your data set is a little unreliable at it stands, but not (hopefully) terribly so. And then you use these numbers to state that it is showing possession, but without any regard at all to 'more patient teams' that like to pass the puck around a bit more, or teams that are way more effective on the cycle and have tonnes of zone time but generate few shots on net, etc. That isn't captured, so the numbers are further skewed.


What percentage of actual possession time is accurately captured in CORSI, do you think? Probably lower than it is made out to be.


Still, you can't argue that 'more shots' is usually more conducive to winning. Being out-shot isn't a sustainable winning strategy. So regardless of actual time of possession correlating with CORSI, out-shooting (or attempting to out-shoot) the opposition usually correlates with winning the game, right?


Except when it doesn't. Like Hartley's Calgary Flames for 1.5 seasons (and other teams that unexpectedly didn't make the playoffs, or that made it 'un-deservingly' from an analytical perspective).


The strategy that Calgary employed through most of the season was active sticks, clogging lanes, collapsing down low - essentially they wanted to keep as many HDSC from happening, and when they did happen, they wanted to contest as many of them as possible. They were also intent on creating as many HDSC (unchallenged) events as possible with trying to have a very quick transition. Them having success never contradicted the analytical community, but somehow it sure seemed like it and they were targetted as being "unsustainable". In my argument, it wasn't unsustainable. That's exactly the rationale there - limiting/challenging as many HDSC chances as possible, while trying to create more of them.



They weren't too interested in shots from the perimeter (which annoyed me to some extent because I felt they weren't aggressive enough defensively, and the odd accidental redirection of skates and so on was always a danger).


As for deluding myself - that's quite the jump to a conclusion. I never said that they were a good team - I said that style was sustainable to winning at that rate. That in NO WAY means that they should (or even could have) beat Anaheim - a bigger, more talented team throughout the line-up. Flames didn't stand a chance against them, no matter how much I 'wanted' them to. Anaheim was a damn good team with few holes to be exploited, and Calgary couldn't match them regardless of what the metrics were. You didn't need metrics to see it. You don't have to play Anaheim (or one of the top teams in the NHL) 82 games in a season. There is also effort level and off-game that explain-away some of those wins as well against good teams.


That's one of the places where HDSC falls short - some of the metrics are too subjective, some of them are not subjective enough (and it is pretty damn difficult to differentiate them into categories).


Because the analytics community does not differentiate in HDSC, it won't account for a system or structure that is focused more on limiting these challenges but allowing more outside chances. That's what Calgary's system was all about - I bet you can go back once you figure out what boxes to differentiate the HDSC for and against and see that Calgary usually got a few more solid chances for than they got against in most games for those 1.5 seasons. It was probably measurable, but because the data was not broken down enough, it didn't appear to be.


That's where the disconnect came from with people watching the games and analysts who didn't but just looked at the metrics instead. They didn't agree, and it got ugly, but it didn't agree BECAUSE it wasn't broken down enough. That's why analysts kept saying that "Calgary was lucky" and their playstyle is unsustainable, while people watching the games kept saying: "Calgary was the more dangerous team. The other team sure did a whole lot of nothing on the outside, but they aren't going to score from there".



So am I deluding myself in thinking that Calgary was the better team than Anaheim and should have beat Anaheim that year? Absolutely I would be deluding myself, but that has ZERO to do with the fact that Calgary's winning for that 1.5 seasons was in fact fairly sustainable given what they were trying to accomplish. Changing the conversation to me somehow expecting Calgary to beat Anaheim is off-topic and 'changing the goal posts', and is just distracting from the argument.


Calgary was not the best team in the NHL. They were one of the worst teams on paper, actually, but if not for their system (which is explained away as 'lucky' and unsustainable) which consisted of just trying to limit and interfere with as many high danger chances as possible, while trying to create as many as possible for themselves without caring too much about outside shots, they would have finished in the bottom 3 that year. It wasn't a pretty system, but I argue it was sustainable. The metrics just didn't differentiate enough. Doesn't imply that they should have been President's Cup winners and Stanley Cup champs, but I bet if you further analyze their data, you can see why they experienced better than anticipated success that year, especially when you factor in all the adversity they faced with injuries throughout the year. Their 'comebacks' were 'lucky' for sure, and shouldn't be counted on, but it was also a product of their system, and probably evened out somewhat with their 'bad luck' in regards to injuries. I don't know how to explain away comebacks to be honest, but as tracking increases and more metrics are fine-tuned (or better metrics are created), comebacks can be rationalized more down the road.



You can explain it away as 'luck', but IMO, luck is just a word you use when you don't have solid enough metrics to have a high enough confidence interval. The new tech and better tracking that the NHL is implementing will result in way better metrics (and better confidence in the existing ones), and will make all of this even more interesting.


For now, the metrics that we as fans are privy to are interesting and they correlate enough to have discussions on and are worth paying attention to, but they are far from infallible and do little to explain results that don't fit into the predictive narrative and people start bickering about it as "luck" and people say "metrics are stupid". That's what should happen through the evolution of metrics anyway. Will just take some more time.


But PDO is stupid, no matter how they track it.

[CorsiHockeyLeague]

Again, you have demonstrated that you don't know what you're talking about.

Originally Posted by Calgary4LIfe

PDO IS garbage, sorry. Yeah, I understand where the math part comes from, it is just a completely stupid rationale. It is using simple math to justify something that it shouldn't. No, it doesn't always equal 1.000.

League-wide, it does, by necessity. As I said, it's just a shorthand. If you see a team at the top of the standings you weren't expecting, you look at the league PDO table, and if their PDO is high, then you can say, okay, well, their shooting percentage or save percentage (or both) is probably driving that. Then you'd have to do a lot more looking to figure out what's really going on. It's just an off-the-cuff indicator about how a team is winning (or not).

Also, I am not 'against' metrics, and you really don't have to talk down to me here. Please go and see any actuary and he will look at CORSI and will tell you that the confidence interval is WAY too low to trust. Most risk-based industries will fire you on the spot with confidence numbers that low.

Totally, but then you have to effectively discard stats as a useful tool to tell you anything about how a team is doing and just go with "we won / lost the game, that's all that matters". All these things are useful for is to give you more information about what trends are going on throughout the season, which hopefully will indicate - better than the win / loss record - what's likely to happen if X or Y tweak was made. No one is suggesting that Fenwick or XGF or RelTM can reliably predict the future; otherwise everyone would make a killing betting on hockey.

No, these metrics that you are passionately defending are not infallible

I have neither passionately defended them, nor have I suggested they're "infallible". I suspect your very long post would be shortened somewhat if it just responded to what I said, instead of creating straw men.

CORSI - what is CORSI? What I mean is what is the measurement describing? In the end, it is attempting to describe possession, correct? By using total shots + missed shots + blocked shots, at even strength 5on5, minus the opposing teams total shots + missed shots + blocked shots. In this way, you can infer possession, right? Do I have this right?

No - hilariously wrong, considering this is a basic stat that has been publicly tracked and available for well over a decade now. Corsi is shot attempts including blocked shots; you don't deduct anything. What you've described is Corsi +/-, which I don't know of anyone referencing when Corsi% is available.

It's also not exactly possession - it's offensive zone possession. Or rather, a proxy for offensive zone possession. So are shots on goal, and goals, by the way. Shot attempts just create more data points. It's just more information... You can use it to support well-reasoned conclusions or poorly reasoned ones. There's nothing inherently good or bad about data.

You seem to have strong opinions on these stats but can't actually describe them at even a basic level. I'd hate to see the gears turning as you attempt to figure out RelTM.

Well, now you have to count on the people tallying up these numbers have it right. Some arenas are notorious for poor counting stats - just shots on goal. Less of them to count, but they often get it wrong. "Nah, that was just a dump-in". So your data set is a little unreliable at it stands, but not (hopefully) terribly so.

This is really obvious, which is why you just have to take a look at the home / road splits to determine if anything particularly funny is going on, which it typically isn't.

And then you use these numbers to state that it is showing possession, but without any regard at all to 'more patient teams' that like to pass the puck around a bit more, or teams that are way more effective on the cycle and have tonnes of zone time but generate few shots on net, etc. That isn't captured, so the numbers are further skewed.

Except, first, that's not the case, based on the limited numbers of times that various people ten plus years ago just used a stopwatch to calculate offensive possession and then compared it to the corsi numbers and determined that they're basically the same. But whatever, if you want to just dream up hypothetical teams that cycle the puck around and don't try to shoot it, go ahead. The reality is that those teams tend to a) actually take more shots, and b) spend less time in their own end because they spend it in the other team's, so it winds up being a wash. But you dream up whatever scenarios you like, if that's what makes you happy.

Still, you can't argue that 'more shots' is usually more conducive to winning. Being out-shot isn't a sustainable winning strategy. So regardless of actual time of possession correlating with CORSI, out-shooting (or attempting to out-shoot) the opposition usually correlates with winning the game, right?

Typically, yes. You can get by with your goalie bailing you out and having 12% of your shots go in for a month, and for about one team every year, longer. Long enough to make the playoffs, even. Hockey's a chaotic, frantic game with lots of variables and bounces.

Except when it doesn't. Like Hartley's Calgary Flames for 1.5 seasons (and other teams that unexpectedly didn't make the playoffs, or that made it 'un-deservingly' from an analytical perspective).

How do those teams do in the post-season compared to those that play hockey the "right way" from a possession standpoint? How do they do the following year? Like I said above, are we really doing this again? If you still think there's no correlation between being a good possession team and long-term success I just cannot help you.

Yeah, I'm not bothering with the rest of it. It's just a bunch of motivated reasoning and ex post facto justification of results that didn't hold up over time, just as anyone with a brain in their head could have predicted. You don't need to read the shot attempt differentials to figure out that having the puck more than the other guys is a good way to win.

[northcrunk]

Zach Kassian:

"What's a stat. Did Tkachuk say that about me!!! I'm a stat!!! What does it mean pencil neck!???"

[Textcritic]

Originally Posted by stang

https://youtu.be/YCnHa4zFcWo



5:22-8:20 breaks this down fairly well.

I don't even need to look at an Oilers blog "breakdown" (LOL) to know that it is just as no good as your post.

Kassian is garbage.
The Oilers are garbage.
The Edmonton media is extremely smelly garbage.
Edmonton is no good.

Sent from my SM-G920W8 using Tapatalk

[powderjunkie]

Originally Posted by CorsiHockeyLeague

This one's easy - HDCA (or HDCF%).

Honest question - don't a lot of these stats relate to a 5 man unit on the ice at any given time?

[Calgary4LIfe]

Originally Posted by CorsiHockeyLeague

Again, you have demonstrated that you don't know what you're talking about.






League-wide, it does, by necessity. As I said, it's just a shorthand. If you see a team at the top of the standings you weren't expecting, you look at the league PDO table, and if their PDO is high, then you can say, okay, well, their shooting percentage or save percentage (or both) is probably driving that. Then you'd have to do a lot more looking to figure out what's really going on. It's just an off-the-cuff indicator about how a team is winning (or not).


And this is garbage. What is it telling me that SV% and SH% isn't telling me already? Gee golly, I need someone to tell me how these add up... and why in the holy hell SHOULD it add up to 100? Do you expect a good team to just be 100? And a bad team to be 100? And that any deviation of 100 is either based on good luck and bad luck, and that they will return to 0 because, well, they just absolutely should add up to 100? See the point I am making. It is a garbage stat. SH% is useful. SV% is useful. PDO is garbage. At least qualifying it by saying it is either lucky or unlucky is garbage. So yes, it is a garbage stat.


Totally, but then you have to effectively discard stats as a useful tool to tell you anything about how a team is doing and just go with "we won / lost the game, that's all that matters". All these things are useful for is to give you more information about what trends are going on throughout the season, which hopefully will indicate - better than the win / loss record - what's likely to happen if X or Y tweak was made. No one is suggesting that Fenwick or XGF or RelTM can reliably predict the future; otherwise everyone would make a killing betting on hockey.


But that isn't the argument that I am making here, so don't twist this into that. Re-read what you just replied to - I AGREE with the analytics that are out there (well, most of it), but it isn't ACCURATE or COMPLETE enough. That's it. That's all that I am saying. Though you can see trends - and that is always welcome, there are simply too many outliers given what you are referencing to draw conclusions 100% off of. I wish I remembered - ok, I wish I felt inclined enough to search for it online and post it here - but an actual actuary took at look at CORSI and laughed at it saying he would be fired for providing guidance on anything with this low of confidence intervals. CORSI is useful, it is interesting, and it can help in digesting the guts of the game, but it is hardly the conclusive statistical model that is irrefutable - it certainly isn't what you claim it to be.


I have neither passionately defended them, nor have I suggested they're "infallible". I suspect your very long post would be shortened somewhat if it just responded to what I said, instead of creating straw men.


Pot meet kettle.


No - hilariously wrong, considering this is a basic stat that has been publicly tracked and available for well over a decade now. Corsi is shot attempts including blocked shots; you don't deduct anything. What you've described is Corsi +/-, which I don't know of anyone referencing when Corsi% is available.

It's also not exactly possession - it's offensive zone possession. Or rather, a proxy for offensive zone possession. So are shots on goal, and goals, by the way. Shot attempts just create more data points. It's just more information... You can use it to support well-reasoned conclusions or poorly reasoned ones. There's nothing inherently good or bad about data.


There isn't anything inherently good or bad, but the way in which they are used and the conclusions that can be drawn can be good or bad. That's the point I am making. Agree or disagree?


You seem to have strong opinions on these stats but can't actually describe them at even a basic level. I'd hate to see the gears turning as you attempt to figure out RelTM.


Thanks. Get bent.


This is really obvious, which is why you just have to take a look at the home / road splits to determine if anything particularly funny is going on, which it typically isn't.

Except, first, that's not the case, based on the limited numbers of times that various people ten plus years ago just used a stopwatch to calculate offensive possession and then compared it to the corsi numbers and determined that they're basically the same. But whatever, if you want to just dream up hypothetical teams that cycle the puck around and don't try to shoot it, go ahead. The reality is that those teams tend to a) actually take more shots, and b) spend less time in their own end because they spend it in the other team's, so it winds up being a wash. But you dream up whatever scenarios you like, if that's what makes you happy.


"They're basically the same" is not THE SAME. Sorry, but inferencing possession by how many shots are taken is not the same thing as actual possession. Numbers get skewed. How much does it get skewed? I don't know. I anticipate that systems, how teams are built, how talented teams are, etc., can skew results for an entire season or more. Like I said, CORSI isn't garbage, but people like to take this too far. Heck, just look at Edmonton and how Eakins tried to 'play for CORSI' - Eberle and a couple of other players said as much. You really don't believe that there COULD be discrepancies between CORSI and actual possession? It is still an indirect measurement, and it is also measuring shot attempts - you sure that was a shot not a pass or slap-pass? Are they being judged completely the same from building to building? Etc. There is still a lot of 'noise' in CORSI. But I guess the gears in my head must be getting rusted and need a bit of oil or something, so I must defer to the great CORSIHOCKEYLEAGUE to provide me with all the answers so that I don't have to think for myself.


Typically, yes. You can get by with your goalie bailing you out and having 12% of your shots go in for a month, and for about one team every year, longer. Long enough to make the playoffs, even. Hockey's a chaotic, frantic game with lots of variables and bounces.


Definitely agreed.


How do those teams do in the post-season compared to those that play hockey the "right way" from a possession standpoint? How do they do the following year? Like I said above, are we really doing this again? If you still think there's no correlation between being a good possession team and long-term success I just cannot help you.


FFS, stop extrapolating it into me saying it is the preferred way. My damn argument is that a team like Calgary's at the time - poorly constructed - who have no hope of winning the 'possession game', can figure out a way to still win. Is it optimal? Never said that. Can it be sustained - yes, through sheer hard-work, I believe it can be sustained. But hockey is also not static. Teams will try and figure out how they can put a stop to it. In the end, talent SHOULD win, as long as that talent is working cohesively. Does possession-type hockey result in more wins? Of COURSE IT DOES. It is like you aren't reading what I am saying. What I am saying is that CORSI has a correlation - the correlation is there for anyone to see - but there are simply too many outliers. Whether it is from a goalie playing lights-out for a long period (is it really the goalie, or the team helping the goalie out a lot?), to a team just struggling with confidence and not finishing (slumping), to a team with poor fitness (late night video games, bad team culture that likes to party too much, etc., etc., etc.).



Yeah, I'm not bothering with the rest of it. It's just a bunch of motivated reasoning and ex post facto justification of results that didn't hold up over time, just as anyone with a brain in their head could have predicted. You don't need to read the shot attempt differentials to figure out that having the puck more than the other guys is a good way to win.

General comment #1: Stop being a dick. Just because I don't agree with the conclusions that you have when it comes to the 'advanced' metrics doesn't mean I dislike them or that I hate them, or worse, that I am some idiot.


General comment #2: Looks like teams really don't look at this stuff that you are going on about and getting so terribly upset over to the point that you are having tantrums and calling people stupid. So why are you getting so attached to this when we just heard today that it is essentially 10 seconds out of a minute's worth of hockey at a time? You want to tell me that every 10 seconds of watching hockey is enough for me to comfortably conclude anything about? Go re-read my first post that you decided to explode over - main points are about correlation not being high enough, and that the analytics (that we are privy to) are not complete.


So take your superior, smug, intolerable posting style and shove it where the sun don't shine. Learn some humility and stop talking down to people. FFS, even if I am stupid, does it make you calling me stupid more right or even more wrong? Think about that.



All this, and once again, I actually like analytics. It is posters like you that make me not want to like it.

[Flames Draft Watcher]

Yeah CHL is pretty insufferable on this topic.

[dammage79]

Things are going well in here I see. Just gonna go back now...

[CorsiHockeyLeague]

Originally Posted by Calgary4LIfe

General comment #2: Looks like teams really don't look at this stuff that you are going on about and getting so terribly upset over to the point that you are having tantrums and calling people stupid.

I did neither of those things. If anyone's throwing a tantrum, it's you - look how angry and offended you're getting. Hell, three times in that post you accuse me of calling you stupid. Please feel free to quote where I did that - I'll save you the time; you won't find it.

What I did do was hand-wave away a lot of what you're saying because it's the same tired old nonsense that's been peddled by the people who don't really understand how the stats work (and can't be bothered) for over a decade, and I don't really feel like having that debate again. It's boring, and pretty pointless, particularly at a time when every team in the league has an analytics department and Ray Shero is getting fired for (reportedly) not being able to get on board with what Tyler Dellow thinks the Devils should be doing. We're like two steps away from Eric Tulsky being a GM. This whole debate about the usefulness of these measurements has been over for years.

It's all probably a moot point anyway, because if the NHL is (as they promise) actually implementing the tracking technology they demonstrated at the All Star Game and making all of that raw data available, there's not going to be any need for most of these stats any more, as early as next season.