[Trade] Elias Lindholm + Noah Hanifin to Flames for D. Hamilton, Micheal Ferland, & Adam Fox

[redforever]

Originally Posted by Hemi-Cuda

I was never a fan of Hamilton, I felt his points helped to cover up his poor defensive play and he always had those rumors of being difficult in the room swirling around. It only took a few games to see that Hanifin is a far better defensive player, and now that he's actually putting some points up this trade is a no brainer of a win


EDIT: Nice to see that I'm consistent, https://forum.calgarypuck.com/showpo...postcount=1015

It was not even about losing Hamilton's points if we traded him. He simply never seemed to live up to his potential and he could be a pylon at times, looking totally confused.

I think it is a personality thing. Some defense men have all the tools to control games but don't seem to want to use them. Hamilton just seemed to want to be "too nice" on the ice.

I was extremely glad that Hamilton was traded rather than Brodie.

[Enoch Root]

Originally Posted by Bingo

Kind of hard to come up with a situation where being top 15 in limiting the opposition from a shot attempt isn't a good thing.

True. Maybe.

Problem is that there are team-mates involved, along with 5 competitors - there is a LOT of noise in the numbers.

And taking that to the next step, I have asked many times, but have never seen anyone come back with a statistical reference as to how reliable these numbers are - what is the confidence test? (i.e t-stat, or whatever metric applies to the data).

Again, we have a lot of noise due to team-mates, competition etc. And we have numbers where people claim that someone with shot suppression of 45% is way better than someone with shot suppression of 50%. Is it? Unless we have confidence levels for our data, we don't know. 5% difference way well be within one stdev, and thus simply random noise - we don't know.

Yet the people that present the numbers typically present them as though they are awesome pieces of information that we must heed, if we want to be enlightened hockey fans (NOT saying you do that, Bingo).

[Roof-Daddy]

I know some players, especially swedish players, can take a little longer to hit their full stride, but is this a level that Lindholm can sustain?

Treliving hit the jackpot on this trade if he can. Having Lindholm become a 30 goal 80 point two way forward who can fill in at center and take draws really takes the sting out of Bennett's slow offensive progression so far.

Jolted the rebuild significantly and helped open the window.

[Enoch Root]

Originally Posted by Roof-Daddy

I know some players, especially swedish players, can take a little longer to hit their full stride, but is this a level that Lindholm can sustain?

Treliving hit the jackpot on this trade if he can. Having Lindholm become a 30 goal 80 point two way forward who can fill in at center and take draws really takes the sting out of Bennett's slow offensive progression so far.

Jolted the rebuild significantly and helped open the window.

I don't know what his shooting percentage is (haven't looked and don't care), but I don't see an average player on a hot streak, I see a really good hockey player.

I wouldn't expect him to keep up his 80+ point pace, but he does so many things really, really well that there is no reason to think he won't continue to be a fantastic hockey player. And a top line player.

[Bingo]

Originally Posted by Enoch Root

True. Maybe.

Problem is that there are team-mates involved, along with 5 competitors - there is a LOT of noise in the numbers.

And taking that to the next step, I have asked many times, but have never seen anyone come back with a statistical reference as to how reliable these numbers are - what is the confidence test? (i.e t-stat, or whatever metric applies to the data).

Again, we have a lot of noise due to team-mates, competition etc. And we have numbers where people claim that someone with shot suppression of 45% is way better than someone with shot suppression of 50%. Is it? Unless we have confidence levels for our data, we don't know. 5% difference way well be within one stdev, and thus simply random noise - we don't know.

Yet the people that present the numbers typically present them as though they are awesome pieces of information that we must heed, if we want to be enlightened hockey fans (NOT saying you do that, Bingo).

With Hamilton I broke the events up and got away from percentages because he puts everything on net and greatly affects his own corsi.

So what I was looking at was shot attempts against per 60 minutes of ice time ... which boils down to what he and his partner allow when on the ice five on five.

Those are counted by an independent source that does the same for all teams (players) so it's at least objective, or flawed equally for everyone.

[Enoch Root]

Originally Posted by Bingo

With Hamilton I broke the events up and got away from percentages because he puts everything on net and greatly affects his own corsi.

So what I was looking at was shot attempts against per 60 minutes of ice time ... which boils down to what he and his partner allow when on the ice five on five.

Those are counted by an independent source that does the same for all teams (players) so it's at least objective, or flawed equally for everyone.

No doubt they are objective. But that wasn't the point at all.

My point was with respect to statistics, and how we determine their validity. I am pretty sure that you, as a trader, are very familiar with statistics (as am I).

It is not enough to simply look at two numbers and say: this guy's shot attempts per 60 are better than that guys, therefore he is doing something better.

With stats, we have to show the robustness or validity of the number, relative to simple chance (whether that be a t-stat, z score, or whatever applies).

Until we know that, we can not draw any conclusions whatsoever from the stated number. If one player's score is 50% and the other player's score is 55%, but the stdev of those numbers is 8%, then the stat would fail the null hypothesis test, and we could not determine at all if it was anything other than random noise.

That is my point, and I have never seen anyone post any support or reference for the numbers in question. It takes a lot of data to get past the random noise test - especially, like with hockey, where there are so many other variables at play at the same time.

[Bingo]

Originally Posted by Enoch Root

No doubt they are objective. But that wasn't the point at all.

My point was with respect to statistics, and how we determine their validity. I am pretty sure that you, as a trader, are very familiar with statistics (as am I).

It is not enough to simply look at two numbers and say: this guy's shot attempts per 60 are better than that guys, therefore he is doing something better.

With stats, we have to show the robustness or validity of the number, relative to simple chance (whether that be a t-stat, z score, or whatever applies).

Until we know that, we can not draw any conclusions whatsoever from the stated number. If one player's score is 50% and the other player's score is 55%, but the stdev of those numbers is 8%, then the stat would fail the null hypothesis test, and we could not determine at all if it was anything other than random noise.

That is my point, and I have never seen anyone post any support or reference for the numbers in question. It takes a lot of data to get past the random noise test - especially, like with hockey, where there are so many other variables at play at the same time.

I'd agree to a point.

Nothing can be locked down completely for sure. But there are certain things that just logically when counted would suggest player A is say better defensively than player B.

I don't think they've nailed it when it comes to how to separate value in a scoring chance, but I'm pretty prepared to say I'd rather have a defenseman that allows less shots of any kind from the slot as a rule for example.

[Enoch Root]

If we both flip coins 100 times, and you flip 55 heads and I only flip 48 heads, are you better at flipping heads than I am?

If we don't have measurements to determine or assess the validity of our results, stats are meaningless.

Now, if one defenseman is giving up egregious scoring chances seemingly every shift, and another isn't, we don't need stats to verify what we already know. But when we are talking 55% vs 50%, with other variables at play at the same time, we sure as hell do.

[Bingo]

Originally Posted by Enoch Root

If we both flip coins 100 times, and you flip 55 heads and I only flip 48 heads, are you better at flipping heads than I am?

If we don't have measurements to determine or assess the validity of our results, stats are meaningless.

Now, if one defenseman is giving up egregious scoring chances seemingly every shift, and another isn't, we don't need stats to verify what we already know. But when we are talking 55% vs 50%, with other variables at play at the same time, we sure as hell do.

And I get that.

I guess it comes down to what the gap is, how repeatable it is, and whether you see that as significant.

When it comes to CA/60 this year the top ten average 47.5, the next ten 50.0 so that's not a huge gap and I think it fits into your coin toss example.

But Cody Ceci is at 75 and Mark Giordano is at 48 ... that's a big gap. Last year Ceci was at 68 and Giordano was at 52. The gap is wider this year, but the consistency between the two ois that less bad things happen when Giordano is on the ice when compared to Ceci.

But as you've said ... there are other factors. The Senators suck the Flames don't ... though both teams missed the playoffs last year.

However I'm pretty comfortable in saying Giordano would have better numbers playing for the Senators that Ceci does now.

[Enoch Root]

Originally Posted by Bingo

And I get that.

I guess it comes down to what the gap is, how repeatable it is, and whether you see that as significant.

When it comes to CA/60 this year the top ten average 47.5, the next ten 50.0 so that's not a huge gap and I think it fits into your coin toss example.

But Cody Ceci is at 75 and Mark Giordano is at 48 ... that's a big gap. Last year Ceci was at 68 and Giordano was at 52. The gap is wider this year, but the consistency between the two ois that less bad things happen when Giordano is on the ice when compared to Ceci.

But as you've said ... there are other factors. The Senators suck the Flames don't ... though both teams missed the playoffs last year.

However I'm pretty comfortable in saying Giordano would have better numbers playing for the Senators that Ceci does now.

That is pretty much what I am saying, yes.

If there is a big enough discrepancy, like your example, it is clear that Giordano is better at shot suppression than Ceci. And I would argue that we would all be able to see that by watching. And we can all confidently quote the stats, because they would pass any test for significance.

My point is with respect to the vast majority of quoted stats for hockey. People will look at two players' high danger corsi-for, where, for example, one player is at 55% and another player is a 51%, and conclude from that that player A is better at generating said chances. What I am saying is: no, we cannot conclude that, unless we know that the difference is statistically significant.

And no one ever provides that for these data (probably because it would be damning, and most of their work would go out the window)

[FireGilbert]

At the time I was happy with the trade and thought the improvements to the forwards would more than make up for the hit to the defence. Now I am ecstatic! Lindholm is a star and the defence is even better with the top 2 pairings working perfectly. When you throw in the contracts this is an amazing deal. Well done Treliving! Thinking about it he really took advantage of a budget Canes team.

[Jay Random]

Originally Posted by Enoch Root

My point is with respect to the vast majority of quoted stats for hockey. People will look at two players' high danger corsi-for, where, for example, one player is at 55% and another player is a 51%, and conclude from that that player A is better at generating said chances. What I am saying is: no, we cannot conclude that, unless we know that the difference is statistically significant.

And no one ever provides that for these data (probably because it would be damning, and most of their work would go out the window)

Statistical significance is a concept that applies to samples drawn from populations, and the null hypothesis is that the population did not change when different samples were taken. That concept does not even apply in a case such as this, where the data are not sampled, but collated for the population as a whole.

No one provides what you're asking for because it isn't a possible question. By definition, a sample of 100% of the population is always statistically significant.

[Enoch Root]

Originally Posted by Jay Random

Statistical significance is a concept that applies to samples drawn from populations, and the null hypothesis is that the population did not change when different samples were taken. That concept does not even apply in a case such as this, where the data are not sampled, but collated for the population as a whole.

No one provides what you're asking for because it isn't a possible question. By definition, a sample of 100% of the population is always statistically significant.

Except that people use part of the season (for example, the first 29 games) to represent the season, and as such it is a sample.

Also, if each data set is in fact its own population, then extracting meaningful information from it, with respect to providing insight into the future, is also, by definition, not valid. Either it is a subset of the universe of data, and thus valid for analysis, or it is its own data set, and thus of little value.

This also doesn't solve or address the problem of non-isolated variables, and the amount of noise in the data.

These criticisms are not addressed at you, but at the data.

[Jay Random]

Originally Posted by Enoch Root

Except that people use part of the season (for example, the first 29 games) to represent the season, and as such it is a sample.

If people do that, that's on them and not on the data. The question is how large a population of data you need for it to have predictive value; which is not the same as the question of statistical significance, though obviously some of the same problems arise in dealing with it.

Also, if each data set is in fact its own population, then extracting meaningful information from it, with respect to providing insight into the future, is also, by definition, not valid. Either it is a subset of the universe of data, and thus valid for analysis, or it is its own data set, and thus of little value.

It is the set of all data that exist at the present time for the conditions being observed. The possibility of bias due to an insufficiently random sample does not figure into it. In any case, we were talking about data concerning Hamilton playing as a Calgary Flame, so the statistical universe for that situation is closed and complete. There is not, and probably never will be, an unfinished season in which Hamilton plays for the Flames.

This also doesn't solve or address the problem of non-isolated variables, and the amount of noise in the data.

Yes, now we're down to the meat and potatoes. Hockey statistics are notoriously non-orthogonal, and there is hardly such a thing as an isolated variable in the whole shooting match.

In the case of Hamilton, the best analytic tool I know of offhand is WOWY data, which at least removes the noise introduced by his playing with different teammates. It's been a while since I looked at those data, but as best I can recall, there was no smoking gun: Hamilton did not perform strikingly better with one particular teammate than without him, which pretty much rules out the hypothesis that he was being carried by (for instance) Giordano.

In fact, the basic counting data seem to suggest that Hamilton was not particularly bad defensively. (That is, not noticeably worse than the team as a whole; which is not much of an endorsement.) If the eye test suggests that he was, that may be down to the problem of ungraceful failure. To paraphrase what one of Winston Churchill's friends said about him: ‘Hamilton was usually right, but when he was wrong, my God.’

[Enoch Root]

Originally Posted by Jay Random

If people do that, that's on them and not on the data. The question is how large a population of data you need for it to have predictive value; which is not the same as the question of statistical significance, though obviously some of the same problems arise in dealing with it.

It is the set of all data that exist at the present time for the conditions being observed. The possibility of bias due to an insufficiently random sample does not figure into it. In any case, we were talking about data concerning Hamilton playing as a Calgary Flame, so the statistical universe for that situation is closed and complete. There is not, and probably never will be, an unfinished season in which Hamilton plays for the Flames.

Yes, now we're down to the meat and potatoes. Hockey statistics are notoriously non-orthogonal, and there is hardly such a thing as an isolated variable in the whole shooting match.

In the case of Hamilton, the best analytic tool I know of offhand is WOWY data, which at least removes the noise introduced by his playing with different teammates. It's been a while since I looked at those data, but as best I can recall, there was no smoking gun: Hamilton did not perform strikingly better with one particular teammate than without him, which pretty much rules out the hypothesis that he was being carried by (for instance) Giordano.

In fact, the basic counting data seem to suggest that Hamilton was not particularly bad defensively. (That is, not noticeably worse than the team as a whole; which is not much of an endorsement.) If the eye test suggests that he was, that may be down to the problem of ungraceful failure. To paraphrase what one of Winston Churchill's friends said about him: ‘Hamilton was usually right, but when he was wrong, my God.’

To the first bold, agreed.

To the second... the reason people track the data is - presumably - to try to determine what the player is - are they better or worse at this or that detail of the game, compared to other players?

And the implication of that question is that there is consistency throughout a player's career, as to what skill set they have. Obviously, players evolve. But with respect to looking at the numbers, the assumption is that, at least to a large degree, there is a consistency that can be evaluated.

In the case of Hamilton, yes, his time in Calgary is now a closed population. But the assumption, built into the data analysis, is that the data should continue to apply to him as a player.

And as such, the users of the data are assuming a larger data population - rightly or wrongly (mostly wrongly).

And what most of this leads to, is that, no matter how you slice it, most of the analysis is basically useless (other than to say: for a given finite period, under these conditions, this happened). However, rarely, if ever, do the presenters of the data leave it to such a confined conclusion, with no ability to make implications beyond the said population.

[Bingo]

The biggest problem with Hamilton's career and isolating him vs his teammates is the fact that he's almost always played with good defenseman.

2400 minutes with Giordano
700 minutes with Chara
400 minutes with Slavin

He did have some pretty ugly numbers with Russell (but that isn't isolated to that pair because Russell always has a low CF%) and also with Jokipakka and Engelland.

[Calgary4LIfe]

I can't wait until the NHL starts tracking players on the ice with the new tech. I think it is going to revolutionize counting stats and provide a heck of a lot more insight into players.


Right now, I dislike these stats because:
1) There is simply too much 'noise' around them
2) Annoying people try and use them incorrectly to support their biases, rather than use them to remove biases
3) They don't discern nearly enough by the quality of each chance



I love these stats because:
1) They are, for the most part, simply counting stats and are thus more or less irrefutable
2) They do help to remove biases
3) They do offer further insight into players and the game overall


When all that fancy tracking tech gets implemented, I wonder how much more accurate they will be? I wonder what brand new statistical concepts will materialize? I also think it will further impact the game in terms of coaching strategies and systems. Like it or not, there are some teams that have admitted to playing for CORSI - Edmonton for sure under Eakins according to Eberle and (IIRC) Hall.



What is also funny is that Gulutzan's system aligned with (most) of the advanced metrics that the analytics community at large deem noteworthy. Peters' system also seems aligned with it. Hartley's system was misaligned terribly with what the metrics were stating until his last year. I really do feel that between these 3 coaching strategies/systems, the currently popularized set of numbers has missed something noteworthy. Hartley had a fairly crap roster but still found SUSTAINED success lasting 1.5 seasons, right up until the goalies pooped in the crease. When Ramo was again average-above average, the Flames experienced an uptick in success that lasted until Ramo was injured. I really believe that Hartley's system wasn't all 'luck'. Ramo wasn't stealing every game.



I am actually really excited for these metrics to further evolve. I think it is really going to impact the game as how it is currently played over time.

[shadowlord]

Probably the wrong thread for this...

Ferland on concussions: 'I'm very concerned'
https://www.tsn.ca/ferland-on-concus...rned-1.1224477

[Jacks]

Of course the video that starts playing on that page is about the bloody Leafs. God TSN is insufferable lately.

[dammage79]

Oh man. Ferlands first game back from concussion and right back off the ice with...an upper body injury. Likely another head injury. This may be it for the man.

Chris Johnston
‏
Verified account

@reporterchris
55s56 seconds ago
More
Micheal Ferland is done for the night with an upper-body injury. He missed Carolina's previous four games with a concussion.

I was never the biggest fan of Ferland but from a very human perspective this is so sad. Feel for the guy.