Luca Sbisa Discussion -- Part IV

[Hit the post]

Jovo was mediocre defensively ... for a #1 guy. Much like Dion Phaneuf.

If you gave a prime Jovanovski soft 3rd pairing icetime like Sbisa gets, he would dominate defensively in those minutes.

Fair enough. Phaneuf is a pretty good comparison I'd agree - formed a fairly solid pairing when Kaberle could still play.

 

[Scurr]

Yes they do. And I don't care what numbers you try throwing out to prove otherwise.

A #1 pair consistently matched against top lines of the opposition has a hell of a lot harder job than a 3rd pair playing 11 ES minutes against mainly scrubs.

If you take the time to read the article… it'll make sense to you. You can also choose not to and keep spouting something uninformed if you'd like.

 

[RobertKron]

Part of Jovanovski's defensive problem was because he was always looking to get things going offensively, and the fact that he was really good at that side of the game can't really be ignored when you're trying (for some reason) to compare his game to Sbisa.

 

[MS]

If you take the time to read the article… it'll make sense to you. You can also choose not to and keep spouting something uninformed if you'd like.

I've read the articles in question. And I don't believe everything I read.

NHL advanced statistics are an (sometimes useful) mess with a lot of things going into them. And when someone cherry picks a result that defies common sense, I don't buy it.

Same as the ridiculous 'Steroids in Baseball' article that floats around the baseball forum here claiming steroids do nothing ... based on hopelessly flawed statistical premises.

In this case, just looking superficially, the articles cite forwards, who are used in much less targeted matchup sense than defenders are. And I'm sure there's a lot more than that involved.

If you are going to try and claim that the way Edler/Tanev have been used this year provides zero difference in terms of statistical results from the way Sbisa/Bartkowski are generally used ... I'm going to call BS. Sorry. If your statistics tell you that gravity is pulling upward, there's something wrong with your data.

 

[Scurr]

I've read the articles in question. And I don't believe everything I read.

NHL advanced statistics are an (sometimes useful) mess with a lot of things going into them. And when someone cherry picks a result that defies common sense, I don't buy it.

Same as the ridiculous 'Steroids in Baseball' article that floats around the baseball forum here claiming steroids do nothing ... based on hopelessly flawed statistical premises.

In this case, just looking superficially, the articles cite forwards, who are used in much less targeted matchup sense than defenders are. And I'm sure there's a lot more than that involved.

If you are going to try and claim that the way Edler/Tanev have been used this year provides zero difference in terms of statistical results from the way Sbisa/Bartkowski are generally used ... I'm going to call BS. Sorry. If your statistics tell you that gravity is pulling upward, there's something wrong with your data.

Ah, the 'nah nah nah boo boo I can't hear you' defence.

I wouldn't say "zero difference", it's just not relevant enough that you should be using it in the thread of every player you're not fond of.

 

[MS]

Ah, the 'nah nah nah boo boo I can't hear you' defence.

I wouldn't say "zero difference", it's just not relevant enough that you should be using it in the thread of every player you're not fond of.

It's not the 'nah nah I can't hear you defense', it's the 'believing in very vague and messy statistics like they're a religion is stupid' defense.

This isn't baseball, where statistics are generally repeatable and concrete. These stats are a mess and should always be treated with a massive grain of salt.

Playing top pairing shutdown minutes vs. playing 3rd pairing minutes is not the same thing. At all.

 

[Scurr]

It's not the 'nah nah I can't hear you defense', it's the 'believing in very vague and messy statistics like they're a religion is stupid' defense.

This isn't baseball, where statistics are generally repeatable and concrete. These stats are a mess and should always be treated with a massive grain of salt.

Playing top pairing shutdown minutes vs. playing 3rd pairing minutes is not the same thing. At all.

It has a lot more to do with the cat and mouse game that coaches play than it does the statistics. Sure, most teams try to get their shutdown pair out against front line guys. But what does the other coach do? He tries to get his front line guys away from shutdown guys and against 3rd pairs. In the end, it largely comes out in the wash.

And Willie doesn't play cat… only mouse...

 

[MS]

It's not the 'nah nah I can't hear you defense', it's the 'believing in very vague and messy statistics like they're a religion is stupid' defense.

This isn't baseball, where statistics are generally repeatable and concrete. These stats are a mess and should always be treated with a massive grain of salt.

Playing top pairing shutdown minutes vs. playing 3rd pairing minutes is not the same thing. At all.

To add to this, here's how you can take something statistically significant and completely miss it through lousy compilation of stats.

We're looking at whether playing hard shutdown minutes is easier than playing soft 3rd pairing minutes.

Let's say there are 500 players in the NHL who have played over (x) minutes this year. 300 of those are forwards who are matched less specifically than defenders. Of the 200 defenders, probably half are playing generally 'mid-pairing' minutes when it averages out, either from being consistently on a 2nd pairing of by moving around. So 400+ of the 500 samples don't really have anything going on on this front. Then there are maybe 40-50 guys who consistently play shutdown minutes and 40-50 guys who consistently play soft minutes.

In order for those 80-100 guys to create a statistically significant impact on a sample size that large, the difference is going to have to be absolutely freaking enormous. And even a fairly large difference will be completely buried and become statistically insignificant when you've buried your target demographic in an unrelated sample size that large.

Not saying this is what's happening or isn't. But this is how people get fooled by trusting stupid statistics that don't make sense, and why just trusting blindly that 'someone gives a number that says this is true!' is a very bad idea.

 

[Scurr]

To add to this, here's how you can take something statistically significant and completely miss it through lousy compilation of stats.

We're looking at whether playing hard shutdown minutes is easier than playing soft 3rd pairing minutes.

Let's say there are 500 players in the NHL who have played over (x) minutes this year. 300 of those are forwards who are matched less specifically than defenders. Of the 200 defenders, probably half are playing generally 'mid-pairing' minutes when it averages out, either from being consistently on a 2nd pairing of by moving around. So 400+ of the 500 samples don't really have anything going on on this front. Then there are maybe 40-50 guys who consistently play shutdown minutes and 40-50 guys who consistently play soft minutes.

In order for those 80-100 guys to create a statistically significant impact on a sample size that large, the difference is going to have to be absolutely freaking enormous. And even a fairly large difference will be completely buried and become statistically insignificant when you've buried your target demographic in an unrelated sample size that large.

Not saying this is what's happening or isn't. But this is how people get fooled by trusting stupid statistics that don't make sense, and why just trusting blindly that 'someone gives a number that says this is true!' is a very bad idea.

It feels like you didn't read the article or didn't quite grasp what it was saying. It's not about whether hard minutes are easier than soft minutes… it's about whether anyone can truly play soft or hard minutes. The NHL game is fluid and both sides have an opportunity to affect the matchup.

 

[Michael Dal Swolle]

It feels like you didn't read the article or didn't quite grasp what it was saying.

Jeez, speaking of the "nah nah nah boo boo I can't hear you" defence...

 

[bossram]

Yes they do. And I don't care what numbers you try throwing out to prove otherwise.

A #1 pair consistently matched against top lines of the opposition has a hell of a lot harder job than a 3rd pair playing 11 ES minutes against mainly scrubs.

They do.

It's just over the course of the season, the effects tend to get washed out. The differences end up being very small.

 

[Scurr]

Jeez, speaking of the "nah nah nah boo boo I can't hear you" defence...

I hear what MS is saying and had much the same reaction when this was brought to my attention.

It has a lot more to do with the cat and mouse game that coaches play than it does the statistics. Sure, most teams try to get their shutdown pair out against front line guys. But what does the other coach do? He tries to get his front line guys away from shutdown guys and against 3rd pairs. In the end, it largely comes out in the wash.

And Willie doesn't play cat… only mouse...

This makes perfect sense, though, and matches up with the data.

 

[MS]

I hear what MS is saying and had much the same reaction when this was brought to my attention.



This makes perfect sense, though, and matches up with the data.

And again, until I see data that isolates onto defenders I don't buy it.

Most coaches get their defensive matchup most of the time. And absolutely you can shelter a 3rd pairing.

 

[Ryp37]

I'm kind of surprised scurrs arguing soft minutes don't exist, of course they do. Any coach that can run a bench has his bottom pairing playing soft minutes, they do this in bantam all the way to the NHL.

 

[Scurr]

I'm kind of surprised scurrs arguing soft minutes don't exist, of course they do. Any coach that can run a bench has his bottom pairing playing soft minutes, they do this in bantam all the way to the NHL.

Any coach that can run a bench can get his front line guys out against a 3rd pairing. It doesn't even take any effort at home or with the icing rule -it makes it a no brainer most of the time. The 3rd pairing probably accounts for 50% of the icing calls on most teams. Even Willie gets these kinds of good match ups sometimes, and he's not even trying.

 

[me2]

They do.

It's just over the course of the season, the effects tend to get washed out. The differences end up being very small.

To some degree that assumes both coaches are line matching or both coaches are 1-2-3-4. That sheltered players at home get exploited on the road. When every game is an away game.....


If we are taking Sbisa stats the only guys Sbisa has played 100 min with this year

Hutton with Sbisa 43.1
Hutton without Sbisa 48.9

Bartkowski with Sbisa 42.2
Bartkowski without Sbisa 46.4

Same story as last year. Part of it is the whole LS trying to play right, but that is on his GM for assembling a failed defense.

 

[StrictlyCommercial]

All I see is someone saying its just as easy to play the Sedins every shift as it is to play Linden Vey.

 

[The Extrapolater]

Canucks defense GA/P 60 the holy grail for defense stats:

1. Tanev 1.70
2. Hutton 1.76
3. Sbisa 1.77
4. Hamhuis 1.91
5. Edler 2.23
6. Weber 2.35
7. Biega 2.56
8. Bartkowski 2.71

So flame away...Sbisa has quietly become the 3rd best Canuck at keeping pucks out of our net and leads in physicality......but dont let numbers get in the way of an agenda.

GA/60 is effected by Sv%. Because the goaltending behind Sbisa has improved this year, his GA/60 has also improved.

 

[VanJack]

I'm kind of surprised scurrs arguing soft minutes don't exist, of course they do. Any coach that can run a bench has his bottom pairing playing soft minutes, they do this in bantam all the way to the NHL.

You've sort of hit on why some d-men like Bartkowski look so bad playing for Willie....they don't get sheltered minutes at all, because the coach just seems to open the gate and throw 'em out there....As a 6-7 d-man playing against other team's fourth lines, he might be passable....but in Willie's system he ends up exposed out there against the other team's top-six snipers.....very tough for bottom-pairing d-man to survive playing for this guy...fear that somebody like Biega is eventually going to to go the same route.

 

[Hit the post]

Any coach that can run a bench can get his front line guys out against a 3rd pairing. It doesn't even take any effort at home or with the icing rule -it makes it a no brainer most of the time. The 3rd pairing probably accounts for 50% of the icing calls on most teams. Even Willie gets these kinds of good match ups sometimes, and he's not even trying.

By accident more than design.

 

[mossey3535]

Soft minutes don't exist in the NHL.

You know those articles just highlight that Corsi differences according to ZS/QoC are predictably low right?

The authors are identifying where the chosen metric (corsi) in many ways does not reflect what we would expect the result to be from the eye test.

The main issue being that Corsi is a volume statistic in a sport where shots in general for the TEAM are not that numerous and individual player shots even less so.

For example a typical game has 100 corsi events (for and against). A defence pairing is on for 1/3 of the game so in a crude way they would expect 33 corsi events. There are four forward lines so if they faced each line equally, they would experience ~8 corsi events total per game against each one.

That is already a very low number, and the only measure we have is a differential. So of the 8 corsi events total, we would need to see an extreme swing to have anything really affect the results like 7 for to 1 against. This would have to happen all the time to have an ongoing effect on the differential. But it clearly doesn't.

Anytime we use shot volume as a fine grain analysis tool it's going to have issues. There are just not that many events, and the events are not discrete like in baseball.

Assuming the zone starts are equally biased that's 4 events per zone every game versus any given forward line for a typical d-pairing. So you're going to have similar issues here with regards to the effect of zone starts on corsi.

The first article actually references a contradictory article where Ferrari found that QoC is a driving force so the author fully realizes that the work is ongoing. He also shows that as QoC is currently formulated that there is a clear increase in corsi when going against worse players. His only real argument is that effect seems to go away when you take into account the overall ice time.

Whether or not that disappearance is real or just an effect of the limitations of the metric is not a conclusion of that first article. It's a great article and properly written because he includes the conclusions of other workers, he recognizes that his result is not the same as theirs, and he welcomes future work to decide who is right. He also calls into question whether the metric involved can really tell us what is happening.

 

[Scurr]

CONCLUSION

While competition certainly does play a big factor in determining how a player will do in any given shift, with these competition metrics we see nobody with usage extreme enough to require a major correction factor. Using the curve for the average player on the first chart, we can calculate that Nodl’s 95th percentile usage is only harsh enough to bump an average player’s Corsi down to 49.5%, while Betts’ 13th percentile usage would be soft enough to allow an average player to post a 50.6% Corsi.

Again, it has less to do with the numbers and more to do with the way player ice-time is manipulated by both teams. A coach can choose to put Manny Malhotra out in the defensive end an extreme amount but they don't get carte blanche on who that player gets to play against. There is a coach on the other bench that wants to get his first line out against your 3rd pairing just as much as you want to keep it away. It's not that QOC doesn't have much of an affect... it's that over time they don't really play a much different QOC at all.

There is an article by someone at broadstreethockey that draws a more direct line to what I'm talking about but I haven't been able to find it.

 

[Ryp37]

Again, it has less to do with the numbers and more to do with the way player ice-time is manipulated by both teams. A coach can choose to put Manny Malhotra out in the defensive end an extreme amount but they don't get carte blanche on who that player gets to play against. There is a coach on the other bench that wants to get his first line out against your 3rd pairing just as much as you want to keep it away.

There is an article by someone at broadstreethockey that draws a more direct line to what I'm talking about but I haven't been able to find it.

Yeah I'm still not buying it, it's called changing on the fly. Pairings know who they should and shouldn't be out against and will change at the first chance possible.

 

[Scurr]

Yeah I'm still not buying it, it's called changing on the fly. Pairings know who they should and shouldn't be out against and will change at the first chance possible.

I think you're wrong but I'll admit that after reading that article again I definitely oversold my case Still, how easy is it for third pairings to move the puck out and get a change after an icing… against a top line?

 

[mossey3535]

Again, it has less to do with the numbers and more to do with the way player ice-time is manipulated by both teams. A coach can choose to put Manny Malhotra out in the defensive end an extreme amount but they don't get carte blanche on who that player gets to play against. There is a coach on the other bench that wants to get his first line out against your 3rd pairing just as much as you want to keep it away. It's not that QOC doesn't have much of an affect... it's that over time they don't really play a much different QOC at all.

There is an article by someone at broadstreethockey that draws a more direct line to what I'm talking about but I haven't been able to find it.

You're taking his analysis as gospel when he himself does not.

Some might remember that Ferrari found the exact opposite and concluded that quality of competition is a major driver of a player’s results. These results seem contradictory at first glance, so I have held onto this article for a while trying to figure out how to account for the discrepancy.

Again he allows for contradictory work to his own.

So over the course of a season, differences in competition should even out more and account for less than 50% of the difference between players, which makes me suspicious that random chance played a role in Ferrari's 55% finding. I am looking forward to the work Jared Lunsford is doing to replicate Ferrari’s approach in part because I hope to better understand the numbers Ferrari found.

So he comes up with a hypothesis which is it itself flawed - that differences in competition should even out. Why? This hypothesis is just a restatement of the conclusion he is trying to reach. He then does some serious post-hoc to try to make those two come together.

For example, if a player is 4th line shouldn't their QoC be uniformly poor? If he plays with other 4th line players, shouldn't that normalize whatever results they get? If he is correct about WOWY numbers being more consistent then he should have found correlating data to back up his claim. But he doesn't.

Essentially his base argument is that Ferrari screwed up and that all the corrections he used to come to his 55% were pointless in the first place. But he has almost no data to back that up besides invoking a symmetry argument that might be false to begin with.

But that essentially means that his chosen metric is insufficient to show fine grain changes in these factors.

This is a common problem with these advanced stats guys. Their statistical analyses are frankly very basic. They almost never stop to think about whether Corsi or shot differential is a good predictor to begin with. And remember, I think this is an example of a pretty good article.

I believe part of the explanation arises from the selection of the metrics. In this analysis, I have used a player’s shot differential as the measure of how good he is (in the competition metric) and of his results. However, as we have already discussed, simple shot differential requires various correction factors to account for the impact of usage. Ferrari’s study included corrections for zone starts and quality of teammates, which improves the assessment of both the player’s performance and his competition. It may simply be the case that this more sophisticated metric more precisely identifies the strength of competition and therefore finds more differences between players in competition faced.

Oh yeah, and then at the end he admits he might have been wrong about everything.

Finally, the 95% CI he invokes at the end is missing a major factor. If you don't have enough resolution to your data to discern small changes that might be large relative to the rest of the data (i.e. differences in QoC) then your distribution is incapable of telling you anything.