Confirmed Trade: [OTT/SJS] Erik Karlsson and Francis Perron for Tierney, DeMelo, Norris, Balcers, 1st, 2nd - Part III

[TheWayToRefJose]

His data just proved my point though... CF% is nowehere near as predictive as people make it our to be. It also validated my causation vs correlation hypothesis that the Hawks/Kings inflated the value of corsi in the early to mid 2010's. Now that they're ****, we're seeing a huge regression in the value of corsi.

Maybe not by itself, but coupled with the other fancy stats like PDO and HDSC% usually paints a pretty good picture of if a team is good or not. Of course there's always outliers and it's not perfect, but his data also shows that it does have a higher chance at predicting than just random chance.

Unless you have game breaking players who can make something out of nothing on a regular basis and/or a goalie who is playing out of his mind, it's usually better to not bleed high danger scoring chances and get very few of them.

 

[TomasHertlsRooster]

His data just proved my point though... CF% is nowehere near as predictive as people make it our to be. It also validated my causation vs correlation hypothesis that the Hawks/Kings inflated the value of corsi in the early to mid 2010's. Now that they're ****, we're seeing a huge regression in the value of corsi.

You could just as easily argue that the Penguins and Capitals with their generational superstars deflated the value of Corsi over the past two seasons.

For teams without Ovechkin/Holtby/Kuznetsov/Backstrom or Crosby/Malkin/Kessel/Murray/Fleury, CF% means a hell of a lot more.

 

[Samsquanch]

Maybe not by itself, but coupled with the other fancy stats like PDO and HDSC% usually paints a pretty good picture of if a team is good or not. Of course there's always outliers and it's not perfect, but his data also shows that it does have a higher chance at predicting than just random chance.

Unless you have game breaking players who can make something out of nothing on a regular basis and/or a goalie who is playing out of his mind, it's usually better to not bleed high danger scoring chances and get very few of them.

There is a peg to to hang your hat on!!! Corsi FTW!

 

[pabst blue ribbon]

Of course the better peg to hang your hat on is a player finishing 29th in points over an arbitrary 2 week period

 

[Qward]

Does Perron have NHL potential? He's good in the AHL so far this season.

No. The best you could hope for is a bottom six, league min guy because you have cap issues.

 

[Gecklund]

No. The best you could hope for is a bottom six, league min guy because you have cap issues.

He’s been fantastic with the Cuda to start the year

 

[Samsquanch]

Of course the better peg to hang your hat on is a player finishing 29th in points over an arbitrary 2 week period

Maybe, maybe not. Im under no false impression myself the Tierney is going to finish over a PPG as one of the top centers in the league. And Suomela may in fact end up being the better of the two.

But if the season was over today, and there were still people trying to argue that Suomela was the better player than Tierney that season because of corsi or whatever the hell other nonsense they were focused on, I would be out in full force backhanded slap mode (like I am now).

 

[swiftwin]

Maybe not by itself, but coupled with the other fancy stats like PDO and HDSC% usually paints a pretty good picture of if a team is good or not. Of course there's always outliers and it's not perfect, but his data also shows that it does have a higher chance at predicting than just random chance.

Unless you have game breaking players who can make something out of nothing on a regular basis and/or a goalie who is playing out of his mind, it's usually better to not bleed high danger scoring chances and get very few of them.

...right, and I just showed you that Tierney has a 58 HDCF%, which is the 3rd best forward on the Sens.

Ottawa is 28th in the league in CF% with 42.65
...but 17th in HDCF% with 49.11

Why the big difference? Because that's the system Ottawa plays. They collapse the front of the net keep shots to the outside in the defensive zone. Tierney is playing our system perfectly. I don't give two shits about CF%. Teams can take low % shots against us all day. What matters is high scoring chances. Tierney is dominating in that category.

Also, you can't just compare advanced stats of players from one team to another. They are heavily impacted by the systems teams play, and teammates. That's why I draw the comparison to +/-. Cosri is literally the +/- of shots on goal.

 

[lomiller1]

His data just proved my point though... CF% is nowehere near as predictive as people make it our to be. It also validated my causation vs correlation hypothesis that the Hawks/Kings inflated the value of corsi in the early to mid 2010's. Now that they're ****, we're seeing a huge regression in the value of corsi.

There is more than one line of evidence supporting the predictive power of Corsi.
The main recent change is that the difference between team Corsi has been shrinking. IOW there are very few legitimately good or bad CF% teams, rather almost everyone is in the mushy middle where differences in CF% are small. This suggests teams are paying closer attention to Corsi and taking action when they are struggling with it.

No they aren't big enough sample sizes. When you figure CF% you have to figure that every different player that is on the ice at the same time is a different variable.

There are mathematical techniques that can look at every single matchup/teammate and attribute responsibility for individual events accordingly. In practice this only yields slight improvement over simple Rel CF% because matchups generally won by the better players.

 

[TomasHertlsRooster]

No they aren't big enough sample sizes. When you figure CF% you have to figure that every different player that is on the ice at the same time is a different variable. You are talking equations with literally hundreds of variables. A player will never be able to play enough NHL games to solve that equation, or at least not with the simple formulas the silly stats use.

This is just wrong. Yes, there are plenty of variables, but over the course of nearly 300 games, with countless different teammates and countless different opponents, a player’s individual effectiveness will generally be the primary driving factor behind his CC%.

Teammate effects are part of the reason why CF% Rel - a player’s CF% relative to his teammates - is also a useful metric. It’s also one in which Chris Tierney has always done quite poorly over the course of his career.

For those who are saying that Tierney is better than Suomela because of points in a small sample size, let’s throw out an analogy. When Ottawa first got Matt Duchene, didn’t he score like 0 points in his first 10 games? And, wasn’t it pretty obvious that, despite the lack of actual points, Matt Duchene was a hell of a player, and better than Kyle Turris; right? That’s exactly what’s going on here with Suomela and Tierney.

 

[swiftwin]

There is more than one line of evidence supporting the predictive power of Corsi.
The main recent change is that the difference between team Corsi has been shrinking. IOW there are very few legitimately good or bad CF% teams, rather almost everyone is in the mushy middle where differences in CF% are small. This suggests teams are paying closer attention to Corsi and taking action when they are struggling with it.

So then why are the top corsi teams constantly missing the playoffs? The #1 corsi team has failed to make the playoffs two years in a row. Last year, 3 of the top 6 teams have failed to make the playoffs. That's not "a mushy middle". Again, repeat after me: correlation does not equal causation.

 

[TomasHertlsRooster]

...right, and I just showed you that Tierney has a 58 HDCF%, which is the 3rd best forward on the Sens.

Ottawa is 28th in the league in CF% with 42.65
...but 17th in HDCF% with 49.11

Why the big difference? Because that's the system Ottawa plays. They collapse the front of the net keep shots to the outside in the defensive zone. Tierney is playing our system perfectly. I don't give two ****s about CF%. Teams can take low % shots against us all day. What matters is high scoring chances. Tierney is dominating in that category.

Also, you can't just compare advanced stats of players from one team to another. They are heavily impacted by the systems teams play, and teammates. That's why I draw the comparison to +/-. Cosri is literally the +/- of shots on goal.

It’s actually the +/- of shot attempts and the fact that you don’t even know that shows that you really aren’t qualified to completely dismiss how valuable these metrics are.

 

[swiftwin]

It’s actually the +/- of shot attempts and the fact that you don’t even know that shows that you really aren’t qualified to completely dismiss how valuable these metrics are.

Sigh.... I know that. I meant shots as a whole, not specifically shots that hit the net. That fact that you're now trying to drive this argument towards semantics shows that you've lost.

 

[Masked]

He’s been fantastic with the Cuda to start the year

But how's his Corsi and PDO?

 

[lomiller1]

Maybe not by itself, but coupled with the other fancy stats like PDO and HDSC% usually paints a pretty good picture of if a team is good or not. Of course there's always outliers and it's not perfect, but his data also shows that it does have a higher chance at predicting than just random chance.

Unless you have game breaking players who can make something out of nothing on a regular basis and/or a goalie who is playing out of his mind, it's usually better to not bleed high danger scoring chances and get very few of them.

HDSC% is less predictive than score adjusted Corsi. In fact most models that integrate shot quality data underperform score adjusted Corsi. I look at shot location anyway, because very often you can rule out shot quality arguments from the get go. People tend to make claims like "Player X gives up lost of shots but they are all from the outside!" but when you look at the actual data they do just the opposite, and are more prone to giving up high danger shots.

The biggest problem with any stat is that at least 40% of what controls the outcome of an NHL game is what can best be described as luck. By that I mean randomly distributed and non-reproducible on an ongoing basis.

 

[Samsquanch]

Sigh.... I know that. I meant shots as a whole, not specifically shots that hit the net. That fact that you're now trying to drive this argument towards semantics shows that you've lost.

Its how it always goes with the advanced stats crowd... there is and will always be a never ending supply of new and useless scenarios from them to draw upon if their argument starts to goes south.

 

[Masked]

This is just wrong. Yes, there are plenty of variables, but over the course of nearly 300 games, with countless different teammates and countless different opponents, a player’s individual effectiveness will generally be the primary driving factor behind his CC%.

Teammate effects are part of the reason why CF% Rel - a player’s CF% relative to his teammates - is also a useful metric. It’s also one in which Chris Tierney has always done quite poorly over the course of his career.

It's a useless metric. Do all teammates play the same amount of time with each other? No. Everything is dependent upon everything and you don't have enough data to figure out how they all relate to each other.

For those who are saying that Tierney is better than Suomela because of points in a small sample size, let’s throw out an analogy. When Ottawa first got Matt Duchene, didn’t he score like 0 points in his first 10 games? And, wasn’t it pretty obvious that, despite the lack of actual points, Matt Duchene was a hell of a player, and better than Kyle Turris; right? That’s exactly what’s going on here with Suomela and Tierney.

Terrible analogy. Duchene had a proven track record before he was traded to Ottawa. Tierney is proving he can put up points in his current situation. Maybe it's just a lucky streak but that not a guarantee. Suomela hasn't proven anything other than that he can make an NHL roster.

 

[Masked]

Its how it always goes with the advanced stats crowd... there is and will always be a never ending supply of new and useless scenarios from them to draw upon if their argument starts to goes south.

If a player did nothing but hop on the ice during a breakout, get the puck, fire a lazy wrister at the goalie from the blueline and then go back to the bench; well, most of these stats would think he's a god that's unlucky.

 

[NyQuil]

I didn't mention anything about Corsi in regards to Soumela being better than Tierney.

From watching most of the 200 plus games Tierney played in teal and watching the few soumela has played, it's pretty obvious to the eye test. Tierney is and always has been a possession black hole. He'll get manhandled along the boards and the cycle stops and the other team goes the other way. Soumela has way, way, better hands, stronger on the puck, better chemistry with Donskoi. Their passing skills are about on par and Tierney has a slightly higher IQ, but doesn't have the tools to use it to his advantage.
I'm not saying Tierney isn't a good player, but he's not and most likely will never be the "first line Center" some fans that are too blinded by the hot start and completely discrediting the underlying numbers seem to think.

Also, if you really believe advanced stats don't matter, than you aren't critically thinking. Yes, obviously goals mean more than having a good Corsi number, but getting caved in your own zone consistently (Bad Corsi) and having an unsustainable PDO usually points to regression to the mean, which in this case will be very, very bad for Tierney. It's only a matter of time until he averages out back to his usual shooting percentage and with it being so high now, that means it will crater in order to go back to his average. His "Pins" will go do drastically while his "PINs" against will raise dramatically.

But yes, it's "intellectually dishonest" to support my claim with logic and data trends instead of using a short hot streak

You're trying to argue something completely different from what he took issue with.

We don't think Tierney is a first line centre. But he's producing like one in this short spell.

For you to say that Soumela has been comparably good over that span of time is a real head-scratcher.

As I said earlier, Tierney would be the top scorer on your team based on his hot streak.

I like how his PP production is being held against him. Don't you want your guys to produce on the PP?

HF sometimes.

For those who are saying that Tierney is better than Suomela because of points in a small sample size, let’s throw out an analogy. When Ottawa first got Matt Duchene, didn’t he score like 0 points in his first 10 games? And, wasn’t it pretty obvious that, despite the lack of actual points, Matt Duchene was a hell of a player, and better than Kyle Turris; right? That’s exactly what’s going on here with Suomela and Tierney.

Again, apparently we need this in bold:

1. Virtually no one is saying that Chris Tierney is a #1 centre.

2. No one knows if Tierney will keep up his pace, but based on his track record, it's very unlikely.

3. Soumela may end up being the better player beyond this span of games.

4. What has been taken issue with is that, when it comes to this small, minute and statistically insignificant number of games, to say that Tierney who would lead either team in scoring has put up a comparable performance to Soumela is debatable at best and requires some pretty "creative" arguments IMO.

5. Unlike in this thread, I don't remember Ottawa fans saying that Duchene was playing better than Turris over the course of his slump - just that they saw his potential.

I'm not sure why admitting that Tierney has had a solid handful of games is such a painful exercise.

In the meantime, people are actually arguing two different things.

 

[Revelation]

Maybe if the Sharks keep faltering they can replace DeBoer with Dallas Eakins and have him pump their corsi up to over 9000

 

[Samsquanch]

For those who are saying that Tierney is better than Suomela because of points in a small sample size, let’s throw out an analogy. When Ottawa first got Matt Duchene, didn’t he score like 0 points in his first 10 games? And, wasn’t it pretty obvious that, despite the lack of actual points, Matt Duchene was a hell of a player, and better than Kyle Turris; right? That’s exactly what’s going on here with Suomela and Tierney.

LOL YES!!!

You seem to get it now. Hypothetical here, but if Sens fans were trying to argue that Matt Duchene and his 3pts in 8gp was BETTER than Turris and his 9pts in 7games, they would have been dead wrong.

YES, we could look at the advanced stats to predict how this might go in the future. Thats totally fair.

But if some moron was trying to argue that in those 7 games Duchene was a better player for his team than Turris was for his own, well you see where Im going with this (I hope)....

 

[lomiller1]

So then why are the top corsi teams constantly missing the playoffs? .

What are you basing this on? Are you just cherry picking anecdotes of do you have some actual data showing that CF% doesn’t correlate to winning?

The #1 corsi team has failed to make the playoffs two years in a row. Last year, 3 of the top 6 teams have failed to make the playoffs.

cherry picking anecdotes it is...

 

[Samsquanch]

What are you basing this on? Are you just cherry picking anecdotes of do you have some actual data showing that CF% doesn’t correlate to winning?


cherry picking anecdotes it is...

Is that not the entire nature of advanced stats when you boil it right down to the basics?

Seems like it only goes one way with the stats crowd (ie when it benefits their own position).

 

[Machinehead]

The Sharks aren't just the top CF% team. They're also the top xGF% team (by quite a bit) and the top K-rating team (by quite a bit)

Their GF% is in the top 10, and really, their PDO isn't even near the bottom.

It's a statistical anomaly that the Sharks are a .500 team. Over very small samples, stupid shit happens for no reason.

 

[lomiller1]

Sigh.... I know that. I meant shots as a whole, not specifically shots that hit the net.

It’s not semantics. "Shot" already has a meaning in hockey and in NHL stats, and not all shot attempts qualify as shots.

That's why I draw the comparison to +/-. Cosri is literally the +/- of shots on goal.

Again plus minus already has a meaning and Corsi is not the same, even after accounting for your conflating shots and shot attempts it’s not the same.

Also, you can't just compare advanced stats of players from one team to another. They are heavily impacted by the systems teams play, and teammates.

Not true. Generally when people give "Corsi" numbers for a player they are using that as shorthand for "Score adjusted Relative Corsi". Rel Corsi has been shown to follow players though their careers, it’s not perfect as there are sill confounds but the main component to Rel CF% is the players own skill.