Who's better, Levi or UPL?

[LGR4GM]

52 minutes ago, dudacek said:

The thread concept is silly and you were absolutely correct that we put the cart before the horse last summer.

What do you think of Levi’s ability right now?

Comparing the two players sharing the crease next year is silly? 

18 hours ago, #freejame said:

I don’t know if you meant this to be funny but I laughed 

I did

[dudacek]

15 minutes ago, LGR4GM said:

Comparing the two players sharing the crease next year is silly?

Quite the opposite. Sorry, I did not mean to offend.

It's any argument that Levi is better than UPL right now that I find silly.

And this is coming from the man who may be this forum's biggest Levi booster.

[LGR4GM]

11 minutes ago, dudacek said:

Quite the opposite. Sorry, I did not mean to offend.

It's any argument that Levi is better than UPL right now that I find silly.

And this is coming from the man who may be this forum's biggest Levi booster.

Not offended. Should have added a tongue out emoji 😝

[Pimlach]

I wont be critical the fancy stats, they are another tool, but I don't fully understand how they get the data for them.  Who decides "expected" goals for example?  

Lets say UPL faces 35 shots and lets in 2 goals playing the entire 60 minutes.   For the game the GAA = 2.00, and the SV % = 33/35*100 = 94.3%.  That is easy enough. 

Lets say he made 3 amazing saves in that game that someone thinks should have been goals.   Who decides that for determination of "goals saved above expected"?  Because they are saying he gave up 2 but should have given up 5 and in some way they are looking for a way to account for that.   Right?  

OR, lets say he let in an easy goal that deflected off his glove on a routine catch?  How do they determine and account for that?  

I understand basics of statistics and probabilities,  so when I see the data I can usually think through the meaning, I just don't know who comes up with the data for these stats and how.   Is there a set of criteria that is applied to all situations in a uniform manner.    Does the league do this?  or the teams?  or independent services?  

Expected Goals - I pass the puck to Kucherov in the slot with the goalie pinned deep in the crease.  There is tons of net.  He shoots cleanly with no defender on him.  Is that an expected goal?   Is it still an expected goal if that player is Eric Robinson?   I assume it is.   If Kuchrov dribbles it into the goalies pads is that a goal below expected for him?  Is it a save above expected for the goalie?  

What are the best fancy stats to tell us effectiveness of a player by position?   

I am probably not alone in wondering about some of this.  

[dudacek]

11 minutes ago, Pimlach said:

I wont be critical the fancy stats, they are another tool, but I don't fully understand how they get the data for them.  Who decides "expected" goals for example?  

Lets say UPL faces 35 shots and lets in 2 goals playing the entire 60 minutes.   For the game the GAA = 2.00, and the SV % = 33/35*100 = 94.3%.  That is easy enough. 

Lets say he made 3 amazing saves in that game that someone thinks should have been goals.   Who decides that for determination of "goals saved above expected"?  Because they are saying he gave up 2 but should have given up 5 and in some way they are looking for a way to account for that.   Right?  

OR, lets say he let in an easy goal that deflected off his glove on a routine catch?  How do they determine and account for that?  

I understand basics of statistics and probabilities,  so when I see the data I can usually think through the meaning, I just don't know who comes up with the data for these stats and how.   Is there a set of criteria that is applied to all situations in a uniform manner.    Does the league do this?  or the teams?  or independent services?  

Expected Goals - I pass the puck to Kucherov in the slot with the goalie pinned deep in the crease.  There is tons of net.  He shoots cleanly with no defender on him.  Is that an expected goal?   Is it still an expected goal if that player is Eric Robinson?   I assume it is.   If Kuchrov dribbles it into the goalies pads is that a goal below expected for him?  Is it a save above expected for the goalie?  

What are the best fancy stats to tell us effectiveness of a player by position?   

I am probably not alone in wondering about some of this.  

I’m kinda like you, in that I don’t know for sure, but I always thought expected goals was similar to the high-danger concept and was largely based on where the chance was coming from.

I appreciate the effort put into fancystats and respect their worth.

My peeve would be the devaluation of actual goals for and against that accompanies some of their use.

Because those are the stats that matter most.

[inkman]

Regardless, unless Upie takes a big step back and Levi in undeniable, I’d like to see Devon spend next season in Rochester.  Plenty of dudes out there they can get to play 20 games if need be. 

[Mango]

Levi's sample size is half of UPL's. 

UPL has been the better overall goaltender by some margin this year. 

[Mango]

On 4/19/2024 at 1:58 PM, LGR4GM said:

Notice how I am not at all in the Sabres Sales mongering rumor thread, that's because I think the topic isn't worth discussing. You should follow my lead and stay out of threads you don't care about. 

Nothing says "I don't care" like bringing up a conversation you aren't having just to tell everybody it is a topic you don't care about. 

[Pimlach]

2 hours ago, dudacek said:

I’m kinda like you, in that I don’t know for sure, but I always thought expected goals was similar to the high-danger concept and was largely based on where the chance was coming from.

I appreciate the effort put into fancystats and respect their worth.

My peeve would be the devaluation of actual goals for and against that accompanies some of their use.

Because those are the stats that matter most.

Indeed.  I expected that you would know answers to these questions.  
 

Edited April 20 by Pimlach

[Pimlach]

On 4/19/2024 at 1:44 PM, Buffalonill said:

Who honestly knows goaltenders are so weird

 

I think I know it.  Can you hum a few bars? 

Edited April 20 by Pimlach

[GoPuckYourself]

8 hours ago, Pimlach said:

I wont be critical the fancy stats, they are another tool, but I don't fully understand how they get the data for them.  Who decides "expected" goals for example?  

Lets say UPL faces 35 shots and lets in 2 goals playing the entire 60 minutes.   For the game the GAA = 2.00, and the SV % = 33/35*100 = 94.3%.  That is easy enough. 

Lets say he made 3 amazing saves in that game that someone thinks should have been goals.   Who decides that for determination of "goals saved above expected"?  Because they are saying he gave up 2 but should have given up 5 and in some way they are looking for a way to account for that.   Right?  

OR, lets say he let in an easy goal that deflected off his glove on a routine catch?  How do they determine and account for that?  

I understand basics of statistics and probabilities,  so when I see the data I can usually think through the meaning, I just don't know who comes up with the data for these stats and how.   Is there a set of criteria that is applied to all situations in a uniform manner.    Does the league do this?  or the teams?  or independent services?  

Expected Goals - I pass the puck to Kucherov in the slot with the goalie pinned deep in the crease.  There is tons of net.  He shoots cleanly with no defender on him.  Is that an expected goal?   Is it still an expected goal if that player is Eric Robinson?   I assume it is.   If Kuchrov dribbles it into the goalies pads is that a goal below expected for him?  Is it a save above expected for the goalie?  

What are the best fancy stats to tell us effectiveness of a player by position?   

I am probably not alone in wondering about some of this.  

It's a bunch of nerds getting together to see who's pocket protector is bigger. I'm sure if you deep dive into this there is some fatal flaw like anything. 

[LGR4GM]

8 hours ago, Pimlach said:

I wont be critical the fancy stats, they are another tool, but I don't fully understand how they get the data for them.  Who decides "expected" goals for example?  

Lets say UPL faces 35 shots and lets in 2 goals playing the entire 60 minutes.   For the game the GAA = 2.00, and the SV % = 33/35*100 = 94.3%.  That is easy enough. 

Lets say he made 3 amazing saves in that game that someone thinks should have been goals.   Who decides that for determination of "goals saved above expected"?  Because they are saying he gave up 2 but should have given up 5 and in some way they are looking for a way to account for that.   Right?  

OR, lets say he let in an easy goal that deflected off his glove on a routine catch?  How do they determine and account for that?  

I understand basics of statistics and probabilities,  so when I see the data I can usually think through the meaning, I just don't know who comes up with the data for these stats and how.   Is there a set of criteria that is applied to all situations in a uniform manner.    Does the league do this?  or the teams?  or independent services?  

Expected Goals - I pass the puck to Kucherov in the slot with the goalie pinned deep in the crease.  There is tons of net.  He shoots cleanly with no defender on him.  Is that an expected goal?   Is it still an expected goal if that player is Eric Robinson?   I assume it is.   If Kuchrov dribbles it into the goalies pads is that a goal below expected for him?  Is it a save above expected for the goalie?  

What are the best fancy stats to tell us effectiveness of a player by position?   

I am probably not alone in wondering about some of this.  

They look it shot data from around the league. So if 100 players take a wrist shot from spot X on the ice and 5 go in, then wrist shots on that spot have a .05 chance of being a goal or a 0.05 expected goal. So you add all the shots in a game up.  The more advanced models can take into account things like puck movement before the shot and player in front of the net but you still get a percentage. 

[Pimlach]

15 minutes ago, LGR4GM said:

They look it shot data from around the league. So if 100 players take a wrist shot from spot X on the ice and 5 go in, then wrist shots on that spot have a .05 chance of being a goal or a 0.05 expected goal. So you add all the shots in a game up.  The more advanced models can take into account things like puck movement before the shot and player in front of the net but you still get a percentage. 

Awesome. They build a data base and assign a weighting to shot position, adding in the other factors like screens is really cool.  Over time and with enough data, this should give reasonable results.  

 

 

 

 

 

 

 

[LGR4GM]

9 hours ago, Pimlach said:

Awesome. They build a data base and assign a weighting to shot position, adding in the other factors like screens is really cool.  Over time and with enough data, this should give reasonable results.  

 

The catch is I don't think they account for shooting talent. Meaning a Krebs shot is worth a Tage shot in xgf because it's based on league average. Maybe team models do? That's why you need a data engineer in your analytics team. 

[LGR4GM]

I suppose you could use a players sh% as some kind of multi plier or you could compare that player to league average from that spot and adjust. So if Tage has 100 shots from the spot and 7 go in but the league only has 5, you could average the two and give it a .06xgf. Hmm, I'd have to think about it and would need a statistician to really do that work. 

Yes goals are the end all be all but goals are rare. Cozens for example just had a strange season he's gonna bounce back from but his xgf helps there. It's also why last offseason everyone public model predicted a Sabres scoring regression. Sure, shooting talent matters and you can consistently be above your xgf but if you're multiple standard deviations above, that doesn't work. 

The other thing if I were a team, I'd classify each player into a cohort 1-4 based on their shooting talent. You could then adjust the xgf by player cohort. If you're a top 25% shooter in the league the avg sh% is .08 but if you're a bottom 25% it's .04 and that would really help adjust the xgf. I'm unsure of how granular the private nhl tracking data is. Once you have 3 years of data though, I'd bet the numbers could be stabilized with a confidence interval because they should fall on a bell curve. 

[Believer]

1 hour ago, LGR4GM said:

I suppose you could use a players sh% as some kind of multi plier or you could compare that player to league average from that spot and adjust. So if Tage has 100 shots from the spot and 7 go in but the league only has 5, you could average the two and give it a .06xgf. Hmm, I'd have to think about it and would need a statistician to really do that work. 

Yes goals are the end all be all but goals are rare. Cozens for example just had a strange season he's gonna bounce back from but his xgf helps there. It's also why last offseason everyone public model predicted a Sabres scoring regression. Sure, shooting talent matters and you can consistently be above your xgf but if you're multiple standard deviations above, that doesn't work. 

The other thing if I were a team, I'd classify each player into a cohort 1-4 based on their shooting talent. You could then adjust the xgf by player cohort. If you're a top 25% shooter in the league the avg sh% is .08 but if you're a bottom 25% it's .04 and that would really help adjust the xgf. I'm unsure of how granular the private nhl tracking data is. Once you have 3 years of data though, I'd bet the numbers could be stabilized with a confidence interval because they should fall on a bell curve. 

Standard deviation, Regression, Cohort, Confidence Interval…

You boys are deep into the weeds this morning!…

Took a couple statistics courses 40 years ago…

You’re bringing back some awful memories 😱

[Pimlach]

1 hour ago, LGR4GM said:

The catch is I don't think they account for shooting talent. Meaning a Krebs shot is worth a Tage shot in xgf because it's based on league average. Maybe team models do? That's why you need a data engineer in your analytics team. 

That is why I asked about Kucherov vs Robinson as shooters.   There is a case to not consider shooting ability and call it an expected goal for all.  Then you compare xgf and actual goals to find who gets chances but does not finish and work on that.  
 

You can get this information other ways,  but plotting it on a map of the ice surface give the coaches and players hard data to work with. 

Way back in the 70’s, I did shot charts for our school basketball team.   I marked who shot and from where on pages with the court layout on it, I circled the ones that went in. The coach used the data and kept the sheets in a big binder.  Of course, was at the mercy of the charters for accuracy.  No PCs and spreadsheets to compile the data, no statistical analysis tools at that time, but he figured out what he wanted  from it.  He had students tracking and charting assists and rebounds too.   He was a fantastic history teacher and our basketball coach.  Great guy.   He recruited the charters from his classes.   This conversation takes me back to a good place.  

Edited April 21 by Pimlach

[SABRES 0311]

Right now I say UPL is generally better in that he is more developed. Raw talent goes to Levi.

[Demoted]

Levi needs to stop being sent down, I don't know if that was stupid Granato or not but UPL is the better goalie right now.

[inkman]

On 4/20/2024 at 9:39 PM, GoPuckYourself said:

It's a bunch of nerds getting together to see who's pocket protector is bigger. I'm sure if you deep dive into this there is some fatal flaw like anything. 

I can get a compass (ouch), protractor and and abacus in mine.  What about you?  

53 minutes ago, Demoted said:

Levi needs to stop being sent down, I don't know if that was stupid Granato or not but UPL is the better goalie right now.

Like he should have never left Buffalo?  Because he absolutely needed to be sent down. His stint in Buffalo went about as poorly as it could have. 

[Drag0nDan]

On 4/20/2024 at 2:02 PM, inkman said:

Regardless, unless Upie takes a big step back and Levi in undeniable, I’d like to see Devon spend next season in Rochester.  Plenty of dudes out there they can get to play 20 games if need be. 

See thats my issue.  The "dudes" out there are terrible when they play here for whatever reason. 3 dudes played 60 or more games this past year - UPL was tied for 11th at 54 starts.  I'm fine bringing someone in I guess, but I don't think its fair to levi to basically say "you're going to the AHL out of camp because 56 (plus another 34 in college) starts over two seasons isn't enough and we want you to see more pucks".  

As a team that is at best "on the bubble", can we really afford to continue to trot out goalies who are worse than the one in the AHL because "development"?  Or should we be looking at something closer to an even split ideally?  

Comrie was 11-16-1 here 

Hutton was 31-49-10

Dell was 1-8-1

Tokarski was 12-20-7

[dudacek]

If the object is to win next season, any dude we bring in has to be at least as good as Levi if we are pencilling in for 25-30 games.

Winning, not Levi’s development takes precedence.

[inkman]

2 hours ago, dudacek said:

If the object is to win next season, any dude we bring in has to be at least as good as Levi if we are pencilling in for 25-30 games.

Winning, not Levi’s development takes precedence.

Yes. Have an open competition.  If Levi is better, then keep him up.  If not, Amerks fans rejoice. Given how Adams has handled goaltending to date, no chance he gets this right. 

[bob_sauve28]

 

 

[PerreaultForever]

Both young. Both have weaknesses. Neither is a proven commodity. I am hopeful they will continue to develop and become solid NHL goalies but they might have some ups and downs still getting there. Goalies are weird and fragile (mentally more than physically).