Improve the Pens by making a bold coaching change using advanced stats as justification

[wgknestrick]

Rules are this:

1. You have Sully's job and can make roster / utilization moves (IE no trades) to improve the team that MUST be backed by some advanced stat that ultimately helps the Pens win more games.

2. Moves must be "bold" ie unconventional and maybe not what most would think would be beneficial

3. Don't forget to provide advanced stats proof and link.


Mine (these are the most beneficial I could find in terms of expected goal diff / win%) :

1. Bench Zucker, and Lafferty.
(2 worst GF% even strength forwards). Benching Laff may not be "bold", but I can never see Zucker being benched. They both consistently lose PIT games when they are on ice at 5v5 and are greatly outscored. Not breaking 40% GF% at 5v5 is very poor and not helping us win games. Zucker has the worst GA/60 of any forward on the team! He's already at -4 for this month, and he's not necessarily an offensive powerhouse.
Player Season Totals - Natural Stat Trick

2. Remove Letang and Rust from all PK duties.
(2 worst skaters at 4v5)
JUSTIFICATION: Letang has the worst GA/60 of any Penguin D by a mile at 4v5 at 12.5 GA/60. Rust is even more at 15GA/60. I wouldn't have thought this would be a smart move, but neither has been getting good results so far and PIT has a poor PK that needs solutions. I don't like our skilled guys eating pucks either and risking injuries.
Player Season Totals - Natural Stat Trick

This allows these 2 players more minutes to contribute where they have been more effective this year. Both are fantastic players that could help with minutes in other situations where they are both dominant (PP and 5v5)

Click to read more...

 

[Empoleon8771]

Ways to improve the Penguins: don't do something stupid based purely on analytics when the team is firing on all cylinders.

 

[Peat]

I'm not sure using GF% really counts as an advanced stat...

 

[ChaosAgent]

It amazes me that GF% is an advanced stat and +/- isn't. They're basically the same thing.

Then again most hockey analytics are just more granular +/-. Count & ratio.

 

[Gurglesons]

Lol. Remove Letang from the PK because he plays the most minutes in a short handed situation and gets scored on because of it.

 

[SomeDude]

Ways to improve the Penguins: don't do something stupid based purely on analytics when the team is firing on all cylinders.

You forgot to provide your source/proof.

Here you go:

 

[ronduguayshair]

Terrible post.

 

[SEALBound]

Easy, don't do it.

Justification: Arizona Coyotes under Chayka.

 

[ziggyjoe212]

My bold coaching move would be to pump Sid, Geno, and Tang with steroids.

And also to demote Lafferty to the AHL.

 

[wgknestrick]

Lol. Remove Letang from the PK because he plays the most minutes in a short handed situation and gets scored on because of it.

Gets scored on the most PER MINUTE of PK time.

 

[theicebox]

 

[CheckingLineCenter]

If I was coach I’d play Kap on the PK. It sucks. He’s good at it. And more in the 3v3. The only thing Sully does or won’t do that I truly don’t understand.

My healthy lines would look different but that’s just personally how I’d set up the team, not necessarily a gripe at Sully.

Other than that Sullivan once again has shown he can press the correct buttons when this team is short handed and banged up. Really hard to criticize the coaching too much even if I haven’t loved the lines this year. It just doesn’t matter when he’s getting results. Hopefully he can get back to his 16/17 postseason ways in a month.

I think your changes aren’t warranted at this moment in time. Lafferty should be scratched when/if the Fs return to full health though.

 

[Peat]

If I was coach I’d play Kap on the PK. It sucks. He’s good at it. And more in the 3v3. The only thing Sully does or won’t do that I truly don’t understand.

My healthy lines would look different but that’s just personally how I’d set up the team, not necessarily a gripe at Sully.

Other than that Sullivan once again has shown he can press the correct buttons when this team is short handed and banged up. Really hard to criticize the coaching too much even if I haven’t loved the lines this year. It just doesn’t matter when he’s getting results. Hopefully he can get back to his 16/17 postseason ways in a month.

I think your changes aren’t warranted at this moment in time. Lafferty should be scratched when/if the Fs return to full health though.

A tricky one that. Kapanen was one of Toronto's worse forwards for expected goals against, but that's balanced by being ludicrously dangerous. This year's version of the PK has been pretty conservative (in fact, the bold coaching change based on advanced stats should be to tell Vellucci to run the PK like Martin did, as we've just got worse there), and I'm not sure whether Kapanen would make that better, or be stymied by the system.

Incidentally, Goalton Waviour is probably our best PKing forward based on this season. But I wouldn't recommend playing him more.

You can also make a case for Pettersson-Letang and Dumoulin-Letang being the same thing based on fancy stats, so why not play the former and let Dumo rejoin Marino in our best pairing to play 100 minutes over the last three seasons... but I'm fairly sure actually using stats involves muttering the word "sample size" at that one over and over

 

[orby]

If the worst problem with the team is that Jason Zucker is having a disappointing season...I think they're probably in pretty good shape.

 

[wgknestrick]

It amazes me that GF% is an advanced stat and +/- isn't. They're basically the same thing.

Then again most hockey analytics are just more granular +/-. Count & ratio.

Please understand the HUGE issue with +- that isn't present in GF% is that GF% is consistent and fair across all players. +- penalizes PP players because they can only record a - in those situations while PKers get the benefit of recording extra +s when they score short handed. While both of those instances are indeed + and - to the team and players, they are captured in what is primarily an even strength stat. You can't compare +- between players because it is based on the same basis. NHL needs to remove +- all together as it is very dated, but was made out of good intentions. It is tough to communicate the "idea" of goal differential (while on ice) to normal posters and fans here without using it though.

Many people don't understand that a player with a GF% of 60% is a dominant player, while a 40% one is losing you a game about once a week. The 20% GF difference just doesn't seem to be that large because it lacks "goal currency" compared to the +15 to -15 that a +- stat might show between them.

I admit that it's difficult to learn about these thing now as the golden age of 2007-2012 ish hockey analytics has been bought up by the NHL. Sites go down left and right as bloggers and data miners are hired by clubs.

 

[ChaosAgent]

Please understand the HUGE issue with +- that isn't present in GF% is that GF% is consistent and fair across all players. +- penalizes PP players because they can only record a - in those situations while PKers get the benefit of recording extra +s when they score short handed. While both of those instances are indeed + and - to the team and players, they are captured in what is primarily an even strength stat. You can't compare +- between players because it is based on the same basis. NHL needs to remove +- all together as it is very dated, but was made out of good intentions. It is tough to communicate the "idea" of goal differential (while on ice) to normal posters and fans here without using it though.

Many people don't understand that a player with a GF% of 60% is a dominant player, while a 40% one is losing you a game about once a week. The 20% GF difference just doesn't seem to be that large because it lacks "goal currency" compared to the +15 to -15 that a +- stat might show between them.

I admit that it's difficult to learn about these thing now as the golden age of 2007-2012 ish hockey analytics has been bought up by the NHL. Sites go down left and right as bloggers and data miners are hired by clubs.

I'm well aware that +/- has the PP/PK idiocy.

It doesn't affect guys numbers that much so I just mentally adjust for it when thinking about +/-.

GF, xGF, Corsi, Fenwick etc. etc. all they're doing is just looking at outcomes while people are on the ice.

It's like when McDonald's puts McRib on the menu. All the above stats just tell you "sales are up." They don't actually tell you how much of those sales were driven by the McRib.

Edit: I also understand the magnitude of 40% being terrible and 60% being great. I just make fun of hockey stats nerdz who made a life of sh1tting on +/- then list a bunch of stats that are just mathematical transformations or more granular outcomes than +/- is.

 

[Peat]

I'm well aware that +/- has the PP/PK idiocy.

It doesn't affect guys numbers that much so I just mentally adjust for it when thinking about +/-.

GF, xGF, Corsi, Fenwick etc. etc. all they're doing is just looking at outcomes while people are on the ice.

It's like when McDonald's puts McRib on the menu. All the above stats just tell you "sales are up." They don't actually tell you how much of those sales were driven by the McRib.

Edit: I also understand the magnitude of 40% being terrible and 60% being great. I just make fun of hockey stats nerdz who made a life of sh1tting on +/- then list a bunch of stats that are just mathematical transformations or more granular outcomes than +/- is.

Mm.

You could make a case that GF% is "Sales Are Up"

Then Corsi, Fenwick, Shots, are "Customer Interactions Are Up" "Drive Thru Service Times Are Up", then Low Danger, High Danger etc.etc. become your time of day - "Customers Down During Dinner Rush" "Customers Up During Morning". xGF% becomes "Projected Sales Based on Customers At Various Times of Day". Stuff that will point you to why GF% might be up.

The McRib is an individual player. Points are its sales. Adverts up are... okay, I've run out of this.

But I think you have the chain of tools wrong. And are also pointing at why stuff like Sznajder's microstats for zone exits/entries etc.etc. are useful, and why the pro hockey stats companies use a bunch of that that isn't available through public NHL data.

 

[ChaosAgent]

Mm.

You could make a case that GF% is "Sales Are Up"

Then Corsi, Fenwick, Shots, are "Customer Interactions Are Up" "Drive Thru Service Times Are Up", then Low Danger, High Danger etc.etc. become your time of day - "Customers Down During Dinner Rush" "Customers Up During Morning". xGF% becomes "Projected Sales Based on Customers At Various Times of Day". Stuff that will point you to why GF% might be up.

The McRib is an individual player. Points are its sales. Adverts up are... okay, I've run out of this.

But I think you have the chain of tools wrong. And are also pointing at why stuff like Sznajder's microstats for zone exits/entries etc.etc. are useful, and why the pro hockey stats companies use a bunch of that that isn't available through public NHL data.

I've never had a McRib, believe it or not. But pretty sure it's a fascinating case where it is more profitable as a limited-time only offering as a drives an incremental trip to McDonald's. Whereas if it were always on people would just shift behavior.

I think points are too limited as direct metrics honestly. It just measures the last 3 guys to touch the puck before a goal. Even an arbitrary weighting of the last 3 and qualitative attribution would probably do better. Like Guentzel tapped in a goal on a redonkulous feed from Rust last game. Credit should like 80% go to Rust.

Points also categorically undervalues defensemen by its nature.

But BTW if all you knew is that when you ran McRib over months and years, your sales were up...keep running McRib (as a limited time offer). In conclusion, this is why you bench Zuc...jk.

 

[ChaosAgent]

Mm.

You could make a case that GF% is "Sales Are Up"

Then Corsi, Fenwick, Shots, are "Customer Interactions Are Up" "Drive Thru Service Times Are Up", then Low Danger, High Danger etc.etc. become your time of day - "Customers Down During Dinner Rush" "Customers Up During Morning". xGF% becomes "Projected Sales Based on Customers At Various Times of Day". Stuff that will point you to why GF% might be up.

The McRib is an individual player. Points are its sales. Adverts up are... okay, I've run out of this.

But I think you have the chain of tools wrong. And are also pointing at why stuff like Sznajder's microstats for zone exits/entries etc.etc. are useful, and why the pro hockey stats companies use a bunch of that that isn't available through public NHL data.

I've never had a McRib, believe it or not. But pretty sure it's a fascinating case where it is more profitable as a limited-time only offering as a drives an incremental trip to McDonald's. Whereas if it were always on people would just shift behavior.

I think points are too limited as direct metrics honestly. It just measures the last 3 guys to touch the puck before a goal. Even an arbitrary weighting of the last 3 and qualitative attribution would probably do better. Like Guentzel tapped in a goal on a redonkulous feed from Rust last game. Credit should like 80% go to Rust.

Points also categorically undervalues defensemen by its nature.

But BTW if all you knew is that when you ran McRib over months and years, your sales were up...keep running McRib (as a limited time offer). In conclusion, this is why you bench Zuc...jk.

 

[CrosbyMalkin]

Rules are this:

1. You have Sully's job and can make roster / utilization moves (IE no trades) to improve the team that MUST be backed by some advanced stat that ultimately helps the Pens win more games.

2. Moves must be "bold" ie unconventional and maybe not what most would think would be beneficial

3. Don't forget to provide advanced stats proof and link.


Mine (these are the most beneficial I could find in terms of expected goal diff / win%) :

1. Bench Zucker, and Lafferty.
(2 worst GF% even strength forwards). Benching Laff may not be "bold", but I can never see Zucker being benched. They both consistently lose PIT games when they are on ice at 5v5 and are greatly outscored. Not breaking 40% GF% at 5v5 is very poor and not helping us win games. Zucker has the worst GA/60 of any forward on the team! He's already at -4 for this month, and he's not necessarily an offensive powerhouse.
Player Season Totals - Natural Stat Trick

2. Remove Letang and Rust from all PK duties.
(2 worst skaters at 4v5)
JUSTIFICATION: Letang has the worst GA/60 of any Penguin D by a mile at 4v5 at 12.5 GA/60. Rust is even more at 15GA/60. I wouldn't have thought this would be a smart move, but neither has been getting good results so far and PIT has a poor PK that needs solutions. I don't like our skilled guys eating pucks either and risking injuries.
Player Season Totals - Natural Stat Trick

This allows these 2 players more minutes to contribute where they have been more effective this year. Both are fantastic players that could help with minutes in other situations where they are both dominant (PP and 5v5)

Sullivan coached this team to a 16-3-1 record over the past 20 games while missing as many as 5 of his top 9 forwards. Yet some arm chair coaches believe they can do better.

This is ridiculous.

 

[Tom Hanks]

I've never had a McRib, believe it or not. But pretty sure it's a fascinating case where it is more profitable as a limited-time only offering as a drives an incremental trip to McDonald's. Whereas if it were always on people would just shift behavior.

I think points are too limited as direct metrics honestly. It just measures the last 3 guys to touch the puck before a goal. Even an arbitrary weighting of the last 3 and qualitative attribution would probably do better. Like Guentzel tapped in a goal on a redonkulous feed from Rust last game. Credit should like 80% go to Rust.

Points also categorically undervalues defensemen by its nature.

But BTW if all you knew is that when you ran McRib over months and years, your sales were up...keep running McRib (as a limited time offer). In conclusion, this is why you bench Zuc...jk.

I’d stick with the poutine. Really good chemistry. Also very flexible adding pulled pork or butter chicken.

Will get to the dirty areas..........eventually.

 

[Peat]

I've never had a McRib, believe it or not. But pretty sure it's a fascinating case where it is more profitable as a limited-time only offering as a drives an incremental trip to McDonald's. Whereas if it were always on people would just shift behavior.

I think points are too limited as direct metrics honestly. It just measures the last 3 guys to touch the puck before a goal. Even an arbitrary weighting of the last 3 and qualitative attribution would probably do better. Like Guentzel tapped in a goal on a redonkulous feed from Rust last game. Credit should like 80% go to Rust.

Points also categorically undervalues defensemen by its nature.

But BTW if all you knew is that when you ran McRib over months and years, your sales were up...keep running McRib (as a limited time offer). In conclusion, this is why you bench Zuc...jk.

Yes and no. I think - gut instinct - it has value as a canary in a coal mine. It's a good quick way to know something's going on or not. In terms of what's actually going on then yeah, I think you want to use more granular process stats - individual chances, point shots vs high danger shots set up etc.etc. - and maybe assign values into what goes into a point. I know I've seen someone do that assigning values but have lost the post where I saw it.

I do every now and again think about going through the Pens goals this season trying to do arbitrary weightings of who contributes to each goal but the problem is the NHL goal clips aren't long enough to get every detail.