2016-17 Centennial Leafs General Team Discussion

[mr grieves]

Some stats to throw at you (all 5-on-5 unadjusted from this season):

Gardiner CA/60 55.41  FA/60  42.81  SA/60  31.14  GA/60  2.78
Carrick  CA/60 53.41  FA/60 40.15  SA/60  29.55  GA/60  3.16
Hunwick  CA/60  61.18  FA/60  42.41  SA/60  29.55  GA/60  2.43
Polak  CA/60  65.38  FA/60  48.76  SA/60  34.05  GA/60  2.45

Take a look at the last column.  Why do Hunwick and Polak, despite being our shittiest "possession" defensemen, have GA/60 rates that are much better than the rest of our defenders?  (Gardiner is 3rd, the rest are all above 3.0 and Marincin is sitting at an ugly 3.92) 

In Hunwick's case he may allow 6 more shot attempts than Gardiner over 60 minutes, but less of them hit the net.  Good positioning?  Forcing shots from further out?  Luck?  Well, luck does have a bit to play into it as his PDO is 101.9 but Carrick and Gardiner also sit above 101. 

But those GA/60 numbers are only half the story, and, to go back to that Extra 2% article, that Hunwick and Polak are slightly better at limiting goals than other defensemen is off-set by their inability to generate anything in the way of offense. They're not giving you much that'll help win a game.

.................	GF60.....	GA60.....	GF%......
Carrick	4.13	3.22	56.25
Gardiner	3.60	2.84	55.88
Hunwick	2.85	2.50	53.33
Polak	1.95	2.51	43.75

 

[herman]

An article echoing the numbers Coco-puffs posted about:
http://hockeyanalysis.com/2016/11/26/theoretical-problems-vs-real-problems-of-the-toronto-maple-leafs/

Here is the problem. Dom has put significant (absolute?) trust in his model and as a result he is certain that making this switch will make the Leafs better. Dom has used his theoretical model to identify a theoretical problem (sub-optimal lineups) with the Leafs and proposed a theoretical solution. In theory it should all work out.

The reality is the four guys Dom identified as being problem players for the Leafs have the best GA60 ratings on the team.

 

[Frank E]

I'm curious to know if the relative numbers of the 4th line/3rd pairing is dissimilar to other teams'?

 

[WhatIfGodWasALeaf]

Good back and forth here folks, high level discussion without anyone being a contrarian just for the sake of it.

Thanks.

 

[Arn]

I see Prust has signed with the Nurnberg Ice Tigers in Germany. Where he'll play alongside David Steckel

 

[herman]

An article echoing the numbers Coco-puffs posted about:
http://hockeyanalysis.com/2016/11/26/theoretical-problems-vs-real-problems-of-the-toronto-maple-leafs/

Here is the problem. Dom has put significant (absolute?) trust in his model and as a result he is certain that making this switch will make the Leafs better. Dom has used his theoretical model to identify a theoretical problem (sub-optimal lineups) with the Leafs and proposed a theoretical solution. In theory it should all work out.

The reality is the four guys Dom identified as being problem players for the Leafs have the best GA60 ratings on the team.

A follow up response to the response.

http://theleafsnation.com/2016/11/26/the-extra-2-continued-a-response-to-a-response

The highlight here is the usage of GA60 or any goal-stat in a 20 game sample being very spiky, as they are high susceptible to shooting percentage swings. It's the reason Corsi/shot attempt stats are favoured over goal-stats as goals are super rare and luck wonky. This is not to say shot attempts are a complete picture either, just a better representation of possession over long periods of time.

 

[herman]

https://mapleleafshotstove.com/2016/11/26/toronto-maple-leafs-q1-player-grades-part-1-forwards/

The most accurate quarter mark assessment I've seen so far. Which is to say the one I agree with the most

 

[mr grieves]

An article echoing the numbers Coco-puffs posted about:
http://hockeyanalysis.com/2016/11/26/theoretical-problems-vs-real-problems-of-the-toronto-maple-leafs/

Here is the problem. Dom has put significant (absolute?) trust in his model and as a result he is certain that making this switch will make the Leafs better. Dom has used his theoretical model to identify a theoretical problem (sub-optimal lineups) with the Leafs and proposed a theoretical solution. In theory it should all work out.

The reality is the four guys Dom identified as being problem players for the Leafs have the best GA60 ratings on the team.

A follow up response to the response.

http://theleafsnation.com/2016/11/26/the-extra-2-continued-a-response-to-a-response

The highlight here is the usage of GA60 or any goal-stat in a 20 game sample being very spiky, as they are high susceptible to shooting percentage swings. It's the reason Corsi/shot attempt stats are favoured over goal-stats as goals are super rare and luck wonky. This is not to say shot attempts are a complete picture either, just a better representation of possession over long periods of time.

I looked at the GF% for the Leafs on Corsica and found the following (first chart is top performers, 2nd chart is worst):

But, as the first column shows, some of those with the best and worst GF% have high and low PDOs.

So, here's xGF%, both best and worst:

Looks to me like it is possible the Leafs aren't good but are getting lucky with Smith, Martin, and Hunwick in the line-up.

 

[ZBBM]

Has anyone taken a objective look at the whole suite of advanced stats to see which ones -- if any -- really correlate with team success?  From where I sit, none of the main ones that get quoted all the time seem terribly advanced -- only in comparison with what came before, which was essential so primitive (+/?) as to be worthless.  Sure, they are better than (the previous) nothing, but taken on their own ...?

 

[herman]

Has anyone taken a objective look at the whole suite of advanced stats to see which ones -- if any -- really correlate with team success?  From where I sit, none of the main ones that get quoted all the time seem terribly advanced -- only in comparison with what came before, which was essential so primitive (+/?) as to be worthless.  Sure, they are better than (the previous) nothing, but taken on their own ...?

Check out this thread: https://m.reddit.com/r/hockey/comments/4xvvjp/what_is_the_most_useful_assist_stat_for/

The first main answer gives a really good jumping off point.

GF% in a short time frame is more descriptive than predictive. It's only one step up in complexity from just counting box stats.

xGF% on the other hand, accounts for more than just goal counting, and tries to algorithmically weigh the context of the goals for repeatability measures. There are currently two versions of xGF: Corsica's and DTM About Heart's. The primary difference is that DTM's accounts for regressed shooting percentage. In the podcast linked up thread, Garret Hohl considers DTM's more predictive than Corisca's.

 

[herman]

https://hockey-graphs.com/2016/10/24/a-primer-on-dtmabouthearts-war-model/

This is where analytics appears to be heading (and probably already there for some teams). It's a wins Above Replacement value for a player. Hit the tag at the end of the post to see the full 5 write ups. I'm still getting through them.

 

[McGarnagle]

https://mapleleafshotstove.com/2016/11/26/toronto-maple-leafs-q1-player-grades-part-1-forwards/

The most accurate quarter mark assessment I've seen so far. Which is to say the one I agree with the most

Agreed, that was pretty much bang-on. Only thing I'd do is drop Martin a half a letter grade.

I have a feeling that the D won't be a rosy a picture though in their "part 2"

 

[ZBBM]

https://hockey-graphs.com/2016/10/24/a-primer-on-dtmabouthearts-war-model/

This is where analytics appears to be heading (and probably already there for some teams). It's a wins Above Replacement value for a player. Hit the tag at the end of the post to see the full 5 write ups. I'm still getting through them.

Thanks for the links herman.  I'm going to delve into them properly soon, but just skimming the above link reminded me of a thought I'd had.  The author talks about how hard hockey is to analyze due to the nature of the game, which (when you think about it) is kind of a staccato fluidity -- lot of player movement but punctuated by all kinds of stops, starts, reversals of direction etc.. 

It seems to me that hockey stats won't really mature until they put GPS chips in each player's skates (or whatever technology allows them to track every movement on the ice) and then build a whole suite of stats around that -- not just looking at each player individually, but where he is on the ice relative to the competition.  Then you'd really  know whether (for example) a defenseman was giving too much space to a particular forward etc.  Then you could begin doing real regression analysis and so forth.

EDIT: Just read the first answer on the reddit thread you referenced.  I'm no statistician but that seems like pretty solid thinking there.

 

[WhatIfGodWasALeaf]

20 games in I know...

Does anyone want to ball-park Zaitsev's next contract?

He is RFA at the end of the season.

He is on pace to score 31 points presently.

He hasn't scored a goal yet and is starting to be used more on the PK than the PP since Marincin has left the lineup.

He is averaging 22 minutes a night and is already a top pairing dman on the team.

Do you think they will risk a bridge contract or try to lock him up at a discounted rate long term?

He makes $925K this year.

 

[Kin]

3 years, 13.5?

 

[WhatIfGodWasALeaf]

3 years, 13.5?

I wouldn't complain too much about that.

I was wondering if just in terms of seniority they ask him to take just under Gardiner's figure.

Maybe something like 4 years 15 million?

 

[bustaheims]

Yeah. Somewhere in the $3.5M-$4.5M per season range sound about right.

 

[herman]

What about a Ryan Ellis-ish contract. 4-5 years @ $2.5-3M AAV?

These RFA Defensemen contracts have all ended up lower than I expected.

I like Zaitsev and I'd be okay going up to 30 years old with him.

 

[Deebo]

I wouldn't be surprised if we see a 6 year deal for him.

 

[WhatIfGodWasALeaf]

I wouldn't be surprised if we see a 6 year deal for him.

Yeah, if Babcock falls in love he tends to make sure his favorites are taken care of.

It's part of what has me a little worried about Hyman's next contract.