Jump to content
Sign in to follow this  
IKnowPhysics

2017-18 Defensemen Pair Analysis

Recommended Posts

The Goal:

To better understand which defensive pairings were successful to inform our guesses about the upcoming defense lineup, I decided to look at how the different defensive pairings performed in 2017-18.  The metric I use for this analysis is CF%, which is Corsi For %.  I pulled the data from here.

About CF%:

Definition: CF% is a calculation of shot attempts for / (shot attempts for + shot attempts against).  It scales from 0-100; 0 is BAD; 100 is GOOD; 50 would mean equal shot attempts for and against; 66.67 would mean out shooting your opponent by a factor of 2 to 1.  Most CF% with good sampling ranges from 40 (BAD - getting outshot by a ratio of 1.5 to 1) to 60 (GOOD - outshooting your opponent by a factor of 1.5 to 1).  The CF% used here only considers 5v5 even strength shot attempts.

Pros of CF%: CF% is a good indicator of team puck possession, because, more or less, if your team has the puck, you're attempting shots, and if your team doesn't have the puck, you're facing shot attempts.  The rate of shot attempts occurring is much higher than say, goals scored, so sample sizes are much larger and statistically useful.

Cons of CF%: CF% can be rough and potentially meaningless without context.  Considering time on ice (or total shots for and against) is critical to establish sample size.  Corsi stats represent the entire group of players on the ice when the shot attempt occurs, so an individual's CF% can be skewed by the quality of teammates, quality of competition, and sometimes game situation.

The Data:

The following is a table of Sabres defensemen who played in 2017-18 and are continuing with the organization in 2017-18.  Each data point has two numbers: time on ice played together (in MM:SS) and CF%.  The single letters are that defenseman's handedness.

TOI/CF%		Ristolainen	Bogosian	Scandella	Beaulieu	McCabe		Nelson		Guhle		Fedun		Tennyson
Ristolainen	R		10:22/45.83	951:31/48.53	170:06/48.78	46:09/44.44	30:49/37.33	77:19/48.70	0:27/50.00	7:25/30.77
Bogosian			R		10:30/50.00	69:25/58.06	170:24/46.25	0:00/00.00	0:45/50.00	0:00/00.00	0:00/00.00
Scandella					L		35:15/50.72	34:40/47.62	198:39/42.39	10:00/64.00	4:04/81.82	46:33/49.38
Beaulieu							L		92:46/53.85	58:07/40.38	5:00/50.00	0:27/00.00	63:22/53.73
McCabe										L		21:09/52.94	27:19/51.85	54:04/48.84	76:31/42.75
Nelson												R		127:15/56.80	0:00/00.00	2:25/42.86
Guhle														L		0:00/00.00	0:00/00.00
Fedun																R		0:51/100.00
Tennyson																	R

Ice Times of Pairs:

Examine the times.  First, disregard any pairing with less than ~15:00 together - this would be less than one game played together, and could easily be incidental time on ice (scattered line changes, etc).

Note the pairing that logged the biggest minutes: Scandella-Ristolainen 951:31.  They were paired together pretty steadily, as they were: LD-RD pair, healthy, and produced an ok-but-not-great CF% of 48.53.  The next longest pairings were Scandella-Nelson 198:39, Bogosian-McCabe 170:24, Beaulieu-Ristolainen 170:06, and Guhle-Nelson 127:15.  That's a big dropoff, and the next set of times is an even bigger dropoff.  It's clear that Housley leaned on Scandella-Ristolainen and was trying to find something -anything- beyond that pair that he liked.  Player Usage Charts confirm this: Scandella-Ristolainen took long, hard minutes, creating a shelter for the rest of the defensemen.

Bogosian logged minutes with Beaulieu and McCabe, but not Fedun or Tennyson- it's clear that he was injured and these players were brought up to replace him; they're also the same side and wouldn't be paired together.  It's also clear that Housley did NOT want the callups playing on the ice together: Fedun only played with McCabe, and Tennyson only had time with veterans -never with Guhle.

 

Veteran Performance via CF%:

Ristolainen had significant time with Scandella, Beaulieu, Guhle, and McCabe.  The near-identical CF%s of ~48.7 show that it doesn't matter who Risto plays with as long as it isn't McCabe at 44.44.

Scandella spent most of his time with Ristolainen, but during some changeups, he played with McCabe, Nelson, and Tennyson, mostly without any notable success.  This meh-ness with others only encouraged Housley to keep him with Ristolainen and eating up the heavy minutes.

Bogosian's peak ice time with McCabe makes intuitive sense: at the start of the season, they seemed like the clear-cut 3-4 guys.  But Beaulieu-Bogosian at 58.06 performed much, much better than McCabe-Bogosian at 46.25.  B-B ended up posting the best CF% of any significant ice time pair.

Beaulieu didn't only have a positive impact on Bogosian's game.  Beaulieu was the best statistical CF% partner for several other players: Ristolainen, Scandella(!), McCabe, and Tennyson.

McCabe has a rough season, in which he struggled to play against better competition with his most steady defense partner, Bogosian.  With Bogosian out, Housley spent the rest of the season trying to match McCabe with literally anybody on the team, left or right.  He saw some success with Beaulieu and, briefly, Nelson and Guhle.

 

Youngins Performance via CF%:

Nelson played with Scandella for a while and that was an awful 42.39.  He was put with Beaulieu and did worse 42.39.  And then Ristolainen(!) - even worse 37.33!  But Nelson played pretty good with McCabe 52.94, and found a real niche with Guhle for the second-best CF% at 56.80.  It think it was quality of competition- hard minutes with Risto, Scandella, and Beaulieu and he couldn't hack it.  Easy minutes with Guhle aren't so bad.

Guhle.  Guhle.  The up-and-comer.  The usage charts show that he was a sheltered noob, and in this role, he found success with Nelson at 56.80 and McCabe 51.85.  But he also did as well as Scandella did with Ristolainen at 48.70- while not great, because he survived those harder minutes, he showed what he hopes to be capable of someday.  Of Ristolainen's LD partners outside of Scandella, Guhle played best.

Fedun, not really a youngin at age 30, played a couple games with McCabe.  He didn't show much.

Housley played Tennyson with Scandella, Beaulieu, and McCabe trying to find a fit.  His best posession was with Beaulieu (53.73).

 

Optimizing Pairs:

Without useful pair data on Dahlin (L) or Hunwick (L), it's hard to know what will actually happen.  But based on historical data: these could be the best pairs without them:

Scandella-Ristolainen 48.53

Beaulieu-Bogosian 58.06

McCabe-Nelson 52.94

(Guhle with Nelson 56.80)

 

 

Edited by IKnowPhysics
  • Like 2
  • Thanks 1

Share this post


Link to post
Share on other sites

Awesome work sir.  

Interesting positive take on Beaulieu.  

Share this post


Link to post
Share on other sites

As you said though, it does not account for the other 3 schmucks on the ice right?

Is it reasonable to assume that the forward pairings would even out somewhat over the course of the season such that it still allows for meaningful comparison?

Share this post


Link to post
Share on other sites

I think my favorite stat for any player really is the heat maps. Corsi just can't account for shot quality, and the numbers are usually so close when comparing them that it's so hard to say what the defining difference

Share this post


Link to post
Share on other sites
2 minutes ago, LTS said:

As you said though, it does not account for the other 3 schmucks on the ice right?

Is it reasonable to assume that the forward pairings would even out somewhat over the course of the season such that it still allows for meaningful comparison?

Right.  Very difficult to pick that apart without further deep diving.  Also difficult to account for is quality of teammate and quality of competition.  Looking at the Player Usage Charts, you see that Housley used what he thought were the best Sabres players in big minutes against the opposition's best players (ie, Risto, ROR, Eichel, Scandella, Reinhart).  Generally, the Sabres players that had worse teammates also fewer minutes and faced easier competition.

I think the sample sizes were large enough, mostly to wash out some of that, but likely not all of it.

9 minutes ago, WildCard said:

I think my favorite stat for any player really is the heat maps. Corsi just can't account for shot quality, and the numbers are usually so close when comparing them that it's so hard to say what the defining difference

Then get to work.

Share this post


Link to post
Share on other sites
4 minutes ago, IKnowPhysics said:

Right.  Very difficult to pick that apart without further deep diving.  Also difficult to account for is quality of teammate and quality of competition.  Looking at the Player Usage Charts, you see that Housley used what he thought were the best Sabres players in big minutes against the opposition's best players (ie, Risto, ROR, Eichel, Scandella, Reinhart).  Generally, the Sabres players that had worse teammates also fewer minutes and faced easier competition.

I think the sample sizes were large enough, mostly to wash out some of that, but likely not all of it.

Then get to work.

Thanks.  All good stuff.  A great post without a doubt.

Also key to point out that cutting and pasting things into this forum is so much easier than before.  It might be the thing I like the most about it.

Share this post


Link to post
Share on other sites

I almost flipped my when I couldn't find the "erase formatting" button before I saw the "paste without formatting" notification.

Share this post


Link to post
Share on other sites
42 minutes ago, IKnowPhysics said:

Then get to work.

I'll get to it in the coming days when I'm not feeling like sh!t. So many charts in there I think I'll just end up making a separate thread for each because of the picture/attachments limits

Edited by WildCard

Share this post


Link to post
Share on other sites

This is where fancy stats and eye test seem to be inconsistent.  Beaulieu looked like dog #### most of the year in my view but this analysis says he should be second pairing and everyone's corsi basically improves with him.  I don't understand but there's some disconnect there.  I do really like that analysis tho. 

Share this post


Link to post
Share on other sites
28 minutes ago, Derrico said:

This is where fancy stats and eye test seem to be inconsistent.  Beaulieu looked like dog #### most of the year in my view but this analysis says he should be second pairing and everyone's corsi basically improves with him.  I don't understand but there's some disconnect there.  I do really like that analysis tho. 

Yah. IKP did a wonderful analysis, and about as much work as somebody not employed in a hockey-stats-postion can, but Beaulieu doesn't look like a bad hockey player from plenty of lenses that don't involve diving into film. It's hard to quantify his bone-head cross-ice passes at the point that get picked off and taken to the house, but they happen. It's hard to quantify how many of his shots are ill-advised and lose possession after hitting shin-pads three feet in front of him, but it happens. It's hard to quantify how many blown assignments he has and how often he misplaces himself as a safety valve for his partner, but he's notoriously horrific at doing both of those things. 

Edited by Randall Flagg

Share this post


Link to post
Share on other sites
17 minutes ago, Randall Flagg said:

Yah. IKP did a wonderful analysis, and about as much work as somebody not employed in a hockey-stats-postion, but Beaulieu doesn't look like a bad hockey player from plenty of lenses that don't involve diving into film. It's hard to quantify his bone-head cross-ice passes at the point that get picked off and taken to the house, but they happen. It's hard to quantify how many of his shots are ill-advised and lose possession after hitting shin-pads three feet in front of him, but it happens. It's hard to quantify how many blown assignments he has and how often he misplaces himself as a safety valve for his partner, but he's notoriously horrific at doing both of those things. 

The thing I always have trouble squaring is...if he does these things with the frequency he is accused, how could he possibly produce useful metrics? It's like the Thomas Vanek is lazy meme that was so popular. My hypothesis is that Beaulieu's mistakes are so comically glaring that they get seared into our minds in a fashion that does not lend itself to the human brain putting it in proper perspective relative to what he does bring to the ice.

This is not to say I think Beaulieu is good. I certainly do not. But I also do think he's a viable 3rd pair NHL defender, whereas I think the consensus is he's an AHL player.

Share this post


Link to post
Share on other sites
1 minute ago, TrueBlueGED said:

The thing I always have trouble squaring is...if he does these things with the frequency he is accused, how could he possibly produce useful metrics? It's like the Thomas Vanek is lazy meme that was so popular. My hypothesis is that Beaulieu's mistakes are so comically glaring that they get seared into our minds in a fashion that does not lend itself to the human brain putting it in proper perspective relative to what he does bring to the ice.

This is not to say I think Beaulieu is good. I certainly do not. But I also do think he's a viable 3rd pair NHL defender, whereas I think the consensus is he's an AHL player.

Bad players can produce "useful metrics" all of the time. Deslauriers put up P/60 equivalent to Okposo and had expected goal metrics, raw and team-relative on a bad team which could suggest similar impact as Eichel, even though he was a bottom 3 regular forward in the entire league while here.

Share this post


Link to post
Share on other sites
1 minute ago, Randall Flagg said:

Bad players can produce "useful metrics" all of the time. Deslauriers put up P/60 equivalent to Okposo and had expected goal metrics, raw and team-relative on a bad team which could suggest similar impact as Eichel, even though he was a bottom 3 regular forward in the entire league while here.

There's gotta be something that accounts for that though. I don't know who  and I don't know when, but I like to believe at some point someone will find the holy grail of stats

Share this post


Link to post
Share on other sites
1 minute ago, Randall Flagg said:

Bad players can produce "useful metrics" all of the time. Deslauriers put up P/60 equivalent to Okposo and had expected goal metrics, raw and team-relative on a bad team which could suggest similar impact as Eichel, even though he was a bottom 3 regular forward in the entire league while here.

Deslauriers did not produce useful metrics when looking at the whole picture. For instance, his possession impact was so negative you'd swear it was a made up number.

2 minutes ago, WildCard said:

There's gotta be something that accounts for that though. I don't know who  and I don't know when, but I like to believe at some point someone will find the holy grail of stats

The player tracking data is gonna be really cool when people start taking a deep dive into it. Holy grail? Probably not, but it will add a ton, I think, to what we already know.

Share this post


Link to post
Share on other sites
Just now, WildCard said:

There's gotta be something that accounts for that though. I don't know who  and I don't know when, but I like to believe at some point someone will find the holy grail of stats

If a doctor prescribed some medicine to you and the medicine had the same correlation factor as the best advanced stats do on their own, and the intended effect was watching Nate Beaulieu skate 82 games a year 3 hours a night, would you take the medicine because it says Beaulieu ain't so bad? 😛

Share this post


Link to post
Share on other sites
1 minute ago, TrueBlueGED said:

Deslauriers did not produce useful metrics when looking at the whole picture. For instance, his possession impact was so negative you'd swear it was a made up number.

The player tracking data is gonna be really cool when people start taking a deep dive into it. Holy grail? Probably not, but it will add a ton, I think, to what we already know.

😂

Mean probably not, there never will be. But we can dream. I always find it amazing that hockey is really the only sport with such a large unknown about advanced stats

2 minutes ago, Randall Flagg said:

If a doctor prescribed some medicine to you and the medicine had the same correlation factor as the best advanced stats do on their own, and the intended effect was watching Nate Beaulieu skate 82 games a year 3 hours a night, would you take the medicine because it says Beaulieu ain't so bad? 😛

I feel like I just took the SAT's all over again after reading that

Share this post


Link to post
Share on other sites
3 minutes ago, TrueBlueGED said:

Deslauriers did not produce useful metrics when looking at the whole picture. For instance, his possession impact was so negative you'd swear it was a made up number.

The player tracking data is gonna be really cool when people start taking a deep dive into it. Holy grail? Probably not, but it will add a ton, I think, to what we already know.

What possession impact is that? His relative corsi is -3% according to Corsica, in the same realm as noted bad forwards Phil Kessel and Jeff Carter.

Share this post


Link to post
Share on other sites
19 minutes ago, Randall Flagg said:

What possession impact is that? His relative corsi is -3% according to Corsica, in the same realm as noted bad forwards Phil Kessel and Jeff Carter.

Are you looking at this year, or when he was with the Sabres? I'm talking about his last full year in Buffalo where he had the metrics of one of the worst players in the league. I believe he actually had a decent season with the Habs. 

Share this post


Link to post
Share on other sites
3 minutes ago, TrueBlueGED said:

Are you looking at this year, or when he was with the Sabres? I'm talking about his last full year in Buffalo where he had the metrics of one of the worst players in the league. I believe he actually had a decent season with the Habs. 

Oh, we're in agreement then. When he was here I argued that he was the literal worst player in the league. I'm saying that this year they claim he ain't so bad for Montreal, as a mostly-joking counter to Beaulieu maybe having some on-ice value, when we should really be politely yet firmly telling him to leave. 

Edited by Randall Flagg

Share this post


Link to post
Share on other sites

As a side note, I would give extreme caution towards drawing hard conclusions from heat maps, but Beaulieu's in the defensive zone is pretty funny, and his offensive one would also suggest correctly that he got most of his ice time with our cycle and shot-generating line (ROR, Reinhart)

Share this post


Link to post
Share on other sites

1). Holy Schmoly, RF and TBPHD.

2).  IKP .. Well done.  A glimpse not only into player analysis, but team construction.

3).  Beaulieu rhymes with who knew?

Share this post


Link to post
Share on other sites
31 minutes ago, Randall Flagg said:

Oh, we're in agreement then. When he was here I argued that he was the literal worst player in the league. I'm saying that this year they claim he ain't so bad for Montreal, as a mostly-joking counter to Beaulieu maybe having some on-ice value, when we should really be politely yet firmly telling him to leave. 

Oh, I'm not about to go to any great lengths to find value in Beaulieu. I just think he's better than most do, but still would like to move on from him after his contract ends. 

 

11 minutes ago, Randall Flagg said:
9 minutes ago, IKnowPhysics said:

Holy .  Nailed it.

My predictions at how people are going to react to certain hockey things >>>> my predictions about actual hockey things. 

  • Haha 1

Share this post


Link to post
Share on other sites
2 minutes ago, TrueBlueGED said:

Oh, I'm not about to go to any great lengths to find value in Beaulieu. I just think he's better than most do, but still would like to move on from him after his contract ends. 

 

My predictions at how people are going to react to certain hockey things >>>> my predictions about actual hockey things. 

And I can pretend to be smart like this because we all know that in reality, I have opinions like loving Jeff Skinner (and believing in the Hurricanes in general), betting everything I own on Vegas being a bottom 5 team even when they had a winning record in December, etc. 

Edited by Randall Flagg

Share this post


Link to post
Share on other sites

Create an account or sign in to comment

You need to be a member in order to leave a comment

Create an account

Sign up for a new account in our community. It's easy!

Register a new account

Sign in

Already have an account? Sign in here.

Sign In Now
Sign in to follow this  

×