RAINBOW RISING

by Lowetide

The Edmonton Oilers defeated the ghost of the Vancouver Canucks last night, and the truth is that the win was a lot tougher than it should have been. Part of it was Ryan Miller—the Oilers had all kinds of chances and it could have been 5-0—but Edmonton had some wobble that didn’t cost them last night. Remember that old line about winning games you should be losing right at the end of a good streak? Yeah. I think that is what we are seeing now. Sunday should be interesting.

CATCH THE RAINBOW, YEAR OVER YEAR

  • Oilers in October 2015: 3-5-0, goal differential -5
  • Oilers in October 2016: 7-1-0, goal differential +12

Edmonton is stretching out the goal differential now, we have talked about shooting percentage, save percentage and PDO and those things will regress to the meat at some point in time. What is this team’s real ability? The answer lies in another question: Is Connor McDavid on the ice? High-Danger Scoring Chances last night were 9-7 Oilers.

DEFENSE, LAST NIGHT

defense-last-night

  • Nurse—Gryba had a fine night, I saw several giving the veteran the lion’s share of the credit but young Nurse had a fine game, too. I am thinking of his clearing of a goal mouth puck during the heart of the game. Both men were rugged along the wall and battled well. They were up against the Granlund—Sutter—Hansen line most often according to NaturalStatTrick.
  • I have mentioned previously that Nurse appears to be taking forward steps, the game seems to be slowing down for him. Last night, on an evening where the Oilers were not especially sharp offensively as a group, I thought he battled through it and made good choices. He is close to 50 percent in Corsi (ranks No. 2 on the team) and has a very good Rel at this time. Now, he is not playing the Klefbom or Sekera minutes, so let’s not get ahead of ourselves. That said, progress is being made.
  • Klefbom—Larsson had a strong night again to my eye, both men passing with aplomb (and each getting an assist). I think this pairing really sets up the rest of the defense, and their TOI last night shows how much Todd McLellan thinks of them. Played a lot (over eight minutes) against Eriksson and the Piss Cutters, and with the Nuge line. That is a long, tall order and they managed it (with help from Cam Talbot). Well done. Only 74 to go.
  • Russell—Sekera got caved by the Corgis (but were even on high-danger scoring chances) and were under pressure for much of the night. The pairing faced Baertschi—Horvat—Virtanen most often, and were out against the pairing of Edler—Stecher and were playing most often with the Leon line. They should have done better, no? I think so. How much of this was score effects? Well, Edmonton didn’t win the possession battle in the third period (or the first), so I think we can assume this duo has had better days. HD SC were 2-2.

talbot-capture5

  • Cam Talbot was outstanding again, this is becoming a major story in this early run. He made one save that will be on the end of season reel, just a fabulous glove save. His .938 EV save percentage ranks him No. 4 in the NHL among goalies who have appeared in five or more games. He is also No. 4 overall among the same goalie group, sitting at .934.

FORWARDS, LAST NIGHT

forwards-last-night

  • Pouliot—Nuge—Kassian line had a great evening, I mean they were checking all over and making themselves difficult to play against. This is the first night I felt this line might have a 20-game run, that’s how well they played. Pretty excited about the idea of an internal solution. Played a ton with Klefbom—Larsson and played over 11 minutes against the Sedin line. I am so pleased with this line.
  • Maroon—Draisaitl—Puljujarvi had a good night by the numbers and were a bit unlucky on the chances. JP didn’t have much in his jet boots on a breakaway (was he gassed? he cannot be that slow) but rang iron later in the game. Leon had a nifty shot that could have been tapped in but no one was home. LD’s brilliant checking play on what would surely have been a goal was fantastic, very strong play by the big man. Played most against the Horvat line and split time with Sekera—Russell and the third pairing. I liked their game, need a little more speed and finnish.
  • Lucic—McDavid—Eberle scored both goals, had seven shots and made some nice defensive plays. Eberle twice interfered with substantial moments for Vancouver, and Milan Lucic used positioning to upset a sortie. The big man’s pass to send away McDavid for the winning goal was very smooth and reflects his strength as an offensive player. Eberle had a quiet night, but I thought he was effective and worked hard on the back check. I am liking his game more and more.
  • Lander—Letestu—Pitlick are like Borg parts these days, up and down the lineup as required. Lander got an assist by taking the faceoff on the McDavid goal, and Letestu went hither and yawn in pursuit of excellence as well. Letestu played over one minute with Lander (7), Pitlick (6), Leon, Maroon, Nuge, Kassian.

This is a large gap. Over at his site, G also had this to say about the game:

  • The gameflow chart shows a huge divergence between the ‘raw’ and the ‘danger’ curves in the third. That’s where that third period ‘keep it to the outside’ really showed. Not to mention the breakaways for! Source

The Oilers do seem to be playing a wiser game, more like the Finnish national style (everything forced to the outside, from the center line in) and it does seem to fit the roster. I also believe we are seeing the benefit of keeping a veteran coach around for more than one season. Remember the improvement we saw year over year with Renney? Perhaps the current boost is partly because of McLellan having a chance to impact the roster for more than one season. The team is still giving up the same number of shots as one year ago (30.5), but if the distance is less severe, well things may be improving for this team defensively.

156 comments
0

You may also like

0 0 votes
Article Rating
156 Comments
Newest
Oldest Most Voted
Inline Feedbacks
View all comments
jm363561

The converse, however, is not necessarily true. If the Oilers were league average at keeping shots to the outside they would have allowed 65 HDSC. They have allowed 61, so only 4 less than average, which is within the range of a single fire drill of a shift. There is much less, or no, evidence of a Larsson effect.

====

To determine a Larsson effect you would compare Oiler’s HDSCA last year to this year, not to league average. There has been a significant improvement.

Woodguy

Note, I wrote HDSCF/FA, but mean FF.

I used FF and not FA for the ratios

Woodguy

Bruce Wayne: IF last year 24% of FA were HDSC and this year it is 23% then there has been no change, at least in this regard.
I’m not saying Larrsson isn’t an effective player.I”m saying that he doesn’t have magical scoring chance reduction abilities.The improvement this year has nothing to do with reducing high danger scoring chances.That is what counting what happens (i.e. watching the game) tells us.

The improvement of the defense is Talbot, not the other way around.

I have no argument that Talbot is driving the good GA.

I’m arguing that Larsson is helping drive down HDSCA/FA as his rate is lower than the team.

Also,

Klef is dreamy

russ99

After seeing how effective Pouliot – RNH – Kassian are without the puck and on the cycle, I just don’t see why so many fans want to change that line.

Best Oilers shutdown line in a decade. As long as we get some scoring out of them I’m happy.

Cassandra

Woodguy:
Bruce Wayne,

So what does this mean. The Oilers are, in fact, creating better scoring chances than we’d expect. We’d expect 59 HDSC and they have instead created 78. This is a big difference. Call this the McDavid effect.

Not so fast……

OIler forwards HDSCF/FA

Ryan Nugent-Hopkins0.379
Jordan Eberle0.370
Milan Lucic0.367
Anton Slepyshev0.333
Benoit Pouliot0.333
Connor McDavid0.333
Patrick Maroon0.313
Jesse Puljujarvi0.286
Zack Kassian0.281
Leon Draisaitl0.262
Tyler Pitlick0.259
Mark Letestu0.257
Anton Lander0.214

Seems to be more team wide.

Nuge!!!!

Just need him to generate more FA/60!!

I’ll accept this modification of the thoery. It does appear to be team wide on offense.

Cassandra

Woodguy:
Bruce Wayne,

The converse, however, is not necessarily true. If the Oilers were league average at keeping shots to the outside they would have allowed 65 HDSC. They have allowed 61, so only 4 less than average, which is within the range of a single fire drill of a shift. There is much less, or no, evidence of a Larsson effect.

There is a good argument that getting to “league average” in terms of HDSCA/FA *IS* a Larsson effect.

Last year Oilers HDSCA/FA was 24%.

Larsson is at 20% so far this year.

Here’s the Oilers HDCA/FA for the Dmen:

Oscar Klefbom0.179
Kris Russell0.198
Adam Larsson0.202
Andrej Sekera0.250
Darnell Nurse0.264
Eric Gryba0.274

Last night Nurse and Gryba were 1 HDSCA on 6 FA.

If they can turn a corner it will help immensely

IF last year 24% of FA were HDSC and this year it is 23% then there has been no change, at least in this regard.
I’m not saying Larrsson isn’t an effective player. I”m saying that he doesn’t have magical scoring chance reduction abilities. The improvement this year has nothing to do with reducing high danger scoring chances. That is what counting what happens (i.e. watching the game) tells us.

The improvement of the defense is Talbot, not the other way around.

Woodguy

Bruce Wayne,

So what does this mean. The Oilers are, in fact, creating better scoring chances than we’d expect. We’d expect 59 HDSC and they have instead created 78. This is a big difference. Call this the McDavid effect.

Not so fast……

OIler forwards HDSCF/FA

Ryan Nugent-Hopkins 0.379
Jordan Eberle 0.370
Milan Lucic 0.367
Anton Slepyshev 0.333
Benoit Pouliot 0.333
Connor McDavid 0.333
Patrick Maroon 0.313
Jesse Puljujarvi 0.286
Zack Kassian 0.281
Leon Draisaitl 0.262
Tyler Pitlick 0.259
Mark Letestu 0.257
Anton Lander 0.214

Seems to be more team wide.

Nuge!!!!

Just need him to generate more FA/60!!

Cassandra

Lowetide: I don’t read David’s posts, so cannot comment on recent changes, but I would suggest that the problem with tracking scoring chances is bias. David’s data isn’t reliable is a debate we can have, but what isn’t in question is that we would be better off getting a bunch of kids from New Mexico State to track them. Less bias, and once the idea of a scoring chance is understood, then bias is the big issue. For David, you and me.

There are two kinds of bias. There is the bias you are talking about. But I don’t that’s the only problem here. The larger problem is outcome bias. NO matter who is doing the judging they are going to see goals as scoring chances, whether they were or not. So you have to define the scoring chance independent of the result, including whether the shot hit the net or not. The judgement of whether it is a scoring chance has to be completed before it gets to the goalie.

Woodguy

Bruce Wayne,

The converse, however, is not necessarily true. If the Oilers were league average at keeping shots to the outside they would have allowed 65 HDSC. They have allowed 61, so only 4 less than average, which is within the range of a single fire drill of a shift. There is much less, or no, evidence of a Larsson effect.

There is a good argument that getting to “league average” in terms of HDSCA/FA *IS* a Larsson effect.

Last year Oilers HDSCA/FA was 24%.

Larsson is at 20% so far this year.

Here’s the Oilers HDCA/FA for the Dmen:

Oscar Klefbom 0.179
Kris Russell 0.198
Adam Larsson 0.202
Andrej Sekera 0.250
Darnell Nurse 0.264
Eric Gryba 0.274

Last night Nurse and Gryba were 1 HDSCA on 6 FA.

If they can turn a corner it will help immensely

hunter1909

As a Canadian, I appreciate the opportunity to view the big game tonight either in French(Montreal) or English(Leafs).

böök¡je

I don’t why. I guess it’s because sometimes people just randomly post comments totally out of context for no reason. Maybe that explains it.

dustrock

Funny that Staples thought Klefbom Larsson were mediocre. Against the Sedins overall they played very well. Staples said probably should have had a goal against – well they didn’t, and that’s probably average for a top pairing.

Klefbom had 2 excellent offensive plays where he could have had the primary assist.

But he’s been harsh on that pairing I think so far this year 🙂

Just IMHO.

Cassandra

On the theory that the Oilers are doing better at limiting scoring chances here is some data:

So far this year 25% of Fenwick events have been scored as HDSC.

For the Oilers that breaks down as:

32% of Fenwick events for have been HDSC.
23% of Fenwick events against have been HDSC.

So what does this mean. The Oilers are, in fact, creating better scoring chances than we’d expect. We’d expect 59 HDSC and they have instead created 78. This is a big difference. Call this the McDavid effect.

The converse, however, is not necessarily true. If the Oilers were league average at keeping shots to the outside they would have allowed 65 HDSC. They have allowed 61, so only 4 less than average, which is within the range of a single fire drill of a shift. There is much less, or no, evidence of a Larsson effect.

Moreover, some of this is going to be score effects, which is going to drive up Fenwick events more than it drives up scoring chances.

Conclusion, this winning streak can be explained almost entirely by three things. 1) McDavid, 2) Talbot, 3) variance in shooting percentage.

Most importantly, the narrative that the winning streak is evidence the Oilers won the trade is demonstrably false.

If you don’t understand why you are successful you won’t continue to be successful.

Cassandra

One of the things stats can do is disprove myths.

For instance it is often repeated that Eakins coached for Corsi, on the one hand, and his teams allowed more scoring chances per shot event because of the swarm.

However, at the time Michael Parkatti was tracking precisely these things. And it turns out that the Oilers of that time did not shoot from farther away than the other team, nor did they allow shots from closer in.

Similarly, David Staples scoring chance data is not reliable. The first problem is defining a scoring chance. No matter what decisions you make here you open yourself up to systematic bias. Worse, it depends upon a judgement on what is a scoring chance which is influence by the result. That is, goals are more likely to be judged a scoring chance than a shot that misses the net, even if the chance itself was exactly the same. Which is to say that judged scoring chances are not independent of goals, but what you are looking for with scoring chance data is to find a measure independent of goals that will predict goals. Using this kind of scoring chance data is selecting on the dependent variable in social science speak.

Woodguy

Zachary DeVine ‏@zakkthebear 16m16 minutes ago
Khaira out on the PK as well. Might be the best forward for the @Condors tonight. Already has a goal on two shots, playing in all situations

Nice to see.

I think he might be better than Slepy and PItlick.

neojanus

Just a couple of observations.

1. Puljujarvi could easily have three or four goals, but his finish is snake-bitten (his own doing). I’d send him down to braise a little bit, but would bring him back after 20 games. His positioning is inconsistent and he fights the puck a little. Saying that, I can see that his defensive awareness is strong, but I think his body is failing his hockey sense. AHL will allow him to come up to speed a little bit better.

2. Defense looks rather good for the most part. Both Russell and Larsson are great at slowing the game down from their zone (usually). I really like how Russell is willing to turn back and start a break out again if it looks poor. Smart hockey. The D coach needs to work on that backhand clearance stuff. Both Nurse and Klefbom aren’t hard enough on that play. The puck doesn’t travel far enough.

3. I’d tinker with the PP option. I don’t like Eberle on that first unit at the moment. Give Nuge a bit of time there?

4. Draisaitl seems frustrated to me. He isn’t quite on his game and is overplaying the puck and forcing too much in the offensive zone. He needs to use his wingers more to finish plays.

5. The trap sucks because it’ll keep Connor from scoring 130, but Vancouver playing that system will give the Oilers a win EVERY time because our speed will always trump the aging one chance antics of the Sedins. Great to see the Oilers win a plodding game. IF (and a big IF) they make the playoffs, it’s good to see that they have the ability to play the patience game.

6. I’m stuck out in Vancouver and the commentary is garbage out here. Garrett and Stonehouse were just brutal, especially Garrett. The bias was brutal.

7. I wish Anderson all my best. Cancer has been an unkind visitor in my family repeatedly and I hope his wife gets the very best care.

Professor Q

I would add that Chychrun has been extremely impressive as well, and a steal by Arizona.

Admiral Ackbar

Woogie63:
Right now Matthews and Laine look head and shoulders above the rest of the draft class.

Agreed. Though Poolparty looked VERY good last night (to my eye). Still had the giveaways and poor neutral zone passes but he’s gaining confidence (quickly) and is looking like the next best kid in that draft.

hunter1909

Woogie63:
Right now Matthews and Laine look head and shoulders above the rest of the draft class.

I wouldn’t want both, if it meant trading away thecaptainconnormcdavid.

Woogie63

Right now Matthews and Laine look head and shoulders above the rest of the draft class.

Admiral Ackbar

Walter Gretzkys Neighbour,

Great post again!

Seems like Russel has 20 legs!

hunter1909

treevojo: But I would be willing to bet there are more posters sitting on their hands right now waiting for the eventual oiler cold streak and hall to continue his offensive dominance.

This is the nature of human beings. Some will always want to stick the boot in, rather than take a chance themselves.

They’re not sitting on their hands. They’re wiping their arses with them.

N64

treevojo: But I would be willing to bet there are more posters sitting on their hands right now waiting for the eventual oiler cold streak and hall to continue his offensive dominance.

If the criteria is playoff games played in Edmonton the next 6 years vs. the last 6 years then Chia wins any trade that does not include 97. Of course the standard is higher than that.

As fans of Hall and the Oilers if we are lucky we land in a place where as many people as possible can believe it was a win win trade. Whatever the evidence ends up saying 😉

hunter1909

Fog of Warts: I looked at the upcoming schedule. If we manage 5-5-2 over the next twelve, I’ll officially downgrade my 85 to “pessimistic”

I’ve decided they’re going to get 115+ points.

Let me check that next 12 games…

H Ottawa – easy win
A Toronto – easy win
A Rangers – competative game
A Islanders – competative game
A Detroit – easy win(this one worries me)
A Pittsburgh – probable Loss
H Dallas – easy win
H Rangers – competative game
A Anaheim – probable split win/loss with next game
A Los Angeles – probable split win/loss with last game
A Dallas – competative game
H Chicago – competative game

My NHL spidey sense, allowing a single point for each competative game gives 15 out of the next 24 points.

Like any fan, I’m usually full of it.

Professor Q

Are there any updates on injured players, and where they might fit in with the current (so as not to mess up the mojo)?

Walter Gretzkys Neighbour

Bruce McCurdy:
One issue I have with advanced stats is that Corsi is based on shot attempts but PDO is based on actual shots.There is a gap in there — what percentage of shot attempts, for & against, wind up being actual shots. From there PDO & its components determine what % of those actual shots go in the nets.

The gap is likely to be more apparent on extreme outliers like Kris Russell who does some of his goalie’s job for him. Not saying Russell is great, just that CF% is not a fair way to rate him.

The advantage of Corsi is that sample size is roughly double that of shots so accumulates to a critical mass quicker. Over a longer term, however, I prefer shots on goal over shot attempts. A big part of the reason is that SF:SA * Sh%:Sv% can be equated to GF:GA. Can’t do that with Corsi.

Last time I checked NHL.com has Corsi & Fenwick equivalents in their Enhanced Stats package but not shots, which to me is a tell that they fundamentally do not understand the concepts.

If I may derail the discussion with an anecdote that this post reminded me of – I grew up in Edmonton (North End – “cover me I’m going for groceries”) watching hockey with my dad who HATED the Canadiens!

In those days before the WHA, you could be a Habs fan or a Leafs fan – or occasionally as I and a few friends were, Bruins fans. (Flyers got popular for a while with their smash mouth stuff!)

Anyway we’d be watching the Saturday night game listening to the announcers and frequently (here’s the connection to this particular post) we’d hear the announcer say “and the shot from the point goes off the leg of the Canadian defender in front of the net”. Very often – particularly if the Leafs were losing – my dad would jump up and scream at the TV “Off a leg, off a leg – how many G#&da#% legs do those bas#$#ds have anyway?”

Fun times! Now back to your regular programming!

Woodguy

I guess another way to say this is:

Goal Share (GF%) is a terrible predictor of future goal share, so goal +/- is not a good stat.

Shot Attempt Share (CF%) is a much better predictor of future goal share so shot attempt +/- is a useful stat.

New Improved Darkness

I looked at the upcoming schedule. If we manage 5-5-2 over the next twelve, I’ll officially downgrade my 85 to “pessimistic”—enter the building through the door marked “Realistic”, then make a hard left at the first foyer (avoiding, if possible, the tantalizing greased-banister rimming the flight down).

N64

Ryan: If save percentage is primarily luck driven, then the Canaiens must have been horribly unlucky after Price was injured last season

Primarily as in the whole population demonstrates more randomness than sustain. Exactly why elite G is soooooo important to identify.

Woodguy

Gret99zky:
Also, do shots off the posts/crossbar count as shots?

No.

If they did they’d have to be a save because its wasn’t a goal, but if the goalie didn’t save it, it can’t be a save.

Woodguy

N64: Some folks hobby is to run a metal detector over fields looking for precious metals.

Woodguy”s is trolling through emerging G and D separating gold from fools gold.

Hehe.

Thanks,

Its the truth.

Woodguy

Admiral Ackbar:
Woodguy,

Brilliant! Thanks. I’ll take a look.

Admiral Ackbar: :

You’re welcome.

Read this too: : https://hockey-graphs.com/2015/10/01/expected-goals-are-a-better-predictor-of-future-scoring-than-corsi-goals/

Hockey-graphs is a very good site and I recommend reading all the posts.

Frank the dog

jm363561: For what it is worth I totally agree. Most people on this blog know at least 10 times more about hockey than me, but when it comes to Corsi I just do not get it. For the 2015.16 season the Oilers Corsi profile was just below league average; identical to Florida’s; and way better than New York’s. Until someone can explain these, and other, mega anomalies I am an HDSC guy. A team’s shooting approach (41 vs 29 feet) arguably dictates its Corsi number more than its ability. With a good goalie who cares how many shots the opposition takes from 40+ feet?

EDIT: Hat tip also to David Staples. He was the first writer who I followed and always liked his focus on chances.

There is a school of thought here that Eakins was/is a better coach than Krieger, based on Corsi,. To my eye, Eakins at one stage thought he could win games by circling the o zone and shooting from afar at oblique angles. Krueger, on the other hand had inferior corsi but much higher sh%. The argument against Krueger is/was that sh% is random and therefore will always regress to the mean, so Krueger was just lucky and Eakins in fact was the better coach. I believe that Krueger was good rather than lucky, and has a strategy that produced a higher sh%.
Edited for phone finger auto correct.

N64

Ryan: If save percentage is primarily luck driven, then the Canaiens must have been horribly unlucky after Price was injured last season.

Why waste cap space on Price or Lundqvist?

Hire two league minimum cap hit AHL goalies and find a lucky penny.

Some folks hobby is to run a metal detector over fields looking for precious metals.

Woodguy”s is trolling through emerging G and D separating gold from fools gold.

Gret99zky

Also, do shots off the posts/crossbar count as shots?

Woodguy

All,

Corsi is not an end within itself.

Its is used because it predicts future goal share better than any metric in small samples.

Its all about goal share, which drives winning.

Teams can run hot (high PDO) while losing the shot attempt share, but its not sustainable.

Newer metrics like xGF and DFF *describe* the actual play better imo, but in small samples they do not predict goal share better than corsi.

This is important.

Now, not all xGF are created equal. @DTMAboutHeart has his own xGF which includes regressed shooting percentage of each player and it actually does predict future GF% better than corsi (for a while)

Read DTMAboutHeart’s stuff here: https://hockey-graphs.com/2015/10/01/expected-goals-are-a-better-predictor-of-future-scoring-than-corsi-goals/

None of the fancystat sites use his xGF metric, so understand that when anyone uses xGF, its not his, but @Mannyelk’s from corsica.hockey so it doesn’t predict future GF% better than corsi.

GMoney’s DFF is very similar to Manny’s xGF and G is going to have 6 years of data when he is done with the database so we will be able to see how well DFF/xGF predicts future GF%, but TODAY WE DO NOT KNOW IF IT IS BETTER THAN CF%.

We do know that with a full season of data DFF predicted GF% better then CF% in a split-half correlation (randomly take half the sample and compare it to the other half), but that was just one season we we can be *positive* that it is true over multiple seasons.

We will run that too.

Here is GMoney’s post on DFF: https://oilersnerdalert.wordpress.com/2016/07/17/a-brief-statistical-look-at-dangerous-fenwick/

To sum up:

Corsi is all about predicting future goal share and its still the best we have in public data bases.

Use score, zone, venue adjusted corsi when possible as it can remove these effects from the data.

I want to run regressions to see how SC% predicts future goal share as well but I need to find someone with the time and inclination to do it.

Material Elvis

Woodguy:
I forgot the 13/14 Avalanche with Jack Adams behind the bench.

CF% 47.69% –25th
PDO 1017.5 – 3rd
SCF% 47.81% – 22nd

PDO took a lower end team and vaulted them to 2nd in the WC.

Roy should have handed that trophy to Varlamov right on the stage.

treevojo

stephen sheps:
oilerman1000,

I don’t think it’s necessary to rip on Hall in order to justify the Oilers current levels of success. The trade happened, he’s a Devil now, contributing their team as best as he can, and is doing quite well considering how much of the team’s scoring he’s responsible for. He made a bad play but he also had to be out on the ice in that critical moment. But really, that’s apropos of nothing. Hall was not solely responsible for the Oilers failures in the past. No need to make him into a scapegoat or a straw man.

The Oilers are winning, but it’s an almost entirely new team compared to the roster 8 games into last season. The past can inform our collective unease about the sustainability of the present but it is also something to let go of as we look forward to what this version of the team can do long term.

I agree with this post.

If I had to guess I would say it was a large majority not in favour of the hall-larsson trade in this community.

There is a vocal minority that seem to like to get their hall digs in at this time.

But I would be willing to bet there are more posters sitting on their hands right now waiting for the eventual oiler cold streak and hall to continue his offensive dominance.

We will see at that point just how level headed this community is.

Hopefully they will follow the blog authors lead and continue with class.

I have my doubts.

Ryan

Georges:
Shafty19,
Lowetide,

Goal differential has the highest correlation to winning or points percentage, and

goal_differential = (shots_for x shooting_%) – (shots_against x save_%)

If you can move one of these four variables in your favour and hold the others constant, you bump up your goal differential and get more wins. (At the team level, the four variables appear uncorrelated, so you have four independent sliders on your NHL GM and coaching control panel.)

It seems to be accepted that the only reliable way to move the variables in your direction is to win the shot battle. Reliable because shot metrics correlate (positively) year to year while the (PDO) percentage metrics don’t. From this, the feeling seems to be that shot metrics measure skill and can be trusted while percentage metrics are more about luck than skill and can’t be trusted.

The PDO is luck argument is laid out here:

http://www.arcticicehockey.com/2011/10/28/2520115/pdo-if-you-were-going-to-understand-just-one-nhl-statistic

Now, in that post, Hawerchuk does say that shooting and save percentage are primarily and not entirely luck driven. If you’re committed to the entirely luck driven argument, then you’re adopting the view that there’s no way to consistently beat the market on percentage stats. You would be willing to give me even odds for any team in any year on whether the team would be better or worse than the median team on the percentage stats. (I laid out my argument against and linked to my analysis a few days back.)

The difference between primarily and entirely luck driven is very, very interesting, because that’s the gap in which you find elite talent. In the past, teams with elite offensive or goaltending talent consistently beat the market on percentage stats.

So, what Woodguy did earlier in the post is correct. If you ever see elite percentage stats, you should ask is there a possibility that you’re dealing with elite talent. If, as in the case of McDavid, the answer is yes, you should expect to consistently have an edge over the typical NHL team on the percentage stats and get more wins.

If save percentage is primarily luck driven, then the Canaiens must have been horribly unlucky after Price was injured last season.

Why waste cap space on Price or Lundqvist?

Hire two league minimum cap hit AHL goalies and find a lucky penny.

N64

Georges: Goal differential has the highest correlation to winning or points percentage, and
goal_differential = (shots_for x shooting_%) – (shots_against x save_%)
If you can move one of these four variables in your favour and hold the others constant, you bump up your goal differential and get more wins. (At the team level, the four variables appear uncorrelated, so you have four independent sliders on your NHL GM and coaching control panel.)
It seems to be accepted that the only reliable way to move the variables in your direction is to win the shot battle.

THIS x 4.

The shot volume sliders seem more sustainable in general.. At a high level danger adjustments improve the correlation to current results (descriptive power) but not as much for future results (predictive power).
.
Tbe elite players that repeatedly tilt the 4 sliders are the signal buried in the noise of the choir of luck.

Woodguy did some great sleuthing to tease out young G and D emerging as slide tilters. Adding in G’s DFF and binning WoodMoney will also be an interesting tool in the hunt for sustain.

Admiral Ackbar

Woodguy,

Brilliant! Thanks. I’ll take a look.

Admiral Ackbar

Bruce McCurdy:
One issue I have with advanced stats is that Corsi is based on shot attempts but PDO is based on actual shots.There is a gap in there — what percentage of shot attempts, for & against, wind up being actual shots. From there PDO & its components determine what % of those actual shots go in the nets.

The gap is likely to be more apparent on extreme outliers like Kris Russell who does some of his goalie’s job for him. Not saying Russell is great, just that CF% is not a fair way to rate him.

The advantage of Corsi is that sample size is roughly double that of shots so accumulates to a critical mass quicker. Over a longer term, however, I prefer shots on goal over shot attempts. A big part of the reason is that SF:SA * Sh%:Sv% can be equated to GF:GA. Can’t do that with Corsi.

Last time I checked NHL.com has Corsi & Fenwick equivalents in their Enhanced Stats package but not shots, which to me is a tell that they fundamentally do not understand the concepts.

I’m still concerned with acceptance of any of these statistics without evidence that they are strong predictors of success. We employ confidence intervals and p-values in medical clinical trials to show the strength of evidence. That could be applied here. I’m surprised that the individuals that have contrived these measures aren’t looking at the strength of their evidence. With powerful evidence comes proof. I figured that’d be sought after.

Woodguy

Admiral Ackbar,

To all the Corgis and fancy stat lovers. Has anyone studied how any of these stats correlate as predictors of success? Something with P-values and confidence intervals would be the proof you’re looking for. With a powerful P value and narrow CI, the value of your fancy stat would be irrefutable. Until this is done, scientifically, many of these fancy stats (while ingenious) still don’t have powerful evidence to support their use as a predictor of success.

Yes.

All of these stats are about predicting future goal share. (GF%)

Was first used because it predicted future GF% better than GF%, especially in small gp samples.

Still does.

xGF/DFF predict GF% better than CF% when using a split half correlations within a season.

When it comes to predicting future GF% with small samples, CF% and xGF/DFF are all very similar.

They are all better than GF%.

The whole basis of corsi analysis is to predict future goal share.

Tore Purdy (RIP) and Vic Ferarri (Time Barnes) did all the work years ago that show this.

Vic’s site is offline since he was hired by the Captials, but Tore’s is still up if you want to read it: http://objectivenhl.blogspot.ca/

For more recent reading google “Eric Tulsky Broad Street Hockey” and a lot of his stuff is still up too even though CAR hired him a couple years ago.

Bruce McCurdy

One issue I have with advanced stats is that Corsi is based on shot attempts but PDO is based on actual shots. There is a gap in there — what percentage of shot attempts, for & against, wind up being actual shots. From there PDO & its components determine what % of those actual shots go in the nets.

The gap is likely to be more apparent on extreme outliers like Kris Russell who does some of his goalie’s job for him. Not saying Russell is great, just that CF% is not a fair way to rate him.

The advantage of Corsi is that sample size is roughly double that of shots so accumulates to a critical mass quicker. Over a longer term, however, I prefer shots on goal over shot attempts. A big part of the reason is that SF:SA * Sh%:Sv% can be equated to GF:GA. Can’t do that with Corsi.

Last time I checked NHL.com has Corsi & Fenwick equivalents in their Enhanced Stats package but not shots, which to me is a tell that they fundamentally do not understand the concepts.

flea

jm363561,

Staples was on the right track, but his stat tracking was objective. Didn’t he decide what constituted a scoring chance?

It is interesting to see everything come full circle (or maybe 270 degrees) where the shot data is now available and some of the smart people here are using that actual data to determine what is a HDSC. Still probably not perfect but a long ways from objective analysis.

If all the players had chips in their jerseys, and there was one in the puck, I think the NHL could easily start producing stats like player and team possession, contested possession, battles won/lost, individual player time with puck, etc. Combine this with Shot data and you’d really start to see who was controlling the game.

lynn

Lowetide,

I always cull the posts on your wonderful site for opinions rather than stats.

There is room for everyone. Your site is a cut above. Thank you.

LadiesloveSmid

Hard to be frustrated by the (decent) corsi numbers when the HDCF% metrics are so strong right now

jm363561

Sorry to add a second rookie observation – looking at the final 2015.16 standings goal difference seems a far better indicator of success than Corsi. Just an accountant’s approach to looking at things.

We can therefore start to plan the Parade!!

Pouzar

Zelepukin: No idea what game Pouzar was watching by saying they didn’t play well.

Z, if you are out there, I was totally trolling…lol….fckin Boxers