NHL now officially tracking advanced stats

[No More Subban]

These stats are giving me a headache.

 

[Captain Mountain]

François Gagnon already did. As a recognized member of the media, he probably had access to a beta version of this new part of NHL.com.

If François Gagnon understands something, I'm pretty sure a bunch of people do.

I said most, not all. And people like Mirtle from the Globe picked up on the stats years ago. The reason others didn't follow had little to do with the name, more to do with them not understanding them. The label really doesn't mean that much.

 

[That Habs Fan]

USAT Behind

USAT Close

USAT On

USHAT

 

[QuebecPride]

All the site does is confirm what I see with my eyes. The Habs have terrible starts, pick up the pace when their backs are against the wall. In other words, they can play a possession game when they have to, but seem content in getting outplayed for most of the game first. This only reinforces my belief that this is a coaching issue.

Their SAT while ahead is brutal also. They are 25th in league, suggesting the have a hard time at preventing their opponents from picking up the pace when they have the lead. Only Columbus, Buffalo, the Leafs and Calgary are worse. yeesh.

I'd be curious to see how the Corsi ahead was before the Habs started losing their leads lately.

 

[Haburger]

These stats are giving me a headache.

come on!!!havent u heard?science and math are the only way .this isnt the 90's.

 

[lou4gehrig]

come on!!!havent u heard?science and math are the only way .this isnt the 90's.

I'm an engineer by profession and I find these new stats uninteresting. Some guys know how to play winning hockey and some don't. Guys will now float just to keep a high Corsi.

 

[Lebowski]

I'm an engineer by profession and I find these new stats uninteresting. Some guys know how to play winning hockey and some don't. Guys will now float just to keep a high Corsi.

How is floating supposed to help someone's CORSI?

 

[sharks9]

I'm an engineer by profession and I find these new stats uninteresting. Some guys know how to play winning hockey and some don't. Guys will now float just to keep a high Corsi.

If anything, floating would hurt your Corsi.

 

[OnTheRun]

Totally agree. I have no problem with advanced stats. My problem is when people call them puck possession stats. This is not what they are. Saying those are puck possession stats you assume the team attempting the most shots have the puck most often. It might be true 90-95% of the time but there will always be time where it's not true. This is not how you are doing stats. There's only one puck possession stat. It's a stat where you time how long a player/team have the puck in a game.

Some players prefer to keep the puck, cycle and attempt only high % shots/plays. Couple of games ago Sekac and Eller spent like 1 minutes in the offensive zone without attempting a single shot. Good possession number, no so good advanced stats number.

Well it is a possession stat, it's not possession time however.

If you take your Eller/Sekac example, for a whole minute they prevented the other team from registering a possession event, so it work as intended. You just need to keep in mind that the stat don't intend to measure how long you hold to the puck but what you do when you have or don't have the puck.

I'm an engineer by profession and I find these new stats uninteresting. Some guys know how to play winning hockey and some don't. Guys will now float just to keep a high Corsi.

That doesn't make any sense. Plus you're not going to get/keep a "high Corsi" by sheer force of will, it's not that simple.

 

[Captain Mountain]

I'm an engineer by profession and I find these new stats uninteresting. Some guys know how to play winning hockey and some don't. Guys will now float just to keep a high Corsi.

That's an impressive and completely inaccurate leap in logic.

 

[Krautso]

We need coach adjusted stats..compute in the MT variable and our guys would look much better. When you ask a team to play collapsing shot blocking defense and transition rush offense you will have crappy advance stats. These stats favour perimeter teams.

 

[LyricalLyricist]

I'm an engineer by profession and I find these new stats uninteresting. Some guys know how to play winning hockey and some don't. Guys will now float just to keep a high Corsi.

Most advanced stats are uninteresting. I'm also an engineer. I think I know what you meant by that last part but it came out making no sense.

 

[MathMan]

Forget it. All the mainstream journalists will pick up on SAT right away.

I dunno about that: https://twitter.com/TSNBobMcKenzie/status/568897175276670976

But really, I don't care. On the French side TVF ("tirs vers le filet") has been a thing since Olivier Bouchard's seminal blog started.

Is it different with the way some of you guys track it? Might explain some of the discrepency.

It's the same definition. Which is weird.

But here's a funny thing. I copied their team stats in a spreadsheet (what can I say, I'm an obsessive nut) and did some sums. Turns out that their dataset has more shots attempts against than shots attempts for -- 1246 to be exact, and proportionately less for the various substats.

Now on a database of 75k shots that may not seem like a huge discrepancy... except this isn't a data collection problem, it's a data interpretation problem. Every shot for is going to be a shot against for someone else. Regardless of your dataset it really should add up to 0.

I think there might be a bug. Or else some sort of undocumented adjustment factor but that would seem weird given the labeling. I'm not sure if that's what causes the discrepancy with the Habs' stats, but it's hardly the only weird thing in their stats.

I think they might need some time to work out the bugs.

 

[lou4gehrig]

Most advanced stats are uninteresting. I'm also an engineer. I think I know what you meant by that last part but it came out making no sense.

Right it makes no sense people around here are trying to sound smart when they know squat. People know which players win players teams. And there's no stat that can quantify what that means. Some players are cluctch, some win with defense, some win with posession or the occassional brilliant bit of skill. Some win because of consistency and making no mistakes.

So the formula for Corsi is the following:
Corsi Number = (Shots on Target For + Missed Shots For + Blocked Shots Against ) - (Shots on Target Against + Missed Shots Against + Blocked Shots For)

Right, so let's think about how somebody could inflate those numbers? Let's see. How about just floating a bunch of worthless shots at the net from the blue line that don't get blocked. Or how about trying to block every shot in site and getting out of position?
And comparing the STAT/Corsi whatever from player to player compares what exactly. That one player is better than another? DD is better than Plek? Dougie Hamilton is better than PK this season? Ribeiro is better than Toews or Crosby?

I think the NBA is great example. 20-25 years ago the best players in the league like Jordon, Karl Malone, Hakeem, Barkley were all able to dominate the game while drawing all the attention from the other team. Yet they all shot over 50% FG and played good D and were leaders. Nowadays, with stats like PER, you might figure the skill level is at an all-time high in the NBA. In fact, guys are awful shooters now and the skill level is decreasing. Who knows why, but maybe it's that their PER is high so they tell themselves they are good enough and don't need to be efficient.

For example Demarcus Cousins (a young player with a bad attitude and who doesn't make his team a winner) has a very high PER on par with Karl Malone during his peak years. Yet nobody would ever confuse those two as far as who is a winner and who isn't.

Stats like PER and Corsi only compare the current players against current competition. It's not a true measure of asbolute skill or performance or ability. Therefore you could simply be performing above your peers now but still below a skill level that was common 10-15 years ago.

 

[DAChampion]

Forget it. All the mainstream journalists will pick up on SAT right away. Clarity. It's important when you're out there on TV talking about something. You don't want to say 'Corsi' and then have to spend 10 minutes to explain what it is.

Everybody knows what a goal is and what an assist is. Even people who don't follow hockey at all can grasp the idea from a generic knowledge of sports. Everybody knows what a shot is. So when you say "shot attempts", it's clear as day what you're talking about.

It might look like a detail to you, because you're inside that community. But the way you name things is extremely important when your job is to communicate information.

Good post.

 

[DAChampion]

To offer a second opinion on MathMan's accusations,

It's entirely possible that NHL advanced stats version 1 has some bugs. It's not a big deal, it's not a huge failure, and nobody should be fired over it. If MM they'll be fixed soon. These things happen.

 

[MathMan]

It's entirely possible that NHL advanced stats version 1 has some bugs. It's not a big deal, it's not a huge failure, and nobody should be fired over it. If MM they'll be fixed soon. These things happen.

Oh, I'm not debating that! I'm in fact quite glad that the NHL has put this on their site, even if I'm a bit underwhelmed by the UI and the accuracy. I know from painful professional experience that bugs happen, and equally painful personal experience that it's really easy to fudge up the queries when collating these numbers.

I'm pointing this out to indicate that in the meantime, the dataset might not be very reliable -- and that we might not want to consider the NHL's data to be authoritative just yet when it contradicts every other stat collector out there.

 

[TRG]

USAT Behind

USAT Close

USAT On

USHAT

It's really cool though. Changing the names democratizes it. Hurt my head all the time when I read articles with advanced stats and had to google all the time to know what the hell it was. Still don't know what PDO is.

Average Shot Length is a really cool stat IMO.

 

[Brainiac]

Yeah, Mathman, maybe you found a bug in their way of compiling data. Looks like it from what you say. I don't know.

So I'd suggest contacting them. Having about a thousand shots going MIA between the hockey blade and the goalie sure is a problem.

Have you tried running some tests to see where the error might come from? Like, any odd way the number 1246 pops out from other calculations? You often find bugs that way.

 

[MathMan]

Yeah, Mathman, maybe you found a bug in their way of compiling data. Looks like it from what you say. I don't know.

Well, they have more shots against than shots for. I don't think anyone needs a lot of math background to figure out why that's wrong.

Have you tried running some tests to see where the error might come from? Like, any odd way the number 1246 pops out from other calculations? You often find bugs that way.

Unfortunately I don't have the datasets or anything like that. This was just adding a column of numbers together.

 

[Brainiac]

Well, they have more shots against than shots for. I don't think anyone needs a lot of math background to figure out why that's wrong.

From what you say, it really sounds like a little bug somewhere in the calculations. Like missed shots being counted one way, but not the other. I don't know.

But if I were you, I'd definitely contact the NHL regarding that. Of course the shots have to balance out in the end.

 

[Yep]

We need coach adjusted stats..compute in the MT variable and our guys would look much better. When you ask a team to play collapsing shot blocking defense and transition rush offense you will have crappy advance stats. These stats favour perimeter teams.

My feeling as well.

I'd like to have true possession time (not zone time) and scoring chances (not shot location). Stathlete successful o-zone passes/game avg. and o-zone puck retrieval would also be very interesting.

---- Mike Kelly @MikeKellyNHL · 10 janv.
Successful OZone passes/game avg. (even-strength)

1. CHI = 60.9
2. BOS = 59.5
3. MTL = 55.3
4. STL = 53
5. SJ = 46.5

@HockeyAnalytics

---


---Mike Kelly @MikeKellyNHL · 5 févr.
Kane can still be a top power F. Top 6 in Ozone puck retrieval metric are:

E. Kane
Bergeron
Williams
Wheeler
Backes
Lucic

@HockeyAnalytics
---

 

[OnTheRun]

My feeling as well.

I'd like to have true possession time (not zone time) and scoring chances (not shot location). Stathlete successful o-zone passes/game avg. and o-zone puck retrieval would also be very interesting.

Scoring chance need to be representative of what a team can actively try to achieve or prevent, if you have a better idea than basing it on shot location then I'm all ears.

---- Mike Kelly @MikeKellyNHL · 10 janv.
Successful OZone passes/game avg. (even-strength)

1. CHI = 60.9
2. BOS = 59.5
3. MTL = 55.3
4. STL = 53
5. SJ = 46.5

@HockeyAnalytics

---

Considering there is over 321 minutes of difference between the 1st place and the 30th in 5-5 TOI, it's not exactly useful by itself. Successful Ozone passes per 60 minutes or % of successful Ozone passes (5on5) would be more useful.

 

[LaP]

USAT Behind

USAT Close

USAT On

USHAT

The cat in the hat?

 

[Yep]

Scoring chance need to be representative of what a team can actively try to achieve or prevent, if you have a better idea than basing it on shot location then I'm all ears.

From Chris Boyle :

"Everybody understands that there is a difference between Steven Stamkos on a breakaway and a Hal Gill point shot. The primary discussion point centers around the impact it has on large samples and whether it is something that can be influenced by a coaching system."
PART 1 : http://www.sportsnet.ca/hockey/nhl/introducing-the-shot-quality-project/
PART 2 : http://www.sportsnet.ca/hockey/introducing-the-shot-quality-project-part-ii/

2 on 1 and breakaways could be interesting too.

Considering there is over 321 minutes of difference between the 1st place and the 30th in 5-5 TOI, it's not exactly useful by itself. Successful Ozone passes per 60 minutes or % of successful Ozone passes (5on5) would be more useful.

I find the per game interesting nonetheless as it hints toward a Hab offense that might not be as putrid as one would think looking at total shots taken.

Anyway, successful ozone pass per 60 minutes (5v5) is not hard to calculate (let's pretend the Stathlete averages are the same at this point of the season) :

1. CHI = 60.9*60/48.28= 75.8
2. BOS = 59.5*60/49.23= 72.5
3. MTL = 55.3*60/47.43= 70.0
4. STL = 53*60/48.04= 66.2
5. SJ = 46.5*60/48.85= 57.1