The Advanced Stats Thread Episode VI: RIP To Our Databases

[chosen]

Nash is a defensive black hole at even strength.

I would avoid him too.

Has he always been or is this a recent development?

 

[Machinehead]

Has he always been or is this a recent development?

Last year was worse, but Nash has pretty much always been a huge problem as far as shots against at even strength.

The reputation he has for being elite defensively is totally absurd.

He's a great penalty killer, but that's penalty killing. It's different.

 

[Machinehead]

I still wanna know what "analytics" set Don off, because our O-line really does suck.

 

[Pavel Buchnevich]

Surprised? Not surprised?

 

[DanielBrassard]

Not surprised

 

[Irishguy42]

Pretty much what I thought. Players will focus on team-developed statistics/analysis more than plain public numbers.

Familiarity plays a big role.

There is nothing wrong with players not caring what their Corsi is.

 

[Leetch3]

Surprised? Not surprised?

isn't that a good thing for analytics though? if guys were aware of and concerned about their advanced stats couldn't they potentially skew their own numbers by throwing pucks at the net from all over the ice?

 

[Machinehead]

Re: That article

15 yards, by the way.

 

[ManUtdTobbe]

Nash is a defensive black hole at even strength.

I would avoid him too.

But he drives offense, which is also something Hayes needs... Fast for the suppression to balance him out, that's my logic atleast.

 

[ManUtdTobbe]

Also, it's a good thing that players don't care/know about analytics imo, not surprised at all. That's hardly anything negative about analytics, which is probably what you want to make it given your history.

 

[Irishguy42]

This article and the Don La Greca implosion will be the cornerstone for the arguments from anti-analytics people for months.

 

[silverfish]

Kane said he believes some players consciously try to game the numbers.
“I do think some guys know in the back of their head it’s a stat, so they just throw the puck on net,” he said.

This is particularly hilarious to me considering how often we hear in the "cliche" hockey interviews that 'the boys' just need to "get pucks on net" and "go to the dirty areas of the ice"... ie. corsi and xG.

“I look at 5-per-cent chances sometimes,” Seguin said. “Basically, you could just say it’s a little box around the net and you have to be in it to have a 5-per-cent chance. You look to see if you’re plus or minus for 5-per-cent chances for the night. As a centre, I use it more to see what is against me and my line for that night, how we did.”

Except that a 5-per-cent chance is more like a shot from the top of the circles, and that "box around the net" is closer to 10% or higher.

These guys believe in "advanced stats" a lot more than they know. How often do we hear AV talk about "scoring chances"? Scoring chances is just a bad xG model.

These things are disdained around the league because they aren't communicated effectively. This falls on both sides of the table. The "stats guys" and the "box-score" fans. Which is essentially the reason why I post in two threads on HF now. Here, and the football thread. Because it's not worth the fight. Even if, in your opinion, you're coming at it from a reasonable angle. But then again, I'm sure everyone thinks they're being reasonable all the time

 

[Mac n Gs]

This is particularly hilarious to me considering how often we hear in the "cliche" hockey interviews that 'the boys' just need to "get pucks on net" and "go to the dirty areas of the ice"... ie. corsi and xG.



Except that a 5-per-cent chance is more like a shot from the top of the circles, and that "box around the net" is closer to 10% or higher.

These guys believe in "advanced stats" a lot more than they know. How often do we hear AV talk about "scoring chances"? Scoring chances is just a bad xG model.

These things are disdained around the league because they aren't communicated effectively. This falls on both sides of the table. The "stats guys" and the "box-score" fans. Which is essentially the reason why I post in two threads on HF now. Here, and the football thread. Because it's not worth the fight. Even if, in your opinion, you're coming at it from a reasonable angle. But then again, I'm sure everyone thinks they're being reasonable all the time

Bingo.

 

[silverfish]

Bingo.

The easiest thing I can think of is Jim Sullivan sitting AV down and teaching him about xG.

JS: "Hey, AV, I hear you mention scoring chances a lot in interviews and when you talk to the players, etc..."
AV: "Yeah, me and the staff use scoring chances all the time"
JS: "Great, that's awesome. There is this "new" metric floating around called expected goals, it's really similar to scoring chances"
AV: *gum chewing intensifies*
JS: "We know with scoring chances that the way it works, it's 1 when it's a scoring chance, and 0 when it isn't. But, everyone knows that all scoring chances aren't created equally. A tip from the top of the crease should be weighed more than a wrist-shot from in between the circles. Expected goals takes care of that"

No one on planet earth should disagree with this.

 

[Irishguy42]

The easiest thing I can think of is Jim Sullivan sitting AV down and teaching him about xG.

JS: "Hey, AV, I hear you mention scoring chances a lot in interviews and when you talk to the players, etc..."
AV: "Yeah, me and the staff use scoring chances all the time"
JS: "Great, that's awesome. There is this "new" metric floating around called expected goals, it's really similar to scoring chances"
AV: *gum chewing intensifies*
JS: "We know with scoring chances that the way it works, it's 1 when it's a scoring chance, and 0 when it isn't. But, everyone knows that all scoring chances aren't created equally. A tip from the top of the crease should be weighed more than a wrist-shot from in between the circles. Expected goals takes care of that"

No one on planet earth should disagree with this.

And yet...

 

[Leetch3]

This article and the Don La Greca implosion will be the cornerstone for the arguments from anti-analytics people for months.

la greca's tirade was hilarious but it will only be used if you are trying to troll analytics people and get them to lose their minds.

 

[silverfish]

And yet...

Yes. And yet.

All these 'hockey guys' can't wrap their heads around the fact that identifying a shot from in between the circles as +1, and a deflection from the top of the crease as +1, and not differentiating them at all, makes no sense to me.

And yet...

I'm not smart enough to be able to communicate this effectively.

 

[Leetch3]

The easiest thing I can think of is Jim Sullivan sitting AV down and teaching him about xG.

JS: "Hey, AV, I hear you mention scoring chances a lot in interviews and when you talk to the players, etc..."
AV: "Yeah, me and the staff use scoring chances all the time"
JS: "Great, that's awesome. There is this "new" metric floating around called expected goals, it's really similar to scoring chances"
AV: *gum chewing intensifies*
JS: "We know with scoring chances that the way it works, it's 1 when it's a scoring chance, and 0 when it isn't. But, everyone knows that all scoring chances aren't created equally. A tip from the top of the crease should be weighed more than a wrist-shot from in between the circles. Expected goals takes care of that"

No one on planet earth should disagree with this.

I think its more that people don't understand it than they disagree with it...people see all of these charts and graphs with pretty colors but don't fully understand the meaning behind the charts and how they are determined. and most people aren't willing to take the time to learn it like its a statistics class. so for the non-analytics people you see a chart and might not know if it is good or bad...

corsi has become more commonly accepted because people know its shots for vs shots against. at least at high level people understand what they are looking at...

as someone that wants to understand them more but is way too lazy to actually research it, i think the analytic community would be better served focusing on a handful of key stats/charts and trying to make the understanding of those stats common knowledge instead of trying to come up with new stats that will just confuse the non-analytic people even more

 

[Leetch3]

Yes. And yet.

All these 'hockey guys' can't wrap their heads around the fact that identifying a shot from in between the circles as +1, and a deflection from the top of the crease as +1, and not differentiating them at all, makes no sense to me.

And yet...

I'm not smart enough to be able to communicate this effectively.

ironically your sarcastic take might be the best explanation of expected goals i've seen yet LOL

 

[Filthy Dangles]

lol how would anyone be surprised

Why would a player ever be expected to know or care what % of Shots their team directs toward the goal when they are on the ice?

They have to focus on the technical aspects of hockey to improve their game and if they do that it should show in the numbers but they should only be concerned with the former and not the latter.

 

[silverfish]

I think its more that people don't understand it than they disagree with it...people see all of these charts and graphs with pretty colors but don't fully understand the meaning behind the charts and how they are determined. and most people aren't willing to take the time to learn it like its a statistics class. so for the non-analytics people you see a chart and might not know if it is good or bad...

corsi has become more commonly accepted because people know its shots for vs shots against. at least at high level people understand what they are looking at...

as someone that wants to understand them more but is way too lazy to actually research it, i think the analytic community would be better served focusing on a handful of key stats/charts and trying to make the understanding of those stats common knowledge instead of trying to come up with new stats that will just confuse the non-analytic people even more

ironically your sarcastic take might be the best explanation of expected goals i've seen yet LOL

To respond to both of these in one. Essentially, expected goals is scoring chances on steroids. Scoring chances are binary, they either are a scoring chance, or they are not a scoring chance. The general rule of thumb on scoring chances is a shot taken from anywhere within the homeplate area of the ice. Any shot. Any situation. If the shot location is here, then it's a scoring chance. On the rare occasion that I track NWHL games, this is what we use, because the PBP data that we're working with isn't as robust as the NHL's (but getting there!)

There are flaws with this, obviously. The most glaring 'error' being that all scoring chances, all shots within this location, are not created equal.

However, the most underrated part of scoring chances is that they have to be tracked manually. This is someone watching the game, and going: "this is a scoring chance, this is not a scoring chance." This is dreadful for two reasons.

1) It's subjective. A scoring chance in AV's head might not be a scoring chance in another coaches head. Who's right? Who's wrong?

2) And this, to me, is the most glaring thing of all. Scoring chances are quite possibly the most heavily impacted metric in terms of weighing the output against the input. What I mean by this, is the person tracking a game might be inclined to denote a shot as a scoring chance, when it wasn't, simply because it was a goal. Further, the person tracking a game might be inclined to not classify a shot as a scoring chance, when it should've been, because the goalie made a save or made the save look easy. (If this doesn't make sense, let me know and I can go into further clarification).

An expected goals model eliminates the noise on this completely. xG is built off of the raw play-by-play data, so it isn't a manually tracked metric. What xG boils down to is a shots percent chance of becoming a goal based off of the inputs in a model.

At the end of the day, it's xG that will (more) correctly weigh how dangerous a shot actually is. No more 1's and 0's, but anywhere from 0 to 1 and everything in between.

Boils down to this:

 

[Leetch3]

To respond to both of these in one. Essentially, expected goals is scoring chances on steroids. Scoring chances are binary, they either are a scoring chance, or they are not a scoring chance. The general rule of thumb on scoring chances is a shot taken from anywhere within the homeplate area of the ice. Any shot. Any situation. If the shot location is here, then it's a scoring chance. On the rare occasion that I track NWHL games, this is what we use, because the PBP data that we're working with isn't as robust as the NHL's (but getting there!)

There are flaws with this, obviously. The most glaring 'error' being that all scoring chances, all shots within this location, are not created equal.

However, the most underrated part of scoring chances is that they have to be tracked manually. This is someone watching the game, and going: "this is a scoring chance, this is not a scoring chance." This is dreadful for two reasons.

1) It's subjective. A scoring chance in AV's head might not be a scoring chance in another coaches head. Who's right? Who's wrong?

2) And this, to me, is the most glaring thing of all. Scoring chances are quite possibly the most heavily impacted metric in terms of weighing the output against the input. What I mean by this, is the person tracking a game might be inclined to denote a shot as a scoring chance, when it wasn't, simply because it was a goal. Further, the person tracking a game might be inclined to not classify a shot as a scoring chance, when it should've been, because the goalie made a save or made the save look easy. (If this doesn't make sense, let me know and I can go into further clarification).

An expected goals model eliminates the noise on this completely. xG is built off of the raw play-by-play data, so it isn't a manually tracked metric. What xG boils down to is a shots percent chance of becoming a goal based off of the inputs in a model.

At the end of the day, it's xG that will (more) correctly weigh how dangerous a shot actually is. No more 1's and 0's, but anywhere from 0 to 1 and everything in between.

Boils down to this:

 

[BlueshirtBlitz]

I remember an article re: baseball and metrics where some beef-headed KC Royals player basically bullied a Fangraphs writer (i wanna say Eno Sarris, could be wrong...i'll look for a link. Found it! http://www.fangraphs.com/tht/learning-the-language-of-the-clubhouse/) for talking about metrics. If that's happening in 2014-15 in BASEBALL, where the framework of WAR is accepted by every analyst worth their salt (even on ESPN these days!) so there's no way the work these guys are doing should be totally ridiculed, I think there's even less of a chance for it to get hegemonic acceptance in hockey.

 

[Machinehead]

But he drives offense, which is also something Hayes needs... Fast for the suppression to balance him out, that's my logic atleast.

My logic is, trade Hayes for anything before you have to give his garbage ass money.

But I like yours too.

 

[Machinehead]

The easiest thing I can think of is Jim Sullivan sitting AV down and teaching him about xG.

JS: "Hey, AV, I hear you mention scoring chances a lot in interviews and when you talk to the players, etc..."
AV: "Yeah, me and the staff use scoring chances all the time"
JS: "Great, that's awesome. There is this "new" metric floating around called expected goals, it's really similar to scoring chances"
AV: *gum chewing intensifies*
JS: "We know with scoring chances that the way it works, it's 1 when it's a scoring chance, and 0 when it isn't. But, everyone knows that all scoring chances aren't created equally. A tip from the top of the crease should be weighed more than a wrist-shot from in between the circles. Expected goals takes care of that"

No one on planet earth should disagree with this.

The problem is, AV would agree, and then continue to play bad xG players in high leverage situations because AV believes there's some mystical nuance to the game and only he knows it.

The made up nuance is the biggest enemy to analytics. Yes I see that the player is objectively awful, but he "adds something." AV said it himself. We know what Tanner "adds."

There is no "something." This entire game can be quantified. The entire game. Hockey is not special. There's no nuance to hockey that other sports don't have. There's nothing magical going on.