The Advanced Stats Thread Episode III: Return of the Eye Test

[Machinehead]

There's always going to be a team or two that changes drastically midseason. It can happen with any stat. In 2013-14 the Rangers went from a bottom of the league scoring team to an elite scoring team overnight some time in January.

But 27, 28 teams are going to stay relatively level with overall quality of play.

 

[TheTakedown]

Your response is the reason that I believe advanced stats is in its infancy and has yet to evolve into a more meaningful thing.

they had a coaching change. Are you seriously going to ignore that?

 

[Machinehead]

It's not all about stats either.

Part of my prediction was the fact that Columbus has never been a good team since coming into the league.

Pittsburgh has been there and done that.

 

[Hunter Gathers]

.

I guess it was the ridicule of my opinion that Columbus was good, as absurd, made me think his statement was stupid and obnoxious.

We all post our opinions here, but when you stand behind advanced stats and laugh at opinions that differ from these facts, you are sending out the notion that those that disagree with you just don't understand what they are watching.

I understand that advanced stats are a nice tool, but sometimes it seems like some of its proponents treat it like religion in a way that belittles all other religions and non-religious beliefs.

Well said.

 

[Machinehead]

.

I guess it was the ridicule of my opinion that Columbus was good, as absurd, made me think his statement was stupid and obnoxious.

We all post our opinions here, but when you stand behind advanced stats and laugh at opinions that differ from these facts, you are sending out the notion that those that disagree with you just don't understand what they are watching.

I understand that advanced stats are a nice tool, but sometimes it seems like some of its proponents treat it like religion in a way that belittles all other religions and non-religious beliefs.

I wasn't really laughing at people who think Columbus is good.

I was laughing at the notion that we lost to the '77 Habs when we went up to Columbus and got outplayed last month. This was during a stretch where Columbus got outshot in like 10 straight games.

Then a bunch of people started posting "well it's ok cause it's Columbus." It's not ok, and Columbus was not playing well at the time.

 

[Machinehead]

Different topic:

If you want eye-test evidence of why corsi is important in the playoffs, watch this.



Every goal in this game is absolute trash.

 

[sbjnyc]

So two weeks of play by the Blue Jackets holds more weight than a decade of top 5 CF% teams winning the Cup?

I'm assuming by top 5 CF% teams you're referring to the regular season. If so I don't think that's necessarily the right comparison. I'd think CF% in the regular season would be best compared to playoff seed. More than half the NHL teams make the playoffs and the cup is won in the playoffs. So of the teams that won the cup, how did their CF% in the playoffs (a) rank vs the other playoff teams that year and (b) compare to their regular season CF%?

 

[Machinehead]

I'm assuming by top 5 CF% teams you're referring to the regular season. If so I don't think that's necessarily the right comparison. I'd think CF% in the regular season would be best compared to playoff seed. More than half the NHL teams make the playoffs and the cup is won in the playoffs. So of the teams that won the cup, how did their CF% in the playoffs (a) rank vs the other playoff teams that year and (b) compare to their regular season CF%?

CF% is more predictive than results-based, though, and the playoffs are inherently a small sample.

 

[sbjnyc]

CF% is more predictive than results-based, though, and the playoffs are inherently a small sample.

The first comment is kind of hand wavy and the with respect to the second, I don't agree that it's a small sample - "smaller" yes but "small" is not an objective statement.

 

[Machinehead]

The first comment is kind of hand wavy and the with respect to the second, I don't agree that it's a small sample - "smaller" yes but "small" is not an objective statement.

In aggregate it's not, but it's a grouping of 4-7 game samples which are tiny. And evidence shows that CF% round to round can change wildly.

The 82 games before that are a better predictor.

 

[Mac n Gs]

Really liked this tweet from yesterday:



Again, context is needed, but the tweet says it all, it's just counting.

 

[Machinehead]

And you know how everyone likes to talk about "pucks on net."

It's been working for 50 years. It's just that we can track it now.

 

[Mac n Gs]

And you know how everyone likes to talk about "pucks on net."

It's been working for 50 years. It's just that we can track it now.

You're preaching to the choir. I like corsi, even with it's limitations, but I would like to see other trackable info get released.

Something along the lines of factoring in pre-shot movement into some form of xGF/SCF, trackable neutral zone defensive stats, zone entry/exits, etc.

I know it's out there, it's just not as readily available as stuff from corsica and naturalstattrick. Sportlogiq is great when they have a tweetstorm of random microstats they have for each team.

 

[Irishguy42]

I can't wait until we get public data about zone time and stuff like that. You know, info we can get from chip-embedded jerseys and whatnot.

 

[silverfish]

I can't wait until we get public data about zone time and stuff like that. You know, info we can get from chip-embedded jerseys and whatnot.

Possession time has already been proven to be a less effective form of analysis to shot attempts. That's why most analysts aren't saying that shot attempts are a proxy for possession anymore. They're not calling them possession stats. They are their own, better, thing.

 

[Irishguy42]

Possession time has already been proven to be a less effective form of analysis to shot attempts. That's why most analysts aren't saying that shot attempts are a proxy for possession anymore. They're not calling them possession stats. They are their own, better, thing.

I wasn't saying I want it for possession time. I just want it because I think there can be a lot of cool analysis that can be done with it!

 

[TheTakedown]

Different topic:

If you want eye-test evidence of why corsi is important in the playoffs, watch this.



Every goal in this game is absolute trash.

Funny you posted that only to see this tweet below go hand in hand with Corsi:

Really liked this tweet from yesterday:



Again, context is needed, but the tweet says it all, it's just counting.

It's simple, really... the more you get the puck to the net, the better chance you have of scoring, and the less time you're defending that puck.

It's not just about the shots. It's what's happening before those shots are even taken, and you can't take a shot without having the puck.

 

[Raspewtin]

Possession time has already been proven to be a less effective form of analysis to shot attempts. That's why most analysts aren't saying that shot attempts are a proxy for possession anymore. They're not calling them possession stats. They are their own, better, thing.

This.

 

[Hunter Gathers]

Really liked this tweet from yesterday:



Again, context is needed, but the tweet says it all, it's just counting.

I truly do think that half the problem with the advanced stats are the stupid ****ing names (some of which are completely self aggrandizing -- **** off, Glicko). If we got rid of these moronic names for some of them, they'd be 1) a lot easier to understand for the fan; 2) a lot easier to understand for the players; and 3) simply catch on more.

Name your stat what it represents. Stop trying to be cute with the ******** names. It makes you look like a jackass.

 

[Raspewtin]

Name your stat what it represents. Stop trying to be cute with the ******** names. It makes you look like a jackass.

The names are corsi and fenwick literally because of the people that implemented their usage. Jim Corsi and Matt Fenwick. lol it's not that hard to understand.

 

[silverfish]

I've been actively trying to use 'shot attempts' everywhere.

 

[Filthy Dangles]

Why wouldn't Shots instead of just Shot Attempts be better at correlating with goals?

Every goal is both a Shot and a Shot Attempt. Obviously, not every shot or shot attempt is a goal, but a Shots are obviously much more likely to result in goals than Shot Attempts.

Sample size problem?

 

[Filthy Dangles]

That would also be very basic for players to understand and buy into. 'Outshoot the other team.'

Yes, to get a SOG you need to 'attempt a shot' but I am still not sure how that's more valuable than just looking at shots that actually reach the goal.

 

[silverfish]

Why wouldn't Shots instead of just Shot Attempts be better at correlating with goals?

Every goal is both a Shot and a Shot Attempt. Obviously, not every shot or shot attempt is a goal, but a Shots are obviously much more likely to result in goals than Shot Attempts.

Sample size problem?

Don't know. Don't have the time to run it right now. Would imagine shot attempts are used because like you said, more data to work with.

That would also be very basic for players to understand and buy into. 'Outshoot the other team.'

Yes, to get a SOG you need to 'attempt a shot' but I am still not sure how that's more valuable than just looking at shots that actually reach the goal.

It's not that simple tho.

If you tell your players to 'outshoot the other team' they'll be launching shots that aren't quality. You need quantity, quality, suppression, every little thing that leads to an attempt for or an attempt against is just as important as the attempt itself.

 

[Filthy Dangles]

Don't know. Don't have the time to run it right now. Would imagine shot attempts are used because like you said, more data to work with.



It's not that simple tho.

If you tell your players to 'outshoot the other team' they'll be launching shots that aren't quality. You need quantity, quality, suppression, every little thing that leads to an attempt for or an attempt against is just as important as the attempt itself.

Yeah, just like how you can't tell players to 'outattempt the other team'. You need to play solid fundamental hockey and those things should come as a product of it. Obviously when a player is on the ice he's not thinking 'I need to outshoot the other team on this shift'.

The 'outshoot the other team' was moreso meant over a long haul not really a strategy to go into a single game.

Currently, the league average Shots on Goal per game (for a single team) is 30. 30 x 2 is 60

League average goals is 2.69. 2.69 * 2 = 5.38

Thats 60 vs 5.38 events to work with per game. I don't know the average shot attempts per game, though, but on the surface is seems like Shots alone should suffice. Goals is clearly too small of a sample to be predictive. I am obvsiously not a statistician though




EDIT: just realized this is counting special teams and all non 5v5 play so it's actually a smaller sample than listed