Time to reconsider Expected Goal % as a major stat?

[JoeSchmoe]

As I understand it, Expected Goals stats don't take into consideration shooting %. 

Watching Tage Thompson, it's clear he's scoring because he's got such a good shot and such good hands down low. It's not hard to imagine why he scores double the goals for every shot taken as some other players. Yet the stat that so many people hold so dearly does not account for this, so Tage could have below average Expected Goals, but still be putting up top stats. Right now it says Tage has only 16.5 expected goals. That is FLAT OUT WRONG.

Despite people that would tell me otherwise, given the facts I honestly think +/- is a better stat than expected goal % as long as there is a suitable sample size. 

[MattPie]

16 minutes ago, JoeSchmoe said:

As I understand it, Expected Goals stats don't take into consideration shooting %. 

Watching Tage Thompson, it's clear he's scoring because he's got such a good shot and such good hands down low. It's not hard to imagine why he scores double the goals for every shot taken as some other players. Yet the stat that so many people hold so dearly does not account for this, so Tage could have below average Expected Goals, but still be putting up top stats. Right now it says Tage has only 16.5 expected goals. That is FLAT OUT WRONG.

Despite people that would tell me otherwise, given the facts I honestly think +/- is a better stat than expected goal % as long as there is a suitable sample size. 

I don't think the answer to use a really flawed stat because the other stat has a flaw.

[mjd1001]

1 hour ago, JoeSchmoe said:

As I understand it, Expected Goals stats don't take into consideration shooting %. 

Watching Tage Thompson, it's clear he's scoring because he's got such a good shot and such good hands down low. It's not hard to imagine why he scores double the goals for every shot taken as some other players. Yet the stat that so many people hold so dearly does not account for this, so Tage could have below average Expected Goals, but still be putting up top stats. Right now it says Tage has only 16.5 expected goals. That is FLAT OUT WRONG.

Despite people that would tell me otherwise, given the facts I honestly think +/- is a better stat than expected goal % as long as there is a suitable sample size. 

+/- is underrated and right now people are attacking because it isn't perfect.  Is it perfect?  No, but it is a very good/useful stat when used in combination with a little common sense.

[matter2003]

1 hour ago, JoeSchmoe said:

As I understand it, Expected Goals stats don't take into consideration shooting %. 

Watching Tage Thompson, it's clear he's scoring because he's got such a good shot and such good hands down low. It's not hard to imagine why he scores double the goals for every shot taken as some other players. Yet the stat that so many people hold so dearly does not account for this, so Tage could have below average Expected Goals, but still be putting up top stats. Right now it says Tage has only 16.5 expected goals. That is FLAT OUT WRONG.

Despite people that would tell me otherwise, given the facts I honestly think +/- is a better stat than expected goal % as long as there is a suitable sample size. 

I think maybe it just shows how much better a player is than a normal player would be in that situation. McDavid is significantly higher also.  As in, what an NHL average player would be expected to score in similar situations.

Edited December 22, 2022 by matter2003

[K-9]

16 minutes ago, mjd1001 said:

+/- is underrated and right now people are attacking because it isn't perfect.  Is it perfect?  No, but it is a very good/useful stat when used in combination with a little common sense.

Somewhere, @pi2000is smiling and nodding his head in approval.

[JoeSchmoe]

1 hour ago, MattPie said:

I don't think the answer to use a really flawed stat because the other stat has a flaw.

The main point I was trying to make is over a large sample size, expected goals is even more flawed than +/-, but is far more respected.

[Brawndo]

From the Sabres Data Engineer 

[Pimlach]

Back when I managed X00 million dollar projects, I learned not to rely on any one metric.   I looked at a lot of data, some was more effective in early stages, some mid stream in the project, and some when we were closer to completed.   

Sometimes you have to blend data from multiple sources.  Sometimes you have to walk around and ask a bunch of people WTF is really going on?   

Bottom line is don't use one single metric, look at everything.   

[Drag0nDan]

7 minutes ago, Pimlach said:

Back when I managed X00 million dollar projects, I learned not to rely on any one metric.   I looked at a lot of data, some was more effective in early stages, some mid stream in the project, and some when we were closer to completed.   

Sometimes you have to blend data from multiple sources.  Sometimes you have to walk around and ask a bunch of people WTF is really going on?   

Bottom line is don't use one single metric, look at everything.   

Different players need different metrics too.  

I don't need an xGF metric to see what my eyes tell me in that tage has a nasty shot, and some dirty dangles.  He can shoot, create space, and deke the goalie.  Thats a star player.  

But if you wanted to evaluate Krebs, or Mittelstadt utilizing analytics.  This is a nice piece of data to have them digest when they re-watch their games.  Your xGF is low because... your corsi is low because... and what pieces of your game you can improve upon to clean it up.  

[Curt]

It’s a measure of how good a player (or team) is at creating (or preventing) scoring chances.  It does not take into account the shooter’s ability to beat a goalie, or a goalie’s ability to stop the shot.  It just measures how good a scoring opportunity is.

And @Pimlach is absolutely correct.  You can find a flaw in any metric.  There is no number that will perfectly tell you how good a player is.  You need to take into account all of the information available when making an evaluation.

[JoeSchmoe]

7 hours ago, Curt said:

And @Pimlach is absolutely correct.  You can find a flaw in any metric.  There is no number that will perfectly tell you how good a player is.

The best metrics for determining how good a player is, are the good old fashioned goals and assists.

The downside is that these don't let people show off how smart they are when they make claims that guys like Rasmus Asplund are actually great players.

[triumph_communes]

The comments from the Sabres data guy at least give me confidence they don’t have a buffoon doing stats like I often see on message boards. 
 

As much as fancy stats try to help they just as easily overlook major concepts. The simpler concepts are often more telling of what’s going on than anything. PDO being a prime example of an oversimplified concept being majorly telling. Honestly, take +/-, remove empty netters and short handers, and correct for minutes/games played and what fancy stat does that give you?  Comparing that with a shot based statistic that ignores rebounds and I think you have a tool chest to tell many stories. 

Edited December 23, 2022 by triumph_communes

[Pimlach]

37 minutes ago, JoeSchmoe said:

The best metrics for determining how good a player is, are the good old fashioned goals and assists.

The downside is that these don't let people show off how smart they are when they make claims that guys like Rasmus Asplund are actually great players.

Great player?   Granato must be missing that. 

[triumph_communes]

38 minutes ago, JoeSchmoe said:

The best metrics for determining how good a player is, are the good old fashioned goals and assists.

The downside is that these don't let people show off how smart they are when they make claims that guys like Rasmus Asplund are actually great players.

For every Ristolainen they are right about there is a Reinhart they are wrong about. Florida sure ain’t looking special for a team full of analytic darlings. 

[Wyldnwoody44]

I simplify it by.... Watching hockey. 

The world has turned sports into prop bets and ridiculous levels of advanced stats. 

I could care less what a Fenwick or corsi are, I have zero desire to learn. But every GDT is loaded with charts full of nonsensical blips.

Better players, better system= better hockey team with some luck and officiating sprinkled in there a bit. 

[LGR4GM]

15 hours ago, Drag0nDan said:

Different players need different metrics too.  

I don't need an xGF metric to see what my eyes tell me in that tage has a nasty shot, and some dirty dangles.  He can shoot, create space, and deke the goalie.  Thats a star player.  

But if you wanted to evaluate Krebs, or Mittelstadt utilizing analytics.  This is a nice piece of data to have them digest when they re-watch their games.  Your xGF is low because... your corsi is low because... and what pieces of your game you can improve upon to clean it up.  

Nhl teams aren't using xgf the way fans are. It gives an average for a shot so you want to shoot and get pucks to places with higher averages. Tage Thompson isn't average so xgf underestimates him but again, it's about getting shots from the higher xgf spots. 

Really xgf is poorly named. Should be more like shot danger% resulting in goals on average. Which is clunky.

What I think it is good for is goalie as it tells you what an average goalie should do. 

[LGR4GM]

18 hours ago, JoeSchmoe said:

As I understand it, Expected Goals stats don't take into consideration shooting %. 

Watching Tage Thompson, it's clear he's scoring because he's got such a good shot and such good hands down low. It's not hard to imagine why he scores double the goals for every shot taken as some other players. Yet the stat that so many people hold so dearly does not account for this, so Tage could have below average Expected Goals, but still be putting up top stats. Right now it says Tage has only 16.5 expected goals. That is FLAT OUT WRONG.

Despite people that would tell me otherwise, given the facts I honestly think +/- is a better stat than expected goal % as long as there is a suitable sample size. 

Tage only has 16.5 xgf because of where he shoots from, not how good a shooter he is. Look at the xgf leaders, they are all stars. Again that's if he was a league average shooter.

+/- is not a better stat because all it tells you is that a player was on the ice when a goal was scored at even or short handed. You like it better because you understand it better and that's not the same. +/- is really only mildly useful when comparing a team to itself and then digging further for why player x has a good or bad one. For example empty net goals give you a +1.

Xgf quantifies how dangerous where you shoot from is. +/- tells me if you're on a bad team 5v5.

[LGR4GM]

6 hours ago, triumph_communes said:

For every Ristolainen they are right about there is a Reinhart they are wrong about. Florida sure ain’t looking special for a team full of analytic darlings. 

Florida fired their coach, coaching still matters. Reinhart had 82 points last year. 

Here let my prove this, Rasmus Dahlin.

[LGR4GM]

6 hours ago, triumph_communes said:

The comments from the Sabres data guy at least give me confidence they don’t have a buffoon doing stats like I often see on message boards. 
 

As much as fancy stats try to help they just as easily overlook major concepts. The simpler concepts are often more telling of what’s going on than anything. PDO being a prime example of an oversimplified concept being majorly telling. Honestly, take +/-, remove empty netters and short handers, and correct for minutes/games played and what fancy stat does that give you?  Comparing that with a shot based statistic that ignores rebounds and I think you have a tool chest to tell many stories. 

You don't because goals are rare events. 

[Curt]

8 hours ago, JoeSchmoe said:

The best metrics for determining how good a player is, are the good old fashioned goals and assists.

The downside is that these don't let people show off how smart they are when they make claims that guys like Rasmus Asplund are actually great players.

Points are a probably the best metric for determining how good a player is at creating offense, but not necessarily how good a player they are overall.

I can do without the condescending nature of your second sentence.  Really turns me off of engaging in a conversation with you at all.

Edited December 23, 2022 by Curt

[JoeSchmoe]

1 hour ago, LGR4GM said:

Look at the xgf leaders, they are all stars. Again that's if he was a league average shooter.

But the reverse is not true. Tage Thompson is at 49% 5 on 5. He's also a +12.

The 49% suggests he's a net negative player, when in fact he can just score from anywhere, and is playing like one of the league's top players.

It's just not that great of a stat.

[mjd1001]

Here is the issue with advanced stats and people who like to post them or post a chart with more blue than red and say "that shows it, he IS a good/bad player":  Advanced stats from my point of view are just another part of the puzzle, just like the old fashioned/simple stats are.  Plus-minus is not perfect, but that doesn't mean it isn't valuable if used the right way.  Corsi is not perfect (although some advanced stats guys try to tell me over and over that it is.)

There are 3 main categories of ways to evaluate a player/team.  1.) eye test (and there are different levels of that, how 'skilled' a player looks or how they play) 2.) Simple stats (old fashioned stats).  3.) Advanced analytics.    NONE of those 3 alone proves anything. You have to used them together with each other and you can use one of them to counter another that may not be true.

Here is the way I look at it.  Wins and Losses are the ultimate team stat.  HOW you get there can be broken down into various pieces.  Simple goals, assists, and plus minus only give you a part of the picture. Advanced stats help to fill in the holes of that picture that goals, assists, and plus minus don't give you. But advance stats are not a replacement for them.   Just as I view analytics as not a replacement for the 'eye test' in evaluation....I also view advanced stats as not a replacement for the simple stats, but something that MUST go along with them.

What drives me nuts is when I give an opinion on a player (and opinion, I don't come in here like the kool-aid man busting down the walls of the forum) and then someone  responds telling me I'm wrong and they show me ONE chart of their favorite 'advanced stat' to prove I'm wrong.  I'm never going to give my opinion on a player, or a team, based on ONLY the eye test of 'how they look' or ONLY one simple stat...and at the same time if someone wants to change my mind they better show me (not 'come at me' but 'show me') a series of advanced stats that go along with the other two things.

I like advanced stats, they give me more to look at and think about than what was available a decade ago. I just don't think you can use them to make a point while ignoring everything else.

Edited December 24, 2022 by mjd1001

[Doohicksie]

My preferred stat is goals scored.  All the rest is statistical *****.

[Weave]

1 hour ago, Doohickie said:

My preferred stat is goals scored.  All the rest is statistical *****.

[inkman]

1 hour ago, Doohickie said:

My preferred stat is goals scored.  All the rest is statistical *****.

Yeah but if homeboy is in the ice for 2X what he scores, he’s not doing you any favors