Advanced Stats

[Ail]

Yeah, because I just told you.

Hahahaha.

Yeah but I think some people in the team toughness thread might disagree. You're biased.

 

[Zamboner]

Yeah, because I just told you.

I had no idea Toews was better than Colton Orr until you told me.

That's hockey stats in general.

SV% and GAA are complimentary, and without the context of the amount of shots a goalie faced can be misleading.

And that's always going to be a problem with hockey stats. There's too many other factors for advanced stats to ever reach baseball. NBA stats is what it should strive for.

 

[Thirty One]

There is, but I've looked into it before and I don't buy it. That explanation and the 10 second rule applies to goals.

Ask yourself this - ten seconds after a defensive zone faceoff, is the puck equally likely to be in any of the 3 zones? If not, the system is flawed and quite considerably in my opinion.

Looks like it to me:

 

[Thirty One]

I had no idea Toews was better than Colton Orr until you told me.

And that's always going to be a problem with hockey stats. There's too many other factors for advanced stats to ever reach baseball. NBA stats is what it should strive for.

I personally think it should strive to be they best that it can be.

 

[Kwayry]

Advanced stats is an attempt to take the human bias out of evaluating performance. But therein lies its flaw.
You can compare 2 players with the same teammates, same opponents and you still would not get a true view of the difference. Unless you assume that the other players on ice are robots and would perform the exact same way over and over again.
How do you quantify the ebbs and flows, the momentum shifts? Player A can be on the ice when his team is dominating, does it mean he helped or it just means he was there when the momentum shifted. The reverse is true, if player B is on the ice when the other team has momentum what does it really say?
They sure are useful, but not as useful as their proponents would like you to think.

 

[Thirty One]

Advanced stats is an attempt to take the human bias out of evaluating performance. But therein lies its flaw.
You can compare 2 players with the same teammates, same opponents and you still would not get a true view of the difference. Unless you assume that the other players on ice are robots and would perform the exact same way over and over again.
How do you quantify the ebbs and flows, the momentum shifts? Player A can be on the ice when his team is dominating, does it mean he helped or it just means he was there when the momentum shifted. The reverse is true, if player B is on the ice when the other team has momentum what does it really say?
They sure are useful, but not as useful as their proponents would like you to think.

Over a large sample those ups and downs/ebbs and flows should even out, and if they don't, that says something about the player you're evaluating, no?

 

[Kwayry]

Over a large sample those ups and downs/ebbs and flows should even out, and if they don't, that says something about the player you're evaluating, no?

Could be, but it's not an exact science. If your margin is 5-10%, what does it say about the stats. Like I said, they are useful, but not as much as people would like you to think.
I see posts about those stats on a game to game basis and those are definitely off.

 

[Thirty One]

Could be, but it's not an exact science. If your margin is 5-10%, what does it say about the stats. Like I said, they are useful, but not as much as people would like you to think.
I see posts about those stats on a game to game basis and those are definitely off.

Who said the margin is 5-10%

Of course game-by-game stats aren't all too telling (low sample size).

OFF FROM WHAT?

 

[Kwayry]

Who said the margin is 5-10%

Of course game-by-game stats aren't all too telling (low sample size).

OFF FROM WHAT?

When you said over a larger sample size, they tend to even out.
Well, the key word there is "tend". And what point are they exactly even?
You always have to assume there is a discrepancy, how much should it be?
Off in terms of true picture. A player can be having a bad day, a headache, a fight with his GF ... whatever.

 

[Thirty One]

when you said over a larger sample size, they tend to even out.
Well, the key word there is "tend". And what point are they exactly even?

85:42

 

[Kwayry]

85:42

I know you rely on them and that's fine, they add another layer to an intelligent discussion on this board. And this board needs more intelligence.

 

[Ail]

A player can be having a bad day, a headache, a fight with his GF ... whatever.

One bad day/game isn't going to skew an entire season of statistical data to the point they become inaccurate, unless you are suggesting that these inconveniences happen to the same player every game and cannot be accounted for?

You don't look at these stats on a per game basis and draw any kind of definitive conclusions based off one game. You compare them to the average data over many games and determine if said game is an anomaly or fits in line with their play historically. If it is an anomaly then you can start speculating about spousal fights, headaches, flat tires, et al.

 

[Fitzy]

People who think problems can be identified by or solutions can be found through statistics exclusively are the reason the economy is the way it is right now.

Statistics can be useful, but they can lead dogmatic followers down a destructive and hubris-laden path.

 

[Thirty One]

People who think problems can be identified by or solutions can be found through statistics exclusively are the reason the economy is the way it is right now.

Statistics can be useful, but they can lead dogmatic followers down a destructive and hubris-laden path.

People who think solutions can be found exclusively through the eye test are the reason the Flames are the way they are now.

I'll let you decide which is worse.

 

[Fitzy]

People who think solutions can be found exclusively through the eye test are the reason the Flames are the way they are now.

I'll let you decide which is worse.

This is the part where I ask if you have any empirical data to back up that assertion.

 

[SnowblindNYR]

OK, now you're just debating what should be discussed on this message board. And you seemed to conclude that neither the past nor the future should be discussed. How very mindful of you.

Well don't judge the trade before you watch and don't jump to conclusions, you can discuss whatever you want.

 

[Thirty One]

This is the part where I ask if you have any empirical data to back up that assertion.

Data is 4 nErdz

 

[Ail]

This is the part where I ask if you have any empirical data to back up that assertion.

Yeah, the Calgary Flames roster on any given game night.

 

[SingnBluesOnBroadway]

"He uses statistics as a drunken man uses lamp posts — for support rather than illumination."
- Andrew Lang

 

[Chaels Arms]

Except for shot location and quality, you can answer most of those questions right now.

This is my main problem with the advanced stats right now. Seems like a lot of the advanced stats writers and proponents have taken the position that shot quality either isn't a factor at all or isn't a factor worth considering. I understand the concept of high sample size and the theory that shot quality will even out over time and while I may buy that theory as applied to teams I've watched too much hockey to believe that's true for individual players as well.

A one timer from the slot isn't the same as a wrist shot from inside the blue line. A shot from Gaborik in his prime isn't the same as a shot from Scott Gomez in the same exact location.

 

[BlueshirtBlitz]

Did I say any of that?

WAR is unreliable due to a few reasons:

1. They're always tweaking the forumula. There's not even one concrete WAR. Different Sabres come up with different WARs.

2. It uses an unreliable fielding metric.

Sabres will tell you that their defensive metrics take three years to stabilize but it doesn't stop them from using it in a single season of WAR. They may have since changed their opinion on that, they do it a lot.

For example, Sabres concluded that Mark Teixeira was a bad fielder his first year for the Yankees. In reality, Teixeira played closer to second base which allowed Cano to also move over and Jeter benefitted greatly from it. Saying Mark Teixeira was a below average fielder was downright stupid. Ironically, the guy that Sabres love to hate (Jeter) finished the season with the best defensive grade of his career despite being older and slower than his younger years.

You can see how unreliable their defensive metrics are just by looking at the crazy fluctuations individual players have from year to year.

I see what you mean, for some reason I thought "and always being changed" meant that they fluctuate wildly year to year for each player. My bad.

However, I still disagree. No matter how different the WAR is, rWAR and fWAR are consistent in that the best players of the game more often than not have similar WARs on each site. In fact, the only MAJOR major difference is that fWAR uses xFIP while bWAR uses RA-9 I believe. The fielding stats are very easy to understand, and you can subtract them if you want. Also, why was it stupid to say Teixera was a bad fielder? If he had to move over, it was very possible his average to below average D was made even worse in a different position.

That's not really wildly inconsistent, IMO. And for stats like Fenwick and Corsi, they can be very helpful if you look at them like you should WAR and other sabre stats. Juan Lagares on the Mets was worth something close to 3 WAR, mostly due to his defensive metrics. In comparison, a guy like John Mitchell might be considered good because he has a high Corsi (or Fenwick, i'm still iffy on which is which sometimes.)

However, by looking at all bodies of knowledge available, a more complete picture is painted. Baseball stats are much more advanced and easy to quantify, so we can see Lagares was a lousy hitter due to his BABIP and wOBA and stuff like that. For hockey it's more complicated, but we can see that, despite his solid advanced stats, Mitchell isn't THAT good by looking other things.

Nobody wants to look at just the stats. But the stats help quantify things that might be difficult to note otherwise (like David Wright still being the best 3B in the game despite his unflashy dominance )

 

[Off Sides]

This is my main problem with the advanced stats right now. Seems like a lot of the advanced stats writers and proponents have taken the position that shot quality either isn't a factor at all or isn't a factor worth considering. I understand the concept of high sample size and the theory that shot quality will even out over time and while I may buy that theory as applied to teams I've watched too much hockey to believe that's true for individual players as well.

A one timer from the slot isn't the same as a wrist shot from inside the blue line. A shot from Gaborik in his prime isn't the same as a shot from Scott Gomez in the same exact location.

A shot from Boyle should not even count towards anything, unless it's to prove the goalies chest protector actually works.



Overall I agree with what you are saying, they'd need a heat chart and some sort of player shooting percentage on top to get a clearer picture.


On another note, puck possession, why can't they just use a stop watch and literally time how much any player/team/group of players possesses the puck? Would that be more accurate than basing the metric off shots on goal? I mean if they are going to watch it anyway closely enough to see every player who is on the ice and if a shot it taken, why not just actually measure what they are trying to conclude? Time with the puck with a stop watch would remove the whole quality of shot thing from the puck possession thing.

 

[Ail]

This is my main problem with the advanced stats right now. Seems like a lot of the advanced stats writers and proponents have taken the position that shot quality either isn't a factor at all or isn't a factor worth considering. I understand the concept of high sample size and the theory that shot quality will even out over time and while I may buy that theory as applied to teams I've watched too much hockey to believe that's true for individual players as well.

A one timer from the slot isn't the same as a wrist shot from inside the blue line. A shot from Gaborik in his prime isn't the same as a shot from Scott Gomez in the same exact location.

Whose stats does shot quality affect? Not talking about points. Quality of the shot is going to affect the goalie and how he performed more than the forward shooting and how he performed. If one goalie is facing 100 cupcakes from Boyle, and another is facing 100 wristers from Sidney Crosby, who is going to end up looking worse? Does that say something about Boyle? Yeah that he doesn't have a great shot. We all know that, and you don't need Corsi to tell you that. Does and can Boyle have a good game despite not scoring? Of course he can, he is a defensive bottom-6 center who specializes in penalty killing and keeping the puck out of his own end. His Corsi will tell you if he is doing that well.

Look at what Corsi is measuring. Puck possession, and what zone the puck was in when the player was on the ice. The kind of shots they are taking is not important to this. All that matters is that they are able to generate a shot which means they were not playing in their own zone.

 

[TheRightWay]

People who think problems can be identified by or solutions can be found through statistics exclusively are the reason the economy is the way it is right now.

Statistics can be useful, but they can lead dogmatic followers down a destructive and hubris-laden path.

And if people made economic decisions solely based off eyeball judgements we'd all be rich, right? Stop talking in absolutes. Nobody is saying you can get all necessary info from statistics. That's not an argument being made by anyone.

 

[Hi ImHFNYR]

Basically, you're still completely wrong.

You are being way plams right now.

So you think that people were using the advanced stats correctly in that thread? Because most of the convo's revolved around misinterpreting and cherry picking data in that last thread. Which would not be correct usage. Again YOU personally may be more than capable of using but that doesn't change the fact that a significant portion of that convo was dedicated to several people going back and forth solely bc of cherry picking a few advanced stats in a poorly thought out manner.