Advanced Stats

[DirtyDeeds]

When it comes to advanced stats then I am not fond of them. Some love them but it is important to understand how they are compiled and what they represent.

 

Lets discuss shots and all the sub categories:

 

Advanced Hockey Statistics

"What questions does it answer?"

There are a number of new statistics that have been developed in the 21st century (i.e. since about 2007) that have helped analysts and fans to better understand the game. What follows is a brief overview of these statistics.

 

 

Corsi and Fenwick

Corsi and Fenwick are similar to the traditional +/- stat, but instead of goals it is based on shots. They attempt to answer the question, "Who is controlling possession of the puck?". Because there aren't possession stats in hockey (yet?) we instead use shots as a proxy, which has been shown to correlate well with puck possession. This can be presented as a +/- stat (i.e. total shots for - total shots against) or as a percentage (i.e. total shots for / (total shots for + total shots against)).

Corsi For (CF) = Shots + Misses + Blocks Against
Corsi Against (CA) = Shots Against + Misses Against + Blocks For
Corsi For % (CF%) = CF / (CF + CA)
Corsi = CF - CA

Fenwick is essentially the same as Corsi, but it excludes blocked shots.

 

 

Expected GF/GA and +/-

While Corsi and Fenwick count shot attempts, they don't account for quality of the shot. Expected +/- considers the shot location, and uses league-wide averages to determine the likelihood of that shot being a goal. It doesn't factor in whether the shot actually resulted in a goal or not. Instead, it's giving a sense of shot quality by looking at the odds of scoring a goal from that location.

A team which has an Actual Goal Differential exceeding its Expected Goal Differential ('axDiff') indicates a team converting or stopping an inordinate amount of good chances compared to league average. This could indicate the team has great shooters, a prolific goalie, or is just getting lucky.

A negative differential would indicate a team is getting more good chances than its opponent, but is not converting or is allowing more than league norms. This could mean bad shooting, bad goaltending, or just being unlucky.

 

 

Scoring Chances and High-Danger Scoring Chances

As originally defined by War On Ice, 'Scoring Chances' indicate shots attempts that are taken from areas of the ice where goals are more likely to be scored. Attempts made from the attacking team's neutral or defensive zones are excluded.

Inside the zone, a shot is assigned a value of 1,2 or 3, depending on where it was from. A rebound shot (defined as any attempt made within 3 seconds of another blocked, missed or saved attempt without a stoppage in play in between) adds a point to this value. A blocked shot decreases the value by 1.

'Scoring Chances' are any shot attempts with a final value of 2 or higher. 'High-Danger Scoring Chances' are any shot attempt with a final value of 3 or higher.

 

 

From  Advanced Hockey Statistics | Hockey-Reference.com

[travel_dude]

Thanks for the laytman's analysis.  Some obvious less special stats include team shooting %.

We are the worst.  We take the most shots and score on a smaller %.

That makes a lot of sense when we shoot the most.

League average shooting and we lead the league in goals for.

 

So, Sutter preaches shot volume and getting to the net.

The first one takes away from the second one since we are always shooting.

You see a chance like last night and it shows how those also get blurred.

Huberdeau going down the pipeline, is at abpout the top of the circle.

Kadri is closer to the boards than the middle.

Huberdeau sense he will not get a clean shot even though he is free and clear to shoot.

He passes to Kadri, who is expecting shot.

I think Kadri sorta got a shot off but he lost any advantage we had, having to adjust.

How do you even measure that.

And it's dozens more of these game after game.

[conundrumed]

I washed my hands of eGF/eGA. I have zero belief in it as a relevant stat. It's applying laws of averages to data that is seldom, if ever, static.

Take the Flames for example. It isn't leading to conclusions that we can can't otherwise plainly see. To me it's a garbage equation. Meaning I think it will always be if you put flawed data into an equation, the result will be either flawed or unreliable.

So the problem becomes, "but we can prove it in example x". But it isn't proving anything. It's using a result to prove an equation. The unfortunate part, is there isn't a blank result to prove/disprove an equation. Knowing the result, seeing it disprove laws of averages, then gets reasoned away by those who want it to work.

So how do you use or move ahead with the equation when you know the result is potentially flawed, and therefore unreliable? What is the point?

[jjgallow]

20 hours ago, DirtyDeeds said:

When it comes to advanced stats then I am not fond of them. Some love them but it is important to understand how they are compiled and what they represent.

 

Lets discuss shots and all the sub categories:

 

Advanced Hockey Statistics

"What questions does it answer?"

There are a number of new statistics that have been developed in the 21st century (i.e. since about 2007) that have helped analysts and fans to better understand the game. What follows is a brief overview of these statistics.

 

 

Corsi and Fenwick

Corsi and Fenwick are similar to the traditional +/- stat, but instead of goals it is based on shots. They attempt to answer the question, "Who is controlling possession of the puck?". Because there aren't possession stats in hockey (yet?) we instead use shots as a proxy, which has been shown to correlate well with puck possession. This can be presented as a +/- stat (i.e. total shots for - total shots against) or as a percentage (i.e. total shots for / (total shots for + total shots against)).

Corsi For (CF) = Shots + Misses + Blocks Against
Corsi Against (CA) = Shots Against + Misses Against + Blocks For
Corsi For % (CF%) = CF / (CF + CA)
Corsi = CF - CA

Fenwick is essentially the same as Corsi, but it excludes blocked shots.

 

 

Expected GF/GA and +/-

While Corsi and Fenwick count shot attempts, they don't account for quality of the shot. Expected +/- considers the shot location, and uses league-wide averages to determine the likelihood of that shot being a goal. It doesn't factor in whether the shot actually resulted in a goal or not. Instead, it's giving a sense of shot quality by looking at the odds of scoring a goal from that location.

A team which has an Actual Goal Differential exceeding its Expected Goal Differential ('axDiff') indicates a team converting or stopping an inordinate amount of good chances compared to league average. This could indicate the team has great shooters, a prolific goalie, or is just getting lucky.

A negative differential would indicate a team is getting more good chances than its opponent, but is not converting or is allowing more than league norms. This could mean bad shooting, bad goaltending, or just being unlucky.

 

 

Scoring Chances and High-Danger Scoring Chances

As originally defined by War On Ice, 'Scoring Chances' indicate shots attempts that are taken from areas of the ice where goals are more likely to be scored. Attempts made from the attacking team's neutral or defensive zones are excluded.

Inside the zone, a shot is assigned a value of 1,2 or 3, depending on where it was from. A rebound shot (defined as any attempt made within 3 seconds of another blocked, missed or saved attempt without a stoppage in play in between) adds a point to this value. A blocked shot decreases the value by 1.

'Scoring Chances' are any shot attempts with a final value of 2 or higher. 'High-Danger Scoring Chances' are any shot attempt with a final value of 3 or higher.

 

 

From  Advanced Hockey Statistics | Hockey-Reference.com

 

There are lots of things advanced stats are good for though,

 

Like convincing ourselves we won the Weegar trade

[robrob74]

1 hour ago, conundrumed said:

I washed my hands of eGF/eGA. I have zero belief in it as a relevant stat. It's applying laws of averages to data that is seldom, if ever, static.

Take the Flames for example. It isn't leading to conclusions that we can can't otherwise plainly see. To me it's a garbage equation. Meaning I think it will always be if you put flawed data into an equation, the result will be either flawed or unreliable.

So the problem becomes, "but we can prove it in example x". But it isn't proving anything. It's using a result to prove an equation. The unfortunate part, is there isn't a blank result to prove/disprove an equation. Knowing the result, seeing it disprove laws of averages, then gets reasoned away by those who want it to work.

So how do you use or move ahead with the equation when you know the result is potentially flawed, and therefore unreliable? What is the point?

I feel that way with the other stats as well. I think with Johnny Hockey, teams were content to let the flames play with the puck because it wasn't getting them more goals, opponents using the wait and counter attack against the Flames. They scored a lot of goals that way, being content to allow the Flames to play the perimeter. 
 

so what, you have the puck more, to me doesn't equate to dominating. I'm talking the average of all the years with that team, and the current rendition I suppose. Not just picking out ones where other teams took them easy to start seasons and it got them going. But once things tighten up, we saw what those Flames teams did with the puck when it mattered most...

[cross16]

Personal suggestion is these really shouldn't' be called advanced stats as there really isn't anything advance about them but rather just simply a different way of organizing the data. Shot metrics, such as what is quoted in the original post, are pretty standard and straightforward. 

 

As many know I am a big of an of analytics because I think they do a better job of identifying the story of the team and give your more info to understand what is happening. They can identify trends which I think provides both potential for better and different better solutions. I've said this time and time again I think the best use of analytics is to use them to mitigate/combat confirmation bias. They do a very good job of this IMO, where the more traditional counting stats don't. 

 

I think the danger in analytics is not their application but rather the conclusions drawn from them. I really think it just comes down to telling a better story of what is happening out there but I think what got lost is the predictability idea. Analytics were sold as a way to predict hockey, which really cannot be done and thus they got a bad rap and got put on a pedestal as the new gold standing when in reality there are just simply a better way to explain what you are likely already seeing. 

 

Also worth pointing out that we, fans, don't have access to same analytics that teams do so a lot of this debate is a bit pointless. 

[conundrumed]

1 hour ago, cross16 said:

Personal suggestion is these really shouldn't' be called advanced stats as there really isn't anything advance about them but rather just simply a different way of organizing the data. Shot metrics, such as what is quoted in the original post, are pretty standard and straightforward. 

 

As many know I am a big of an of analytics because I think they do a better job of identifying the story of the team and give your more info to understand what is happening. They can identify trends which I think provides both potential for better and different better solutions. I've said this time and time again I think the best use of analytics is to use them to mitigate/combat confirmation bias. They do a very good job of this IMO, where the more traditional counting stats don't. 

 

I think the danger in analytics is not their application but rather the conclusions drawn from them. I really think it just comes down to telling a better story of what is happening out there but I think what got lost is the predictability idea. Analytics were sold as a way to predict hockey, which really cannot be done and thus they got a bad rap and got put on a pedestal as the new gold standing when in reality there are just simply a better way to explain what you are likely already seeing. 

 

Also worth pointing out that we, fans, don't have access to same analytics that teams do so a lot of this debate is a bit pointless. 

I can get with that. Analytics are presented as a, "we should be better/worse than" based on a statistical analysis. That's where the wheels come off. I see the value in how individual teams would algorithm their particular play effectively.

[jjgallow]

34 minutes ago, conundrumed said:

I can get with that. Analytics are presented as a, "we should be better/worse than" based on a statistical analysis. That's where the wheels come off. I see the value in how individual teams would algorithm their particular play effectively.

 

I dunno.  The in-house stuff seems like it's not likely to be on twitter.

 

If it's not on twitter, I just can't trust it as fact.

[JTech780]

I have a hard time with analytics and primarily the models that are publicly available. What I don’t love about them is when they are used to say player A is bad player or player B is a good player, based on the analytic model. 
 

The reason I have an issue with that is there is so much that goes into individual success in a team sport, that I am not sure that it can be quantified in a graph.

 

The biggest factor is coaching, Sutter coached teams always have good shot based metrics, and analytically look strong. We can plainly see this year that something isn’t right with this team. Did Huberdeau become a bad player when coming here or was he used differently by Sutter than Brunette.

 

When you look at a team like Tampa, they don’t necessarily show strong shooting metrics, but everyone knows that they are a very skilled and well balanced team, who have had lots of success the last 3 years.

 

I look at a players shooting metrics and how they are deployed and try to take into account the team around them when I am looking outside the “old school” stats. I combine that with the times I have watched that player and what I know of that player to come up with my conclusion.

 

What I do know is that you can’t have a team full of “skilled” players or a team full of “corsi darlings” or a team of “goons”. You need to have a mix and you need to have the right coach, with the right mix at the right time for a team to be successful.