GDT: Carolina @ Edmonton

[What the Faulk]

Great, great stuff here.

Also worth looking at the D only version of this chart (you can switch positions at the top):

* Hainsey has a way better Corsi than I thought, best on the D corps;
* Murphy and Sekera are slightly negative corsi and essentially identical;
* Bellemore is ok, but against tougher competition than I thought;
* Gleason is awful against equal competition;
* Harrison is bad against poor competition;
* Murphy is, uh, a forward.

--hank

Also Bellemore and Sekera have been trending down and Hainsey up.

I don't know why someone "couldn't care less" about Corsi when they're shot attempts. It's not like it's some fancy made up stat. I do hate the name, but the stat has value.

If your hypothesis is

Jordan Staal is incredible. He's the best defensive forward in the game.

How would you prove it?

Worded like below, how can anyone not find value in advanced stats?

On a typical shift, Jordan will start in the defensive zone against the other team's best player, and by the end of his shift the Canes will have outshot the other team. That is a sign of truly outstanding two-way play, even if it's awkward to watch at times.

 

[CaniacSZN]

QualComp is the y-axis (vertical) and tracks the quality of each player's competition -- the relative Corsi of the players who were on the ice against them.

You're right that about half the team has had more difficult competition than Semin. Ruutu actually has a slight edge over Jordan in this measure, but as you can see from the other numbers he has been facing those opponents in more offensively-favorable situations and his results have been worse.

The bolded comment was to say that Jordan's relative Corsi (the color of his bubble) is better than anyone's but Semin's, even in spite of constantly being used in defensive roles. Semin's results are a little better (darker blue) but he is in much more favorable situations on a regular basis.

On a typical shift, Jordan will start in the defensive zone against the other team's best player, and by the end of his shift the Canes will have outshot the other team. That is a sign of truly outstanding two-way play, even if it's awkward to watch at times.

Thanks! My bubble coloring was on Corsi On, had to change it to Corsi Relative.

 

[Vagrant]

Worded like below, how can anyone not find value in advanced stats?

Because Corsi assumes that giving up a shot is always, by default, a bad thing. And that a shot against is always the result of inferior defense. In truth, shots are taken for a variety of reasons. Sometimes players want long range shots on net to get a whistle. Sometimes a shot from a bad angle is the only thing a defense gives up. It's not always a bad play to let your goaltender touch the puck. Additionally, it's not always a positive when you generate a shot for basically the same reasons as above. There are shots that are simply "give up" plays to get line changes entering the offensive zone.

The goal of the game is not necessarily to outshoot your opposition. I know we've all seen boxscores for goalies that would read like the goalie had to stand on his head but in truth it was a pretty quiet night with 35 nondescript shots against. When you boil it down to something as simply as "shot - good", it undermines the complexity of the game. There are PLENTY of times when shooting the puck is the wrong play.

 

[tarheelhockey]

Because Corsi assumes that giving up a shot is always, by default, a bad thing. And that a shot against is always the result of inferior defense. In truth, shots are taken for a variety of reasons. Sometimes players want long range shots on net to get a whistle. Sometimes a shot from a bad angle is the only thing a defense gives up. It's not always a bad play to let your goaltender touch the puck. Additionally, it's not always a positive when you generate a shot for basically the same reasons as above. There are shots that are simply "give up" plays to get line changes entering the offensive zone.

The goal of the game is not necessarily to outshoot your opposition. I know we've all seen boxscores for goalies that would read like the goalie had to stand on his head but in truth it was a pretty quiet night with 35 nondescript shots against. When you boil it down to something as simply as "shot - good", it undermines the complexity of the game. There are PLENTY of times when shooting the puck is the wrong play.

This is absolutely true and something that I have argued for a long time in relation to analysis at the micro level. "LOL, we outshot them and still lost" is a frustrating thing to read.

That said, the use of Corsi is backed by a large body of evidence. Over a large sample (which is what Corsi is designed to produce), shots on goal correlate strongly with zone time. Zone time correlates with winning hockey games. The real clincher is that possession time has been shown to be more predicative of winning games than a team's own record of previous wins; meaning the advanced numbers are superior to the basic numbers in identifying false postitives. The same applies to player performance; you can more easily identify effective and ineffective players by zone time (represented in large-sample team shot counts) than by looking at their "hockey card" stats.

Now, with all of that said, raw Corsi is still not going to give us a complete picture. At a bare minimum we would want to look at relative Corsi to see how a player compares to his linemates -- otherwise there are team-quality biases in play. And it helps tremendously to know his opponents' relative Corsi as well -- thus we ought to cross-reference the player's relative Corsi against his Relative Corsi Quality of Competition. And then you throw in zone starts, to account for his usage, and his TOI to account for his role on the team. The end result is a chart like the one in the link. It's so deeply cross-referenced, with so many layers of anti-bias factors, that it would take a true fluke for it to be particularly inaccurate.

And the proof is ultimately in the pudding, right? The chart says Jordan has been performing extremely well considering his role and matchups, which is true. It says Lindholm is struggling, Semin is sorely missed and Westgarth is nearly useless. It says Malhotra is leaned on heavily in defensive situations, Skinner rarely touches the ice in his own zone, and Eric has been carrying some fairly deadweight linemates lately. All seem to pass the eye test.

 

[CaniacSZN]

This is absolutely true and something that I have argued for a long time in relation to analysis at the micro level. "LOL, we outshot them and still lost" is a frustrating thing to read.

That said, the use of Corsi is backed by a large body of evidence. Over a large sample (which is what Corsi is designed to produce), shots on goal correlate strongly with zone time. Zone time correlates with winning hockey games. The real clincher is that possession time has been shown to be more predicative of winning games than a team's own record of previous wins; meaning the advanced numbers are superior to the basic numbers in identifying false postitives. The same applies to player performance; you can more easily identify effective and ineffective players by zone time (represented in large-sample team shot counts) than by looking at their "hockey card" stats.

Now, with all of that said, raw Corsi is still not going to give us a complete picture. At a bare minimum we would want to look at relative Corsi to see how a player compares to his linemates -- otherwise there are team-quality biases in play. And it helps tremendously to know his opponents' relative Corsi as well -- thus we ought to cross-reference the player's relative Corsi against his Relative Corsi Quality of Competition. And then you throw in zone starts, to account for his usage, and his TOI to account for his role on the team. The end result is a chart like the one in the link. It's so deeply cross-referenced, with so many layers of anti-bias factors, that it would take a true fluke for it to be particularly inaccurate.

And the proof is ultimately in the pudding, right? The chart says Jordan has been performing extremely well considering his role and matchups, which is true. It says Lindholm is struggling, Semin is sorely missed and Westgarth is nearly useless. It says Malhotra is leaned on heavily in defensive situations, Skinner rarely touches the ice in his own zone, and Eric has been carrying some fairly deadweight linemates lately. All seem to pass the eye test.

I was looking at some other players Corsi Relative.. Ryan Suter is -6.30. Is that normal for D to be in the negative? I know some consider him the best defensemen in the game so it was weird seeing him negative.

 

[What the Faulk]

^^ Spot on. Corsi has its flaws, no doubt, but to completely omit it from any consideration is doing yourself a disservice.

Shot quality absolutely factors in, but until that's tracked to the same extent that everything else is, we can only rely on what we have. But I absolutely do want to see heat maps become a (more?) common thing.

EDIT: I see Suter at -0.6.

Well, that direct link didn't work. Anyway, just scroll down. Also worth noting that when most people take Corsi into account, it's 5 on 5 close.