The value of Corsi and Fenwick

[Thesensation19*]

Why is Corsi and Fenwick #'s so important?
How can a team use these numbers to improve their teams play.

It seems to me that its overrated to know how many shots in total a player takes. The real important is something I have not come across really but its the difference between the two, the number of shots blocked by an individual.

Maybe I am missing a key element in all this.

Click to read more...

 

[NHLPaul]

Why is Corsi and Fenwick #'s so important?
How can a team use these numbers to improve their teams play.

It seems to me that its overrated to know how many shots in total a player takes. The real important is something I have not come across really but its the difference between the two, the number of shots blocked by an individual.

Maybe I am missing a key element in all this.

Corsi and Fenwick numbers measure possession. If a team has more possession, they are likely generating more scoring chances. It doesn't always correlate in one game or in a handful of games, but the larger the sample size, the more it directly correlates. When you are blocking shots, you are defending. You don't have the puck and arent creating any scoring chances. Its a skill to get in the lanes and block shots, but at the same time you are getting outpossessed when you are forced to block attempts.

 

[Thesensation19*]

Corsi and Fenwick numbers measure possession. If a team has more possession, they are likely generating more scoring chances. It doesn't always correlate in one game or in a handful of games, but the larger the sample size, the more it directly correlates. When you are blocking shots, you are defending. You don't have the puck and arent creating any scoring chances. Its a skill to get in the lanes and block shots, but at the same time you are getting outpossessed when you are forced to block attempts.

Okay. Yes, but what I meant about shots blocked was not how much one player blocks a shot but how many times his shots r blocked

 

[Thesensation19*]

Is there a statistic about players who typically shoot from higher percentage areas as oppose to lower chances.

Like shooting from low angles, far distances would be low chances. Where breakaways, rebounds could be higher...

 

[Hivemind]

Corsi & Fenwick do not measure how many shot attempts an individual player takes. They measure how many shot attempts are taken when a player is on the ice.

 

[NHLPaul]

Okay. Yes, but what I meant about shots blocked was not how much one player blocks a shot but how many times his shots r blocked

ahh i got you now. I dont have the numbers in front of me but I believe you could determine it by finding the difference in a player's iCorsi and iFenwick, considering fenwick doesnt count the blocked shots.

 

[NHLPaul]

Is there a statistic about players who typically shoot from higher percentage areas as oppose to lower chances.

Like shooting from low angles, far distances would be low chances. Where breakaways, rebounds could be higher...

i think http://war-on-ice.com/ is going to calculate the shooting areas for each player. And http://www.progressivehockey.com/ has a shot quality stat but im not sure how they calculated it. Those are two new sites that are helping replace extraskater. Still in the new stages of development so there may be some issues, but I like them a lot

 

[schuckers]

To the original point, the reason that Corsi% and Fenwick% are important is that they are good predictors of future goals. They are better predictors of future goal scoring than just past goals. Fenwick is better than Corsi in this regard.

 

[charlie1]

To the original point, the reason that Corsi% and Fenwick% are important is that they are good predictors of future goals. They are better predictors of future goal scoring than just past goals. Fenwick is better than Corsi in this regard.

Can you provide a link to the study that showed that? I must have missed that one. Thanks.

 

[ARSix]

I don't know what the "study" was, but it's pretty widely accepted that Fenwick close is more predictive than anything else. You still use Corsi for shorter periods or individual analyses in order to get the higher sample size because there will always be more corsis in a game than fenwicks, obviously.

 

[Darth Joker]

Why is Corsi and Fenwick #'s so important?
How can a team use these numbers to improve their teams play.

It seems to me that its overrated to know how many shots in total a player takes. The real important is something I have not come across really but its the difference between the two, the number of shots blocked by an individual.

Maybe I am missing a key element in all this.

I've not researched Corsi and Fenwick as much as many people here have, but I've thought about it a bit, and I think I have a decent grasp on it. So here's my take.

It's not just a question of how many shots a player takes. A player could conceivably take a lot of shots, but also be on the ice for a lot of shots against, which may well speak to that player being a bit weak defensively and/or a cheery-picker.

But if a team is getting a lot more shots on the opposing net than the number of shots that they give up, when Player X is on the ice, than that means that Player X is probably contributing to good puck possession for his team.

To try to wrap your head around Corsi and Fenwick, consider the following:

1. For obvious reasons, players would rather be taking shots than giving up shots.

2. So teams with the skill, strength, and effort needed to outshoot the opposition will typically do so.

3. So if Team A typically outshoots other teams then that probably means that Team A is better than most other teams, at least in some key areas.

4. So teams with good Corsi and Fenwick stats will tend to be better than teams with bad Corsi and Fenwick stats. The same is true for individual players, but context has to be considered here, as it does with just about any individual stat.


Hot goaltending can skew things a bit, but not enough to completely compensate for bad Corsi/Fenwick over the long haul (i.e. a full 82-game schedule).


Shotblocking is a valuable skill, and DMen who are great at it are nice to have for penalty killing and helping to protect a lead. However, when I watch my team put up a high number of shot blocks for an entire game, it usually means they're getting outplayed. Shotblocking is obviously better than letting a lot of shots get through to the net, but if your team is doing a lot of shotblocking, then the opposition is probably carrying the play in that game, and your team is probably going to lose that game unless your goalie puts on a standout performance.

 

[charlie1]

I don't know what the "study" was, but it's pretty widely accepted that Fenwick close is more predictive than anything else. You still use Corsi for shorter periods or individual analyses in order to get the higher sample size because there will always be more corsis in a game than fenwicks, obviously.

A blog post? Anything?

 

[Aavco Cup]

A blog post? Anything?

http://www.arcticicehockey.com/2014/7/17/5901395/underlying-metrics-are-stupid-easy

 

[charlie1]

http://www.arcticicehockey.com/2014/7/17/5901395/underlying-metrics-are-stupid-easy

I was referring to the assertion that Fenwick is a better predictor of Goals than Corsi is. Which doesn't seem to be addressed at all in that link.

 

[The Bob Cole]

A blog post? Anything?

http://objectivenhl.blogspot.ca/2011/02/shots-fenwick-and-corsi.html

Think this goes into it. An old JLikens one.

 

[charlie1]

http://objectivenhl.blogspot.ca/2011/02/shots-fenwick-and-corsi.html

Think this goes into it. An old JLikens one.

Thanks for that.

Just to summarize that article-- Corsi is better than Fenwick at predicting Goals in all of his empirical comparisons. The only case where Fenwick is better than Corsi is when he applies a "theoretical" correction to the correlations which attempts to control for the amount of variability in each predictor. He then states that the observed differences after the theoretical correction may not be statistically significant, and that more data is needed to test it.

To me that is pretty weak evidence that Fenwick is better than Corsi at predicting goals, and until I see a study that actually demonstrates Fenwick's superior predictive ability (rather than just a theoretical extrapolation), I'll stick with Corsi.

 

[The Bob Cole]

Thanks for that.

Just to summarize that article-- Corsi is better than Fenwick at predicting Goals in all of his empirical comparisons. The only case where Fenwick is better than Corsi is when he applies a "theoretical" correction to the correlations which attempts to control for the amount of variability in each predictor. He then states that the observed differences after the theoretical correction may not be statistically significant, and that more data is needed to test it.

To me that is pretty weak evidence that Fenwick is better than Corsi at predicting goals, and until I see a study that actually demonstrates Fenwick's superior predictive ability (rather than just a theoretical extrapolation), I'll stick with Corsi.

I don't think there ever was a 'definitive' one or the other study, was just always preference of the bloggers for the most part. Do you believe blocking shots is a skill or not? In general, I think Corsi was used for shorter samples as including blocked shots helped improve sample sizes.

This is one of the few that actually looked into finding out if blocked shots were a skill or not: http://vhockey.blogspot.ca/2010/05/blocked-shots-luck-or-skill.html

I don't think neither yield significantly improved results over the other, so long as you're accounting for score effects in both.

 

[jbones6287]

I don't think there ever was a 'definitive' one or the other study, was just always preference of the bloggers for the most part. Do you believe blocking shots is a skill or not? In general, I think Corsi was used for shorter samples as including blocked shots helped improve sample sizes.

This is one of the few that actually looked into finding out if blocked shots were a skill or not: http://vhockey.blogspot.ca/2010/05/blocked-shots-luck-or-skill.html

I don't think neither yield significantly improved results over the other, so long as you're accounting for score effects in both.

I was one of the "analytics are stupid" people early on, but am really coming around to it over the last few months.

 

[schuckers]

Some notes and links. The evidence is not as clear as I remembered it to be.

From http://objectivenhl.blogspot.com/2011/02/shots-fenwick-and-corsi.html
"Therefore, while corsi has more predictive validity with respect to goal ratio at the within-season level, fenwick and shot percentage appear to correlate more strongly with goal ratio over a sufficiently large sample of games. In other words, in theory, both fenwick and shot percentage seem to serve as better measures of team quality than corsi does."

From http://vhockey.blogspot.com/2009/12/brief-corroboration-of-new-jersey.html#1019843358686265674
See the comments from Hawerchuk (aka Gabe Desjardins)
"Goal+SOG: R^2 = 0.29 with Win %
Goal+SOG+Miss: R^2 = 0.35
Goal+SOG+Miss+Block: R^2 = 0.29"

From http://nhlnumbers.com/2012/6/26/shot-quality-revisited-a-look-at-the-correlation-between-scoring-chances-and-shot-totals
"There is a strong correlation between shot differential [Fenwick] and scoring chance differential, so the simple shot differential tells us most of what we need to know"

 

[charlie1]

I don't think there ever was a 'definitive' one or the other study, was just always preference of the bloggers for the most part. Do you believe blocking shots is a skill or not? In general, I think Corsi was used for shorter samples as including blocked shots helped improve sample sizes.

This is one of the few that actually looked into finding out if blocked shots were a skill or not: http://vhockey.blogspot.ca/2010/05/blocked-shots-luck-or-skill.html

I don't think neither yield significantly improved results over the other, so long as you're accounting for score effects in both.

Now that was an excellent read. Thank you.

That is pretty convincing to me that blocked shots are a skill rather than purely related to possession. Which is support for the idea that we should rely on Fenwick more so than Corsi when data is abundant. Like you said, and like Sunny Mehta said in that post though, the difference in predictive value will probably be quite small.

One thing I'd like to see is the actual histogram of BSF% (i.e., [Blocked Shots For] divided by [Goals For + Saved Shots For + Missed Shots For + Blocked Shots For]). And I'd like to compare that to the distributions of the other metrics used in that article, just to see the shape of the distribution.

 

[Mubiki]

I always use corsi as a superior means of determining possession, but I use Fenwick to determine who will score and allow goals more efficiently within that possession.

I bet a lot, so Fenwick was very useful in picking between two teams that had relatively even corsi scores.

 

[Flamingo]

I always use corsi as a superior means of determining possession, but I use Fenwick to determine who will score and allow goals more efficiently within that possession.

I bet a lot, so Fenwick was very useful in picking between two teams that had relatively even corsi scores.

So your predictive algorithm is to multiply corsi by Fenwick?

 

[wintersej]

i think http://war-on-ice.com/ is going to calculate the shooting areas for each player. And http://www.progressivehockey.com/ has a shot quality stat but im not sure how they calculated it. Those are two new sites that are helping replace extraskater. Still in the new stages of development so there may be some issues, but I like them a lot

I like progressivehockey.com mostly because his Expected GF% (both in Rel and non-Rel form) makes Bergeron look as impressive as I find him (1st last year, 2nd in 11/12, and best over the past three years combined).

 

[TheHockeyRant]

Corsi and Fenwick numbers measure possession. If a team has more possession, they are likely generating more scoring chances. It doesn't always correlate in one game or in a handful of games, but the larger the sample size, the more it directly correlates. When you are blocking shots, you are defending. You don't have the puck and arent creating any scoring chances. Its a skill to get in the lanes and block shots, but at the same time you are getting outpossessed when you are forced to block attempts.

Corsi and Fenwick do not measure possession. They count shots, and put them into a formula.

Corsi Number = (Shots on net FOR + Shots missed FOR - Shots blocked AGAINST) minus (Shots on net AGAINST + Shots missed AGAINST - Shots blocked FOR)

Fenwick is the same except remove blocked shots from the equation. Anyway that gives you a players Corsi number. Plug that into another equation.

Relative Corsi= Corsi number of player- Corsi number of team when player is not on the ice.

Anyway Corsi/ Fenwick are good stats to predict what teams are better than they appear or underrated etc. But can also be misleading just like plus-minus.
Never buy into a single statistic as your be all end all.
http://thehockeywriters.com/corsi-an-overrated-statistic/

 

[NHLPaul]

Corsi and Fenwick do not measure possession. They count shots, and put them into a formula.

Corsi Number = (Shots on net FOR + Shots missed FOR - Shots blocked AGAINST) minus (Shots on net AGAINST + Shots missed AGAINST - Shots blocked FOR)

Fenwick is the same except remove blocked shots from the equation. Anyway that gives you a players Corsi number. Plug that into another equation.

Relative Corsi= Corsi number of player- Corsi number of team when player is not on the ice.

Anyway Corsi/ Fenwick are good stats to predict what teams are better than they appear or underrated etc. But can also be misleading just like plus-minus.
Never buy into a single statistic as your be all end all.
http://thehockeywriters.com/corsi-an-overrated-statistic/

I know the formula for it and I know what they technically measure, but if one team is throwing a lot more shots at the net, they are owning possession. Corsi and Fenwick are all about possession.