Corsi - Are my assumptions and math correct?

[Frank Drebin]

Posted this over on the Habs board but figured I might get some additional insight here.

Corsi = Shot attempts for (CF) vs shot attempts against (CA) (shots, plus missed shots, plus blocked shots), at even strength

A 50% Corsi indicates that the play on the ice is even; an equal amount of time approximately is spent in each teams zone (based on shot attempts)

If you were to outnumber shot attempts for by the following percentages, your corsi would be as shown below: (Corsi is calculated as CF/CF+CA)

5% more CF than CA = 51.2%
10% more = 52.4%
25% more = 55.6%
50% more = 60%

NHL team ratios of CF to actual 5on5 shots are anywhere from 2.18:1 (Pittsburgh) to 2.50:1 (Philly), CF against ratios are in a similar range 2.14:1 for Chicago to 2.56:1 for San Jose. Average ratios are 2.35:1 for both Corsi for and corsi against. Each team on average takes about 23ES shots on goal for the game.

What does this mean over an entire season (all things being equal)? If you were to have a 55.6% Corsi over the entire season, you would attempt 25% more even strength shots than your opponent, which theoretically would result in 10.6% more actual even strength shots on goal.

Both teams on average, are putting together 46 5vs5 shots the positive corsi team would theoretically have 24.1 shots for vs 21.9 shots against.

Over the course of 82 games, that equals out to 180.4 more 5vs5 shots for.

Average 5vs5 save percentage in the league currently sits at .923. Those 180 shots for theoretically should result in 14 more goals for per season.

Breaking it down to a player level, if say a player averages 26 minutes a game (assuming he matches his teams 5vs5 to special teams ratio), a 55.6 Corsi player would account for 6 more GF per year due to possession than a 50% corsi player.

Is my math wrong on this? For all the talk of importance that Corsi is given, it doesn't seem to be a huge factor in the outcome of games.

Click to read more...

 

[Raccoon Jesus]

Is my math wrong on this? For all the talk of importance that Corsi is given, it doesn't seem to be a huge factor in the outcome of games.

Didn't check your math but assuming it's correct, 14 MORE 5v5 goals is pretty significant. But also, don't forget that this possession theoretically also takes offense away from the other team. With the Kings, Sutter preaches possession as a form of defense rather than offense. Let's just pretend the net effect is ONLY a +14--look at the standings and tell me that's not huge, even before special teams/situations are accounted for.

Also breaking that down to a player level to determine offense seems like a complete generality that isn't fair to the stat OR the players and a misuse like the poster below is ranting about.

 

[japhi]

Good question. I've wondered the same.

Corsi seems like such a BS metric - it doesn't adjust for zone starts or quality of competition, not to mention quality of shot. A guy that starts in the o zone a lot with offensive minded players will have a better corsi than a guy that starts in his own end with bottom six guys. A guy that starts ten percent more in the o zone should have ten percent more shots for, and less shots against. Corsi seems to just confirm that players that are used in offensive roles will create more offense for the teams.

Corsi doesn't tell me who the more effective player is. And if the delta between a guy that is a Corsi star and a Corsi bum is a handful of goals then I'm not sure what the point is.

Subban vs Weber is the best example of this. Weber over his career has started in his own end 54 % of the time. Subban 50%. Career Corsi 50.6 for Weber, 52.2 for Subban. And in years where Subban had less than 50 percent o zone starts his Corsi is around 50%, it falls significantly. When he starts 55 percent of the time in the o zone it rises dramatically. It is safe to assume that if Weber started 55 pecent in the o zone he would put up Subban type Corsi numbers. So does Corsi really tell us that Subban is a better possesion player thrn Weber? When deployed the same they have same numbers.

 

[Frank Drebin]

Didn't check your math but assuming it's correct, 14 MORE 5v5 goals is pretty significant. But also, don't forget that this possession theoretically also takes offense away from the other team. With the Kings, Sutter preaches possession as a form of defense rather than offense. Let's just pretend the net effect is ONLY a +14--look at the standings and tell me that's not huge, even before special teams/situations are accounted for.

Also breaking that down to a player level to determine offense seems like a complete generality that isn't fair to the stat OR the players and a misuse like the poster below is ranting about.

That is 14 more ES goals than the opponent, so the stifled offense is taken into consideration. (7 more plus 7 less)

14 more ES goals is a lot. But a 5% corsi advantage is also huge, a 25% advantage in attempted shots. To put it in perspective, with the exception of Minnesota Wild, all playoff teams have a positive Corsi, and only 2 have a Corsi of over 52.5%.

Which means most playoff matchups will have at most a 2.5% Corsi difference, meaning about a 10% difference in shot differential, which over a 25 game playoff season equals out to about 2 ESG net advantage, or negligible difference over a 7 game series.

 

[Raccoon Jesus]

That is why most people look at the value of outshooting/outattempting your opponent as a big-picture/over time effect rather than as a minute, 7-game series effect, where corsi can simply be nullified by a hot goaltender or luck.

 

[Frank Drebin]

That is why most people look at the value of outshooting/outattempting your opponent as a big-picture/over time effect rather than as a minute, 7-game series effect, where corsi can simply be nullified by a hot goaltender or luck.

After ~ 50 games

Corsi Rank - NHL Standings

1 - 14
2 - 18
3 - 5
4 - 25
5 - 16

6th place Corsi is 51.43%
16th place Corsi is 50.03%.

 

[wgknestrick]

Don't ever, ever, ever, ever forget that Corsi is only 1 half of the important equation. #Shots x SH% = goals for. SH% and SV% are just as important even if they require much more data for certainty. If I buy (20) 50/50 tickets vs your (20) powerball tickets, who is more likely to walk away with money?

GF% (or SCF%) are generally more valuable than CF% because it includes more data. There are some corsi gamers out there tricking the system and some that try to only take dangerous shots sacrificing their corsi

Zucker (and many MIN players) are a great example of this. Currently only has 49%CF, but 60% SCF%, and 80% GF.

Conversely, Matt Read is in the top 5 CF% this season (58% CF%), but only has a 49% SCF and 50% GF%

 

[Frank Drebin]

Don't ever, ever, ever, ever forget that Corsi is only 1 half of the important equation. #Shots x SH% = goals for. SH% and SV% are just as important even if they require much more data for certainty. If I buy (20) 50/50 tickets vs your (20) powerball tickets, who is more likely to walk away with money?

GF% (or SCF%) are generally more valuable than CF% because it includes more data. There are some corsi gamers out there tricking the system and some that try to only take dangerous shots sacrificing their corsi

Zucker (and many MIN players) are a great example of this. Currently only has 49%CF, but 60% SCF%, and 80% GF.

Conversely, Matt Read is in the top 5 CF% this season (58% CF%), but only has a 49% SCF and 50% GF%

Uh...dumb question but is GF% essentially goal differential or plus minus?

And SCF is essentially Corsi with scoring chances only vs absolute Corsi?

 

[wgknestrick]

Uh...dumb question but is GF% essentially goal differential or plus minus?

And SCF is essentially Corsi with scoring chances only vs absolute Corsi?

5v5 goals for %. Basically what +- should be. It doesn't penalize PP players by only allowing them to get minuses, while also crediting PKers with only allowing them to get +s. GF% is basically, do you outscore the other team while a certain player is on the ice.

Scoring Chance For as defined from shot location qualifier (basically "home plate" model), not from a arbitrary stat recorder deciding....um yeah that looked like a SC to me. . This is probably the best form of Corsi since it is "qualified" as a shot with a high likelihood of becoming a goal based on the location.

 

[Raccoon Jesus]

After ~ 50 games

Corsi Rank - NHL Standings

1 - 14
2 - 18
3 - 5
4 - 25
5 - 16

6th place Corsi is 51.43%
16th place Corsi is 50.03%.

After 427 games--

1--2 championships
2--2 championships
3--the previous championship.

Also don't see the point in taking the middle third in teams.

 

[Frank Drebin]

delete