Comments on Five Hole Fanatics: On Corsi

Great summary that you've put together here, K...

2010-03-02T00:22:30.564-07:00

Great summary that you've put together here, Kent.

I would liked to have put together a similar article myself but I'm not nearly as good as explaining things.

I think the real reason corsi is controversial (be...

2010-02-26T10:52:26.092-07:00

I think the real reason corsi is controversial (beyond the fact that it's new) is that it doesn't immediately line up with results in the short term. The Flames, for example, were horrible in terms of outshooting to start the year, but the percentages had them winning anyways. People see bad corsi, but a good record and immediately think 'pffft...useless'.

Fair enough, I guess.

Verrry nice post. I kinda started using a variati...

2010-02-26T10:49:16.479-07:00

Verrry nice post.

I kinda started using a variation of corsi, that is I give two numbers, the total shots while a player is on ice and the shot differential.

So instead of writing "Scott Gomez is -33 and Laraque is -68", I say "Scott Gomez is 1330/-38 and Laraque 200/-68". Corsi % is another way of saying that but it shaves away the ice time proxy that is inherent to Corsi's.

The habs, as a team, were on ice for 5821 ES Shots this year and are -396 (jeebus...). Putting those numbers ath the end of the table already takes the reader some way toward Gabe's CorsiREL numbers.

Corsi will remain controversial unless the MSM start using it for what it means. Such is life.

haha, I thought so. Corrected again.

2010-02-26T08:17:35.918-07:00

haha, I thought so. Corrected again.

Whoa whoa Kent hold up! I was making a witty rema...

2010-02-26T08:13:36.966-07:00

Whoa whoa Kent hold up!

I was making a witty remark :-)

r^2 = 0.55 is that stupid-ass lame-ass dumb-ass argument that Jibble and friends keep trying to throw at you regarding Corsi. That was based off Gabe's work earlier this season.

r^2 = 0.40 is correct I think (from JL's work)

haha, crap. I'll fix that now.

2010-02-26T08:10:25.790-07:00

haha, crap. I'll fix that now.

Nice article Kent. But, um, r^2 = 0.55! Seriously...

2010-02-26T07:59:56.832-07:00

Nice article Kent. But, um, r^2 = 0.55!

Seriously, I despise the argument (popularized by JibbleScribbits) that if it doesn't correlate perfectly then fuck it.

Of course these brilliant minds fail to address the fact that current outscoring's correlation to future outscoring is shit fuckin' all, over the time span of one season (so where the fuck's that r^2 now bitches!). And in a completely unexpected twist, most of those who ppf-shaw at Corsi are fans of teams that are winning without outplaying (Colorado I'm looking at you Jibble, Montreal, etc.)

The whole point of this exercise, which you capture nicely here, is to get predictive value. What is sustainable over the course of one season in hockey and what is not.

If the season were five times as long as it is now then we probably wouldn't even be talking about Corsi since outscoring would shed a lot of its noise.

Addendum: since variance is basically noise^2, lower values of r^2 are actually a lot more powerful than they appear at first glance.