Out Hit?!

[DeltaOne81]

NVM

[Josh '99]

[Q]jtwcornell91 Wrote:
Kotyra's the only ECAC ref I would call "good".  J. Murphy is in the second-tier category with Hansen and Dell, refs who are sometimes annoying, but you know where they're coming from, and they don't totally suck.  (The third tier includes D. Murphy and DuPree.)[/q]
I disagree.  Dell is terrible.

[jtwcornell91]

[Q]KeithK Wrote:

 [Q2] But we play whichever team is lowest ranked after the first round.[/q]

Unless Hahvahd runs the table and we get swept in the North Country.  Now go outside, turn around three times, and spit!

[KeithK]

I said "if we finish first".  Honest.  That's what I wrote in the first place.  Please ignore the "Edited' tag - that was to fix a typo.  That's my story and I'm sticking to it.

[jtwcornell91]

[Q]jmh30 Wrote:

I disagree.  Dell is terrible.[/q]

I threw Dell into the second category because he distinguished himself by calling tight games the weekend after Schafer earned the suspension for cursing out Dupree for not calling anything on injury-risking hits.  Basically the league suspended Schafer and then acknowledged his complaint by giving us a ref who would call everything.  So maybe it's just that he's the kind of terrible we needed at that point.

[Trotsky]

I'm going to create the TBRW tables for GF/GP and GA/GP as a ratio of the league means of those statistics.  However, I have a question before proceeding.  Does it have greater statistical integrity and meaning to include the Cornell GA and GF totals in the league totals, or to exclude them (and exclude the Cornell games in the other teams' total, of course)?  Maybe a card-carrying math genius like John Whelan or somebody who uses stats all the time in their work like John Hayes has an opinion/answer?

[billhoward]

[Q]Trotsky Wrote:  I'm going to create the TBRW tables for GF/GP and GA/GP as a ratio of the league means of those statistics.  However, I have a question before proceeding.  Does it have greater statistical integrity and meaning to include the Cornell GA and GF totals in the league totals, or to exclude them (and exclude the Cornell games in the other teams' total, of course)?  Maybe a card-carrying math genius like John Whelan or somebody who uses stats all the time in their work like John Hayes has an opinion/answer?[/q]

You mean if the other teams average 4 goals for and 3.5 against , or 4.5 goals for and 3.0 againtst in games not played against Cornell, and Cornell averages 5 for and 1 against, do you omit Cornell's averages, and then also omit the games the others played against Cornell?

If you're trying to see how the rest of the league compares to Cornell -- you want the snapshot to be from Cornell's point of view -- then I think you have to omit the Cornell-involved games because Cornell can't play itself. You want to see how many goals Brown, Colgate, Harvard, Dartmouth, RPI, etc. average against Brown, Colgate, Harvard, Dartmouth, RPI, etcetera.

OTOH Cornell is only 1/12 of the league so the impact on the averages may be not that great. If we (Cornell) allow 1 goal in the ECAC average and the others average 3 goals against, the average would be the average of (1 team * 1 GA avg) + (11 teams * 3GA avg) or (1+33)/12 = 2.83.  2.83:1 vs. 3.0:1, either way, it's pretty clear.

[DeltaOne81]

From a statistical perspective it doesn't make a different. Using what *should* be the completely valid assumption that Cornell consist of 1/2*x of the average (because Cornell is one half of 1/x of the games), then you can completely calculate one from the other ( x = number of teams that year).

So if you know the average and the Cornell averages, you can calculate the averages w/o Cornell (and averages w/o Cornell & their opponent). And if you give the #s without Cornell, you could calculate the overall. So statistically, none is more 'valid', its just a different perspective.

Now, which is more interesting? Thats up to you, but it seems that taking out Cornell's #s would only be an attempt to make Cornell look only that much better. So I'd be more 'honest' and just do the overall. Or hell, do both!