Browser-busting proto-KRACH thread

ursusminor
John,

Isn't it time for the early-season completely meaningless KRACH?

It might be a little too meaningingles at the moment:

jtwcornell91

ursusminor
John,

Isn't it time for the early-season completely meaningless KRACH?

It might be a little too meaningingles at the moment:

You got that right.

Nothing is ever too meaningingles.

ugarte
Nothing is ever too meaningingles.

But something can be too horizontal.

Thanks, John.

John,

Looking at the group of 4 teams that played in Alaska this weekend, is there an obvious explanation as to why AA would have a better record than Mi? Mi won the game between them, and Mi lost to an undefeated team while AA beat a winless team. At first glance, I would think that the records should be reversed.


Edit: There is an obvious explanation -- the AA-Mi game was not included.

ursusminor
John,

Looking at the group of 4 teams that played in Alaska this weekend, is there an obvious explanation as to why AA would have a better record than Mi? Mi won the game between them, and Mi lost to an undefeated team while AA beat a winless team. At first glance, I would think that the records should be reversed.


Edit: There is an obvious explanation -- the AA-Mi game was not included.

I'll bet the cron job ran before the score from the late game in Alaska had been posted. Oops.

Now including all of the weekend's results

jtwcornell91
I'll bet the cron job ran before the score from the late game in Alaska had been posted. Oops.

Sounds painful.

Including last night's games, except the Robert Morris/Alaska-Fairbanks game, which doesn't seem to be in the database yet:

Including last night's games (Oct 17):

Interesting. RPI's RRWP went up a lot with the tie against UA_. I guess that makes sense because UA_ was 3-0. Too bad that the 'Tute had to take a retaliation penalty to set up the tying goal.

A little more vertical, as expected:

Are RPI and Union's Friday games vs. SHU and AIC (and perhaps others) missing?

Oops, included the wrong image. Fixed now, thanks!

jtwcornell91
Oops, included the wrong image. Fixed now, thanks!

Some day I am going to have to spend the time to understand more fully the inner workings of KRACH. At this time of the year, it is still possible to guess how things should work out. SHU and AIC have each played one game, losing to RPI and Union, respectively. Since RPI has a better RRWP than Union, I would have expected that SHU's RRWP would be better than AIC's, but they are the same. That surprises me.

ursusminor

jtwcornell91
Oops, included the wrong image. Fixed now, thanks!

Some day I am going to have to spend the time to understand more fully the inner workings of KRACH. At this time of the year, it is still possible to guess how things should work out. SHU and AIC have each played one game, losing to RPI and Union, respectively. Since RPI has a better RRWP than Union, I would have expected that SHU's RRWP would be better than AIC's, but they are the same. That surprises me.

It's because this particular model is absolutist. Since SHU and AIC have each lost to a team in the big group, and not beaten anybody, the results are explained by making their ratings zero compared to anyone in that group. So in the hyothetical round-robin used to construct the RRWP, they have a .000 expected winning percentage against anyone in that group, along with the teams in the groups above it (whose ratings are infinite compared to RPI, Union, et al). Their expected winning percentage against each other (and other teams not in or above the big group) is undefined, so we set it to .500 to calculate the RRWP.

Thanks, That explains it.

The mega-group is now related to everyone else (until the Ivies start playing):

Is that vertical enough for ya? Also, we're number one!

With UMass and Quinnipiac losing last night, we now have the best RRWP in the land.

jtwcornell91
With UMass and Quinnipiac losing last night, we now have the best RRWP in the land.

It sure is nice that we got that OT goal. It's made our whole season.

Jim Hyla

jtwcornell91
With UMass and Quinnipiac losing last night, we now have the best RRWP in the land.

It sure is nice that we got that OT goal. It's made our whole season.

Is today selection Sunday?

Woohoo, we won our first game! ... wait, I already knew that.

P.S. the new forum software does a great job of making this thread title inaccurate.

Do you save these images in any one easy-to-reference place over the course of a season? It would be neat to see an animation of how the KRACH rankings change over the course of a season.

ftyuv
Do you save these images in any one easy-to-reference place over the course of a season? It would be neat to see an animation of how the KRACH rankings change over the course of a season.

an animation would get pretty boring, pretty soon when everyone is interconnected.

KeithK

ftyuv
Do you save these images in any one easy-to-reference place over the course of a season? It would be neat to see an animation of how the KRACH rankings change over the course of a season.

an animation would get pretty boring, pretty soon when everyone is interconnected.

Don't judge ftyuv's fun times.

KeithK

ftyuv
Do you save these images in any one easy-to-reference place over the course of a season? It would be neat to see an animation of how the KRACH rankings change over the course of a season.

an animation would get pretty boring, pretty soon when everyone is interconnected.

But it would still be interesting to see teams move up and down the rankings. Sort of like this:

[www.atm.ox.ac.uk]

(tracking the relative positions of the Oxford college crews over time in the Summer Eights bump races)

ugarte

KeithK

ftyuv
Do you save these images in any one easy-to-reference place over the course of a season? It would be neat to see an animation of how the KRACH rankings change over the course of a season.

an animation would get pretty boring, pretty soon when everyone is interconnected.

Don't judge ftyuv's fun times.

They're all I have.

Robb
But it would still be interesting to see teams move up and down the rankings. Sort of like this:

[www.atm.ox.ac.uk]

(tracking the relative positions of the Oxford college crews over time in the Summer Eights bump races)

If you color-coded the bubbles on a per-team basis, you could get a similar effect with an animation; watch the waves of color move this way and that, pretty!

Robb

KeithK

ftyuv
Do you save these images in any one easy-to-reference place over the course of a season? It would be neat to see an animation of how the KRACH rankings change over the course of a season.

an animation would get pretty boring, pretty soon when everyone is interconnected.

But it would still be interesting to see teams move up and down the rankings. Sort of like this:

[www.atm.ox.ac.uk]

(tracking the relative positions of the Oxford college crews over time in the Summer Eights bump races)

Talk about browser-busting......

Hurry, get the KRACH posted while we're the only undefeated and untied team left in the country! (BSU is 6-0-1, having beaten and tied NMU)

Robb
Hurry, get the KRACH posted while we're the only undefeated and untied team left in the country! (BSU is 6-0-1, having beaten and tied NMU)

Ideally, he doesn't need to rush on this. Let's hope he has at least a couple of weeks to get around to it.

CK

CKinsland

Robb
Hurry, get the KRACH posted while we're the only undefeated and untied team left in the country! (BSU is 6-0-1, having beaten and tied NMU)

Ideally, he doesn't need to rush on this. Let's hope he has at least a couple of weeks to get around to it.

CK

Nah, ideal is 37-0-0 giving him all the time in the world to post it.

Yep, now the diagram is linear (since teams like UConn also lost). I could almost do a regular KRACH except we would be infinite, Harvard would be zero, and Dartmouth would be zero-squared.

I like the rating Cornell gets (and it's actually pictured) currently on CHN:

[www.collegehockeynews.com]

Chris 02
I like the rating Cornell gets (and it's actually pictured) currently on CHN:

[www.collegehockeynews.com]

Note that in the Bayesian KRACH, equivalent to two fictitious games, we're 8th.

Chris 02
I like the rating Cornell gets (and it's actually pictured) currently on CHN:

[www.collegehockeynews.com]

The team with the #1 strength of schedule is... Robert Morris?

Lauren '06

Chris 02
I like the rating Cornell gets (and it's actually pictured) currently on CHN:

[www.collegehockeynews.com]

The team with the #1 strength of schedule is... Robert Morris?

[www.slack.net]

They've played 7-0-1 Bemidji State (twice), 6-1-1 Alaska (Fairbanks), 6-1 Quinnipiac (twice), and 7-3 Ferris State (twice). Their only game against an opponent with a losing record was one game against 4-6 Alaska-Anchorage. So, yeah. Their only CHA games so far have been against BSU, who have the best non-perfect record in the country.

Oh I wasn't doubting the data, just pointing it out.

Chris 02
I like the rating Cornell gets (and it's actually pictured) currently on CHN:

[www.collegehockeynews.com]

I also liked the rating Harvard gets.

Seems odd that losing to 3-0 Cornell and 4-2-3 Colgate would do that to them. I suspect that their real crime is only having beaten 0-3 Dartmouth - as soon as Dartmouth wins a game, Harvard will shoot up the rankings.

How do we not have the worst strength of schedule? We've played the #56, #57, and #58 teams (out of 58!).

Robb

Chris 02
I like the rating Cornell gets (and it's actually pictured) currently on CHN:

[www.collegehockeynews.com]

I also liked the rating Harvard gets.

Seems odd that losing to 3-0 Cornell and 4-2-3 Colgate would do that to them. I suspect that their real crime is only having beaten 0-3 Dartmouth - as soon as Dartmouth wins a game, Harvard will shoot up the rankings.

Yeah, that's all determined by the bubbles in the protokrach diagram. Harvard is infinitely better than Dartmouth and infinitely worse than everyone else. (So if you like Dartmouth's rating is 0 times 0.)

ebilmes
How do we not have the worst strength of schedule? We've played the #56, #57, and #58 teams (out of 58!).

I believe SOS is based on the winning percentages of the entire season's schedule, not just those we've already played. I could be wrong, though.

Beeeej

ebilmes
How do we not have the worst strength of schedule? We've played the #56, #57, and #58 teams (out of 58!).

I believe SOS is based on the winning percentages of the entire season's schedule, not just those we've already played. I could be wrong, though.

You are wrong. It's just the games played. (It's literally calculated as KRACH divided by winning ratio, which happens to equal a weighted average of you're opponents' KRACHes.)

jtwcornell91

Beeeej

ebilmes
How do we not have the worst strength of schedule? We've played the #56, #57, and #58 teams (out of 58!).

I believe SOS is based on the winning percentages of the entire season's schedule, not just those we've already played. I could be wrong, though.

You are wrong. It's just the games played. (It's literally calculated as KRACH divided by winning ratio, which happens to equal a weighted average of you're opponents' KRACHes.)

Ok - you gotta 'splain that to me, then. Our opponents' KRACHes are 0, 0, and 4.5, and our scheduled strength is 98.9. What kind of a weighted average is that? Yale's opponents to date (RPI, Union, and Princeton) have KRACHes of 157.6, 71.6, and 42.3, respectively, and Yale's SOS is listed as 77.9. I could buy that Yale's SOS is a weighted average of those 3 numbers, but what sort of weightings of 0, 0, and 4.5 gives 98.9?

The FAQ at CHN doesn't help, either, because it just says that the weighting factor to use for SOS is the number of times you've played that opponent divided by the sum of your ratings. Therefore, the weighting factor for each of our three opponents should be 1/(x+infinity) = zero, right? So then our SOS is 0/0? But if you do it the other way (KRACH divided by winning ratio), then our SOS would be infinity divided by infinity - we'd need to know something about the limits of each of those zeros or infinities. How does it come out to 98.9?

Inquiring minds want to know!

Robb
Inquiring minds want to know!

Well, Cornell is now 48th, so ....

adamw

Robb
Inquiring minds want to know!

Well, Cornell is now 48th, so ....

But in the Bayesian, we're a much more rational #25.