Polls 2019-20

[KenP]

Nice to see the women pick up all first place votes. Go ladies!!

[RichH]

Nice to see the women pick up all first place votes. Go ladies!!

Interesting to look at the SOS numbers differentiated by East vs. West. Cornell has the highest SOS among Eastern schools in the KRACH top-20, but every Western team has a higher SOS than CU, most by a very large margin.

I'm reminded of the former 4-team ACC in lacrosse, where they could really pad the SOS metric by just playing each other so often.

[Jeff Hopkins '82]

Nice to see the women pick up all first place votes. Go ladies!!

Interesting to look at the SOS numbers differentiated by East vs. West. Cornell has the highest SOS among Eastern schools in the KRACH top-20, but every Western team has a higher SOS than CU, most by a very large margin.

I'm reminded of the former 4-team ACC in lacrosse, where they could really pad the SOS metric by just playing each other so often.

If you look at it, the NCHC, WCHA, and Big Ten play each other a lot.  A significant portion of that is driven by geography.  For the same reason, The Hockey East schools tend to play the ECAC a lot and on a limited basis AHA.  So to some extent it is rather "inbred".

[Robb]

Right, but it's not the insularity that creates the disparity in SOS numbers.  If you had two insular groups, their average records would both be .500 (within the group).  Then, if a single game is played where a random team from group A beats a team from group B, then suddenly KRACH will tell you that every team in A is better than every team in B.  There's a little more east-west crossover than that in a hockey season, but the point is that Western teams did well in those crossover games.  That makes the Western group look stronger top-to-bottom than the East.  Want to reverse the trend?  Do better.  Win more.

[ice]

KRACH 2003 vs. KRACH 2020

[Trotsky]

Is KRACH without SOS basically just winning percentage?

[ice]

I used this data:

https://www.collegehockeynews.com/ratings/krach/

[Swampy]

The graphs may be misleading because although Krach & SOS are measured in similar orders of magnitude, their ranges are drastically different. So instead of a stacked bar chart, side-by-side bars with the SOS scale at the right would give a visual impression more reflective of the teams' relative standing. Of course, the visual impression would reflect a tacit linearity in the two scales, but I'm not sure they're linear.

[ice]

I revised the graphs for Swampy.  

I think the graphs show a few things.  One, the ECAC teams had relatively low SOS's in 2003.  Two, ND and Mankato appear to be especially strong right now.  Three, in 2020 the difference between the 6th and 16th teams is a lot smaller now than it was in 2003.  

With regard to Cornell's late period play this year, I think the team is playing to the level of its opponents and they are not trying to win 9-0.  For Matt, he may be feeling out of the game when he doesn't face a lot of shots and isn't pumped for the entire 60 minutes.  I am not part of the team so this is just speculation.  In a close game with a strong opponent, I think they would play very differently.  We will all see on February 29 and then in March.  

The team only has two losses this season.  That's impressive.

[Jeff Hopkins '82]

Right, but it's not the insularity that creates the disparity in SOS numbers.  If you had two insular groups, their average records would both be .500 (within the group).  Then, if a single game is played where a random team from group A beats a team from group B, then suddenly KRACH will tell you that every team in A is better than every team in B.  There's a little more east-west crossover than that in a hockey season, but the point is that Western teams did well in those crossover games.  That makes the Western group look stronger top-to-bottom than the East.  Want to reverse the trend?  Do better.  Win more.

Agreed on all points.  The issue is that the methodology over-emphasizes (IMO) that limited number of cross-over points.

[abmarks]

The graphs may be misleading because although Krach & SOS are measured in similar orders of magnitude, their ranges are drastically different. So instead of a stacked bar chart, side-by-side bars with the SOS scale at the right would give a visual impression more reflective of the teams' relative standing. Of course, the visual impression would reflect a tacit linearity in the two scales, but I'm not sure they're linear.

Agreed. I read the revised graphs and they are still misleading and/or confusing. It's hard to make any obvious visual generations due to the compressed scale of the SOS measure as well as having all schools with the same color.

Would love to see the crotch and SOS on separate graphs, and also perhaps color the bars by the team's league, or one color for each one color for West.

[nshapiro]

The graphs may be misleading because although Krach & SOS are measured in similar orders of magnitude, their ranges are drastically different. So instead of a stacked bar chart, side-by-side bars with the SOS scale at the right would give a visual impression more reflective of the teams' relative standing. Of course, the visual impression would reflect a tacit linearity in the two scales, but I'm not sure they're linear.

Agreed. I read the revised graphs and they are still misleading and/or confusing. It's hard to make any obvious visual generations due to the compressed scale of the SOS measure as well as having all schools with the same color.

Would love to see the crotch and SOS on separate graphs, and also perhaps color the bars by the team's league, or one color for each one color for West.

He said 'crotch'  Heh, heh heh heh

[marty]

The graphs may be misleading because although Krach & SOS are measured in similar orders of magnitude, their ranges are drastically different. So instead of a stacked bar chart, side-by-side bars with the SOS scale at the right would give a visual impression more reflective of the teams' relative standing. Of course, the visual impression would reflect a tacit linearity in the two scales, but I'm not sure they're linear.

Agreed. I read the revised graphs and they are still misleading and/or confusing. It's hard to make any obvious visual generations due to the compressed scale of the SOS measure as well as having all schools with the same color.

Would love to see the crotch and SOS on separate graphs, and also perhaps color the bars by the team's league, or one color for each one color for West.

Halftime show notwithstanding.

[Swampy]

The graphs may be misleading because although Krach & SOS are measured in similar orders of magnitude, their ranges are drastically different. So instead of a stacked bar chart, side-by-side bars with the SOS scale at the right would give a visual impression more reflective of the teams' relative standing. Of course, the visual impression would reflect a tacit linearity in the two scales, but I'm not sure they're linear.

Agreed. I read the revised graphs and they are still misleading and/or confusing. It's hard to make any obvious visual generations due to the compressed scale of the SOS measure as well as having all schools with the same color.

Would love to see the crotch and SOS on separate graphs, and also perhaps color the bars by the team's league, or one color for each one color for West.

I had in mind having a left scale for Krach ranging from zero to 1200 and a scale on the right for SOS ranging from zero to about 230. Alternatively, and perhaps better, just use % of the maximum and normalize both metrics by converting everything to percentages.

[ice]

Alternatively, and perhaps better, just us % of the maximum and normalize both metrics by converting everything to percentages.