PDA

View Full Version : All-Time KRACH Ratings



dxmnkd316
03-24-2016, 09:17 PM
So I've been working for a while trying to run a KRACH calc across every game played for all teams past and present. Finally pulled in the last of the available data for the seasons I started with and cleaned it up the best I could. There are some inconsistencies in the data and a few missing games but overall the impact isn't very great. Maybe a point or two either way for the teams. I'll try to reconcile the data as best I can moving forward.

No weight has been given to site, score, etc. (Though that's my next step.) I excluded all games marked as exhibitions and didn't pull in data pre-1900 (Excel can't handle these dates and the automated data pull didn't catch them) games but there are so few it shouldn't impact the results that much.

Since I've been working on this off and on for the last couple years, I only pulled data in through the 2013-2014 season. I'll need to get that when I have a chance. Anyways, here are the results:

100.0 Minnesota
86.0 North Dakota
80.3 Wisconsin
76.6 Michigan
75.7 Denver
63.1 Boston College
61.5 Michigan State
60.4 Boston University
54.6 Northern Michigan
54.2 Colorado College
52.4 Maine
51.2 Minnesota-Duluth
50.6 St. Cloud State
50.3 Clarkson
49.4 Michigan Tech
47.7 New Hampshire
46.6 Harvard
46.0 Cornell
41.3 Lake Superior
39.8 Notre Dame
38.3 Miami
38.3 Bowling Green
37.2 Minnesota State
36.3 St. Lawrence
36.3 Nebraska-Omaha
35.4 Providence
34.8 Ohio State
34.8 Western Michigan
34.4 Ferris State
33.9 Rensselaer
33.0 Massachusetts-Lowell
32.1 Bemidji State
31.7 Alaska-Anchorage
29.9 Colgate
29.7 Vermont
29.2 Yale
28.7 Alaska
26.5 Northeastern
26.3 Dartmouth
24.5 Niagara
23.7 Brown
22.7 Union
21.1 Princeton
20.7 Merrimack
19.9 RIT
19.6 Quinnipiac
17.6 Alabama-Huntsville
17.2 Massachusetts
16.5 Wayne State
16.4 Mercyhurst
15.5 Robert Morris
14.6 Air Force
13.6 Penn State
12.6 Army
10.8 Holy Cross
9.8 Findlay
9.8 Canisius
8.8 Connecticut
8.6 Sacred Heart
8.5 Iona
7.8 Bentley
6.3 American International
4.7 Fairfield

Spartanforlife4
03-24-2016, 09:24 PM
What about Kent State, UIC, Ohio, Illinois, etc.?

dxmnkd316
03-24-2016, 09:52 PM
Haven't added them yet. Have their data just haven't updated the formulas.

Shirtless Guy
03-24-2016, 10:51 PM
St Louis?

Biddco
03-24-2016, 11:03 PM
suck it Fairfield!

brassbonanza
03-24-2016, 11:10 PM
Is all-time KRACH aware that North Dakota is going down the tubes though?

Tipsy McStagger
03-24-2016, 11:44 PM
Northern Arizona?

Tipsy McStagger
03-24-2016, 11:46 PM
Is all-time KRACH aware that North Dakota is going down the tubes though?
Going Veruca Salt on the bit.

FreshFish
03-25-2016, 06:54 AM
here are the results:


22.7 Union


are you sure that's not 26.0? ;)

Kepler
03-25-2016, 07:30 AM
Northern Arizona?

God bless the Lumberjacks.

Kepler
03-25-2016, 07:34 AM
St Louis?

God bless the Bilikens.

Some other outliers from the 30s-40s: USC, California, Stanford. More recently (60s-70s): Lake Forest, Penn, Colby, MIT, Williams, Middlebury, Bowdoin, Norwich, Amherst College, Hamilton.

dxmnkd316
03-25-2016, 08:36 AM
So in parsing the data yesterday I noticed a bunch of teams like MIT but wasn't sure if they were actual D1 varsity teams or club teams. It's why I was hesitant to include them.

Does anyone know if they were varsity squads?

I did add up the games that were included. Something like 42,000 games were used in the calcs. There are probably another 10-15k against Canadian teams and club teams.

Edit: the other reason so many teams were excluded was because the original dataset I was working with only went back to the 99-2000 season. As I pulled in more data I had to add more teams.

dxmnkd316
03-25-2016, 08:37 AM
Plus d3.

Would it make sense to include all the teams?

dxmnkd316
03-25-2016, 08:41 AM
The other project I'm working on is all time pair wise. Might be interesting. Not sure though.

Ralph Baer
03-25-2016, 09:08 AM
So in parsing the data yesterday I noticed a bunch of teams like MIT but wasn't sure if they were actual D1 varsity teams or club teams. It's why I was hesitant to include them.

Does anyone know if they were varsity squads?

I did add up the games that were included. Something like 42,000 games were used in the calcs. There are probably another 10-15k against Canadian teams and club teams.

Edit: the other reason so many teams were excluded was because the original dataset I was working with only went back to the 99-2000 season. As I pulled in more data I had to add more teams.
MIT did at one time have a D-III program. The first three years that the ECAC existed from 61-62 to 63-64, there was no division of the ECAC. See http://www.augenblick.org/chha/ecac_stn.html.

Almington
03-25-2016, 09:22 AM
The other project I'm working on is all time pair wise. Might be interesting. Not sure though.

I just don't understand what the results mean, mathematically. Because we are looking over decades, is this showing who has had the most consistant teams but not necessarily the best teams?

Since each season for each team is relatively independent, particularly as time increases, I wonder if running KRACH for each season, than running statistical analysis on that dataset would not be more valid. But my statistics knowledge is rusty and I'm no expert.

As you go back before the NCAA era, definitely pre-WWII, I think that it becomes very difficult to differentiate between a club team and a varsity team.

I would like to see how UW has cratered in the last two seasons or is that a sufficently small part of the data set that it has little overall impact.

dxmnkd316
03-25-2016, 09:33 AM
I wondered the same thing myself. But I thought it was an interesting mathematical exercise. I wanted to see what programs have had the best long term success as measured by KRACH and pair wise.

Regarding seasons, I have a macro in place to automatically progress through each season and record the results. I should have those up this weekend I'm hoping.

dxmnkd316
03-25-2016, 09:36 AM
Oh, and I will say this, it's been a massive pain trying to reconcile school name changes. For instance, Michigan Tech has changed names a couple of times.