College Policy Debate Forums
November 20, 2017, 06:57:27 PM *
Welcome, Guest. Please login or register.
Did you miss your activation email?

Login with username, password and session length
News: IF YOU EXPERIENCE PROBLEMS WITH THE SITE, INCLUDING LOGGING IN, PLEASE LET ME KNOW IMMEDIATELY.  EMAIL ME DIRECTLY OR USE THE CONTACT US LINK AT THE TOP.
 
   Home   Help Search Login Register CEDA caselist Debate Results Council of Tournament Directors Edebate Archive  
Pages: [1]
  Print  
Author Topic: Glicko Debate Ratings  (Read 5324 times)
jregnier
Jr. Member
**
Posts: 94


« on: October 07, 2015, 10:45:05 AM »

Fresh for 2015...

The ratings are still very rough due to the fact that we have virtually the bare minimum of data to run them. There *are* teams that you may find on the ratings and believe that they are much better or much worse than their current standing. Hopefully, those instances will decline as more data becomes available. There is also a FAQ and a series of blog posts on the website which helps to explain how the ratings are calculated and the reasons behind some of the decisions that have been made.

Also, before anybody pulls out their "jump to conclusions" mat, please be aware that the results of neither round robin have yet been included. Those results will be part of the next round of ratings, which will come out after Harvard.

http://collegedebateratings.weebly.com/home/ratings-10072015
Logged
jregnier
Jr. Member
**
Posts: 94


« Reply #1 on: November 04, 2015, 12:45:10 PM »

Updates: http://collegedebateratings.weebly.com/home/debate-ratings-november-2015
Logged
jregnier
Jr. Member
**
Posts: 94


« Reply #2 on: November 23, 2015, 09:39:26 AM »

Updates: http://collegedebateratings.weebly.com/home/final-ratings-for-first-semester-2015
Logged
jregnier
Jr. Member
**
Posts: 94


« Reply #3 on: January 13, 2016, 10:36:58 AM »

Update: http://collegedebateratings.weebly.com/home/post-swing-ratings-2016
Logged
repko
Full Member
***
Posts: 108


« Reply #4 on: January 13, 2016, 02:07:08 PM »

Kirby seems puzzled by this wave of rankings.

... his reaction is attached.


* KirbyIsConfused.jpg (172.19 KB, 640x852 - viewed 639 times.)
Logged
jregnier
Jr. Member
**
Posts: 94


« Reply #5 on: January 14, 2016, 01:08:52 AM »

That was kind of what my face looked like at first as well.  I didn't see TS's fall to 5th coming.  After looking more closely, it was because they took on 2-3 more losses than 3 of the 4 teams ahead of them, and the fourth fell quite a bit but just not quite enough for TS to pass them.
Logged
repko
Full Member
***
Posts: 108


« Reply #6 on: January 14, 2016, 03:09:42 AM »

1. Oh, the post wasn't limited to MSU TS's ranking. On balance, I think there's a great deal to question here.
2. The most constructive thing I can say is not some inside-the-math-bubble disagreements... Instead, it is a simple comment to students:

At this time of year, there are many students that worry about First-Round math, Second-Round security, etc.

... to those students I would say this: I suspect most First-Round and Second-Round voters will not rely heavily upon the Regnier-Glicko system. Thus, there's a lot of potential to read false signals here.

Tone can be difficult to detect in a post. I actually do like some of what Jason is attempting to do here - as I like the broad idea of ELO and the notion of importing certain mathematical tools used in other competitions. That's intriguing. I just happen to think the 2014-16 iterations of Glicko need to reconsider a host of variables. I neither have the time or inclination to get into all of it. And, I am not going to defend the resume of an MSU team (I'm not even sure that's an appropriate thing to do - given that I am a FRALB voter). But, my take is that some of the community "Huh!!!" reactions to these rankings are indicative of something beyond mere instinct. I suspect many mathematical approaches would/will confirm those reactions.

... we'll find out when the FRALB and SRALB rankings are released in April. But, I guess I wanted to say - to students - that Glicko is but one approach.

Best,

  Will


Logged
jregnier
Jr. Member
**
Posts: 94


« Reply #7 on: January 14, 2016, 08:55:03 AM »

To a large extent, I don't disagree.  It's one metric.  There are others - no different than, say, basketball where one can look at PER, RPM, BPM, WinShares, ringzzz, the eye test, etc.  Each one reveals something and perhaps conceals something else.  

I think that a conscientious and well-informed person who sits down to really take into account all the information can "beat" the ratings.  Hopefully, that's what bid voters are.  More importantly, though, the ratings aren't exactly attempting to replicate bid voting, which in my estimation isn't really about picking the "best teams" so much as rewarding debaters for successful seasons.  Nevertheless, fwiw as one data point, the ratings for last year were actually closer to the final aggregate vote for second round bids than the majority of individual voters were [**EDIT: actually among teams that received second round bids, the ratings were closer to the aggregate vote than your own rankings, with a 2.19 average error compared to your 2.81].

Finally, I don't consider the ratings a finished product yet.  While I locked in the methodology at the start of this year, I have already planned some changes for next year that are significant - most notably, a way to better account for strength of opposition at the beginning of the season.  But ultimately, if somebody came along with a system that is better, I would welcome it.  I'm not trained as a mathematician and try to be upfront about the places where the ratings might have weaknesses, but I do think that they fill a need.  An important rejoinder to concerns about the accuracy of these (or any) ratings is: compared to what?
« Last Edit: January 14, 2016, 09:49:43 AM by jregnier » Logged
jregnier
Jr. Member
**
Posts: 94


« Reply #8 on: February 10, 2016, 02:13:44 PM »

Updates: http://collegedebateratings.weebly.com/home/finalish-regular-season-ratings-bid-voting
Logged
jregnier
Jr. Member
**
Posts: 94


« Reply #9 on: February 12, 2016, 04:03:54 AM »

Updated to include Dartmouth RR.  Basically, the effect was as predicted.
Logged
kylevint
Newbie
*
Posts: 1


« Reply #10 on: April 06, 2016, 01:07:38 PM »

Now that FRALB voting has been released I think it might be instructive to return to this topic to let debaters know what factors go into different methods for FRALB voting. Will, you seemed to indicate that you found the Glicko/ELO system, as it is currently constructed, to have major flaws. What different metrics and considerations did you take into account when doing your voting? If I remember correctly, you used to post the system that you used to go about your FRALB voting and it could be helpful to discuss the differences between how voters go about the process in relation to how Jason's model functions throughout the course of the year.

Kyle
« Last Edit: April 06, 2016, 01:11:38 PM by kylevint » Logged
jregnier
Jr. Member
**
Posts: 94


« Reply #11 on: April 08, 2016, 08:40:55 AM »

final 2015-16 ratings: http://collegedebateratings.weebly.com/home/final-2016-ratings
Logged
Pages: [1]
  Print  
 
Jump to:  

Powered by MySQL Powered by PHP Powered by SMF 1.1.19 | SMF © 2013, Simple Machines
SMF customization services by 2by2host.com
Valid XHTML 1.0! Valid CSS!