EPTAwesome job!
[CSL] Playoffs Week 3: Ro16 - Page 4
| Forum Index > StarCraft 2 Tournaments |
|
yenta
Poland1142 Posts
Awesome job! | ||
|
T.O.P.
Hong Kong4685 Posts
| ||
|
hazelynut
United States2196 Posts
 | ||
|
T.O.P.
Hong Kong4685 Posts
| ||
|
Slardar
Canada7595 Posts
 Was a pretty baddly played series on our part. gj RPI | ||
|
yenta
Poland1142 Posts
On April 10 2011 19:40 T.O.P. wrote: This Glicko system is bad. University of Maryland College Park is #1 and University of Alberta #5. They're not in the playoffs so they can't lose anymore games (can't lose any points). Do you have a suggestion for a better ranking? I'll gladly switch it out. | ||
|
CuriousMoose
United States73 Posts
| ||
|
shindigs
United States4795 Posts
Finally, UCI! | ||
|
hazelynut
United States2196 Posts
exo vs. nerzhul | ||
|
Nizzy
United States839 Posts
| ||
|
T.O.P.
Hong Kong4685 Posts
On April 11 2011 02:11 yenta wrote: Do you have a suggestion for a better ranking? I'll gladly switch it out. I don't know if there's a better ranking system. It's kinda weird when UCSD beats Indiana University 3-2. UCSD falls from 1787 to 1779. Indiana University falls from 1782 to 1759. | ||
|
hazelynut
United States2196 Posts
 4s now haha | ||
|
ExO_
United States2316 Posts
| ||
|
DarthThienAn
United States2734 Posts
| ||
|
yenta
Poland1142 Posts
On April 11 2011 12:30 T.O.P. wrote: I don't know if there's a better ranking system. It's kinda weird when UCSD beats Indiana University 3-2. UCSD falls from 1787 to 1779. Indiana University falls from 1782 to 1759. The RD values initially being set too high coupled with a close margin victory having a different function for determining the outcome value. Even though the ratings fell, so did the RD values (which are hidden). A lower rating with a lower RD is "better" under Glicko than a higher rating with a high RD because you are more sure of the rating. On April 11 2011 12:58 ExO_ wrote: not a big fan of the ranking system. UH is down at #27. Explain how that happened Despite being ranked 1st in their division due to tiebreakers, UH was actually 3rd when considering set differentials. Texas A&M would than be the highest seed from Leviathan and the rankings reflect that. --- I think that Glicko was a bad choice because the relatively small amount of matches played between teams per season does not really allow for the RD values to converge. It would be better for a individual league with many matches played over a short set of time. Also, From Page 22 of A Comprehensive Guide to Chess Ratings by Mark Glickman One of the fundamental problems with using the rating system as a predictor of performace is that it is only accurate on a within- ... The only true remedy to this problem is to ensure While the playoffs sort of address this issue, they are not sound enough to address the problem. We are thinking about heavily jumbling teams between divisions (aside from regional rivalries) which would lead to better ratings eventually. However, as more seasons pass, the performance of a team changes due to graduates and incoming frosh. Reflecting that is tricky in itself. And From DavasiaN's comment under the Stats Announcement I understand that power ranking is just for fun and one of the few metrics we have to compare schools in different divisions, but there are many variables that the glicko system simply can not account for that the rankings become quite inaccurate. Additionally, the current system takes into account playoff results, which will artificially rank teams to be similar to the end results of the tournament. This is simply because the playoffs is single-elimination style tournament. Since almost every team who made the playoffs was at most 2 games apart in record, the winner of the tournament will pretty much be guaranteed to be ranked #1. It also punishes teams that make the playoffs and lose in the first round. Take my school, University of Illinois - Urbana Champaign (Dragon division), for example. We ended up making the playoffs, beating out University of Alberta and McGill University for the last playoff spot based on set differential. Since we were the #4 seed, we played a higher seed in Waterloo and ended up losing. Because of this we have an extra loss compared to Alberta and McGill, and invariably got punished in the rankings for making the playoffs. Both University of Alberta and McGill are ahead of us in the power rankings when they did not even make the playoffs. I'm going to keep experimenting and replace the glicko system with just a plain elo system with varying K values. I will probably skew the results by:
My ultimate goal is for the rating to reflect how teams performed during the season. Looking back, the rating system was mislabeled as a "Power Rank". My idea for the future is to provide
and to perhaps reach a "Power Rank" by weighting and combining the 3. | ||
|
inn5013orecl
United States227 Posts
-SoFLUFFY | ||
|
pasdefromage
Canada11 Posts
http://torontostarcraft.tumblr.com/post/4542053489/csl-interview-university-of-waterloo Thanks again to Monk for his time, and gogogo UW! There's also an interview with another Canadian uni (Wilfred Laurier) on the site, for those interested in the little team that could! | ||
|
SniXSniPe
United States1938 Posts
Q: What or who are the biggest challenges your team is currently facing in the CSL? A: Cheeses and snipe builds. In a best of 7, no school other than UToronto poses any threat to UW playing straight up. One of the main things that I tell my players to practise against is cheese and early aggression. We’re fully aware of all cheese and tailoring the builds accordingly. If my players stay committed and keeps practising, UW shouldn’t even drop a single set until Grand Finals. | ||
|
stalife
Canada1222 Posts
| ||
|
hazelynut
United States2196 Posts
| ||
| ||
