EPTDecember 26 rating list: now tweaked - Page 5
| Forum Index > SC2 General |
|
kollin
United Kingdom8380 Posts
| ||
|
Aunvilgod
2653 Posts
| ||
|
Markwerf
Netherlands3728 Posts
It would probably be better if it was some bayesian estimate or the K value would be lower so it would take longer to reach high ranks. A player like combatex is probably ranked highly because he played few ranked games and did quite well in the most recent ones. Of course this is just conjecture since I didn't go through the data but the foreigner list just feels off | ||
|
Leviance
Germany4079 Posts
| ||
|
jermmanDOTA
Canada45 Posts
| ||
|
kollin
United Kingdom8380 Posts
On December 28 2012 23:27 jermmanDOTA wrote: I think combatex is up there because of how often he cheeses (and how damn good he is at it. Seriously, that shit is ridiculous.) No he's there because he placed top 8 at WCG. This system can't magically tell your play unless it has tournament results to back it up. | ||
|
Nihilnovi
Sweden696 Posts
I can't even fathom what type of alghoritm is used here, how is Grubby not above them considering his recent IEM win, and same for BabyK @ wcs, this whole topic just seems like a giant flamebait :S | ||
|
crbox
Canada1181 Posts
| ||
|
kollin
United Kingdom8380 Posts
On December 28 2012 23:44 Nihilnovi wrote: I can't take any list that ranks Combatex and MacSed as the two top foreign protoss seriously. I can't even fathom what type of alghoritm is used here, how is Grubby not above them considering his recent IEM win, and same for BabyK @ wcs, this whole topic just seems like a giant flamebait :S Grubby didn't win IEM. | ||
|
Corrosive
Canada3741 Posts
On December 28 2012 23:44 Nihilnovi wrote: I can't take any list that ranks Combatex and MacSed as the two top foreign protoss seriously. I can't even fathom what type of alghoritm is used here, how is Grubby not above them considering his recent IEM win, and same for BabyK @ wcs, this whole topic just seems like a giant flamebait :S what i've learned from this thread so far is people don't understand statistics. | ||
|
Salient
United States876 Posts
| ||
|
Whatson
United States5357 Posts
On December 28 2012 23:57 Corrosive wrote: what i've learned from this thread so far is people don't understand statistics. Yup we like our rankings to be an accurate representation of a player's skill, and somewhat realistic. This "ranking" just further proves that statistics are completely overrated. | ||
|
Conti
Germany2516 Posts
On December 29 2012 00:33 Salient wrote: The problem seems to be that it doesn't apply the correct weight to various wins. The fact that MC was #2 in a recent GSL and #3 at OSL should put him in the top 25 players despite his otherwise slumping results. Conversely, the recent top 10 performances by CombatEX at a few tournaments s not really that impressive. The ranking does not weight any games, as there's no automatic (and sensible) way to do so. | ||
|
Salient
United States876 Posts
| ||
|
Grovbolle
Denmark3822 Posts
On December 29 2012 00:33 Salient wrote: The problem seems to be that it doesn't apply the correct weight to various wins. The fact that MC was #2 in a recent GSL and #3 at OSL should put him in the top 25 players despite his otherwise slumping results. Conversely, the recent top 10 performances by CombatEX at a few tournaments s not really that impressive. On December 28 2012 23:44 Nihilnovi wrote: I can't take any list that ranks Combatex and MacSed as the two top foreign protoss seriously. I can't even fathom what type of alghoritm is used here, how is Grubby not above them considering his recent IEM win, and same for BabyK @ wcs, this whole topic just seems like a giant flamebait :S Please TRY at least to understand it. Do You Weigh Games Differently? No, I don't. Korean tournaments and players receive no special treatment. The GSL is difficult because good players play there; the players aren't good because they play in the GSL. The more data we have describing a player, the more we can expect them to play at their given rating. Since both CombatEx and MacSeed aren't that well documented, it still "surprises" the model when they perform well, meaning that the model will think they are better than they are. When we receive more and more information about players, we will have a more accurate representation of a players true skill level on a pretty current level. Winning tournaments does NOT give you a higher rating, beating higher ranked players does. (Which you usually do to win a tournament, but it takes wins, not "overall wins") | ||
|
Conti
Germany2516 Posts
On December 29 2012 00:51 Salient wrote: Any ranking system that places CombatEX above MC is ridiculous. MC won multiple major tournaments and took home well over 100 thousand dollars this year. CombatEX? Not so much. That particular problem will fix itself. Promised. | ||
|
Salient
United States876 Posts
| ||
Milkis
5003 Posts
| ||
|
Grovbolle
Denmark3822 Posts
On December 29 2012 00:54 Salient wrote: That's why you don't rank players at all until you have sufficient data to rank them in a meaningful manner. So please define the sufficient amount of games played before we can rank a player? What about the already ranked player the unranked player played, how would he have points assigned? | ||
|
Grovbolle
Denmark3822 Posts
On December 29 2012 00:55 Milkis wrote: For the results ticker, I'd add in some variable for "offline" play or whether or not players are playing from the same region, just to see how much lag effects "skill". would be something interesting to see. (so EU vs EU matches are fine, but EU vs NA matches would have this flag, etc) Would require an entire rework of the way results are currently added. Maybe it can be done when we "start" fresh with HoTS | ||
| ||
