EPTDivision leader rankings - Page 13
| Forum Index > SC2 General |
|
Slayer-
United States113 Posts
| ||
|
Disarray
United States1164 Posts
On February 23 2010 13:16 choboPEon wrote: he worked for them but no longer I think you're confusing pillars with zileas | ||
|
LaLuSh
Sweden2358 Posts
On February 23 2010 13:50 damenmofa wrote: Ok here are screens of all 6 EU Platinum Divisions (if somebody is in platinum 7 or higher pls correct me, but seeing there are only 8 people in platinum 6 I guess thats it): + Show Spoiler + ![[image loading]](http://i49.tinypic.com/o9l8n8.jpg) ![[image loading]](http://i50.tinypic.com/24cgw3t.jpg) ![[image loading]](http://i46.tinypic.com/29ylimf.jpg) ![[image loading]](http://i50.tinypic.com/23wo3eo.jpg) ![[image loading]](http://i46.tinypic.com/t9uy6o.jpg) ![[image loading]](http://i48.tinypic.com/33k65w6.jpg) Took me like 2 hours to find a gamer for each division, Blizz really could have made checking on rankings a lot easier 8[ Me = diehilde in case you didnt know, obviously im not that selfless to make myself that much work without a lil braggin involved ;P Swedes leading 4/6 of those divisions. Haypro, MinD, Akisto and RunA. Nice work gathering this info. | ||
|
Foxhawk
United States41 Posts
| ||
|
Pillars
United States147 Posts
Any replays of Columns? Haven't uploaded any replays yet though I might in the future. I was gonna ask that. Does he work for Blizzard now? I worked on the design team for a year or so in 2007. | ||
|
Roes
Belgium10 Posts
| ||
|
nicoaldo
Argentina939 Posts
| ||
|
Konni
Germany3044 Posts
 It seems not only rank but also matchup dependent stats and ingame stats are used to determine rank and points. | ||
|
ZpuX
Sweden1230 Posts
On February 23 2010 22:47 Konni wrote: It's really weird. Sometimes you get only matched against lower level players and earn like 4-10 points per match. Sometimes you get equally ranked or higher ranked players and earn like 20 points. Yesterday I had a match where I got 38 points. Don't understand that It seems not only rank but also matchup dependent stats and ingame stats are used to determine rank and points. you really think it's strange? battle.net searches for players closest to you in terms of points and league. If it cant match you with someone with the exact same stats/points/league they will have to make the search bigger... Eventually you will have to play someone with much less points than you because there are no one else near you searching for a game... And if you think it's strange that you get less points for beating worse players and more points for beating better players, well just look at the iccup ranking system, it works like the exact same way. If you cant find someone at your own rank, you will have to look for players either above or below you. | ||
|
SoLaR[i.C]
United States2969 Posts
| ||
|
rANDY
United Kingdom748 Posts
On February 23 2010 09:36 Jyvblamo wrote: Points lost =/= points gained by other guy, by the way. If points lost =/= points gained then its not ELO? Or am I not remembering things correctly here? | ||
|
thOr6136
Slovenia1774 Posts
| ||
|
LaLuSh
Sweden2358 Posts
On February 24 2010 01:44 no_re wrote: If points lost =/= points gained then its not ELO? Or am I not remembering things correctly here? Elo is only absolute within your division. Effectively you are only competing against the people in your own division. And as more players are expected to drop into most leagues/divisions, blizzard don't want the existing players in divisions with a small amount of players to have too much of a head start. As your division fills up with players, elo gained/lost will increase. You can see this pretty clearly watching the bronze leagues, where people reach 1400 elo with crap stats like 19-12. And the top players are at 1800-1900 elo. While people in platinum divisions with a small amount of players can barely break 1200 with stats like 100-10. | ||
|
oberon
United States1320 Posts
On February 24 2010 01:44 no_re wrote: If points lost =/= points gained then its not ELO? Or am I not remembering things correctly here? Blizzard claimed their rating system was "one of the best in existence" or somesuch. The obvious comp is Microsoft's TrueSkill, which is a slightly modified Glicko2, I believe. It seems unlikely Blizzard has decided to adopt something non-iterative (e.g. WHR), but people say their ratings have changed without playing games (doesn't happen in Glicko2), so who knows. --oberon | ||
|
On_Slaught
United States12190 Posts
| ||
|
Asta
Germany3491 Posts
On February 24 2010 04:01 oberon wrote: Blizzard claimed their rating system was "one of the best in existence" or somesuch. The obvious comp is Microsoft's TrueSkill, which is a slightly modified Glicko2, I believe. It seems unlikely Blizzard has decided to adopt something non-iterative (e.g. WHR), but people say their ratings have changed without playing games (doesn't happen in Glicko2), so who knows. --oberon After reading a bit on wikipedia it seems to me that the non-trivial part of the TrueSkill system lies within the fact that players can join different groups to play together. As this doesn't happen in SC2 (every 2on2 team you play in gets its own ranking... or actually, how is it for random teams? doesn't matter...) the rest of the system seems very simple. However there is no mention of considering other variables than just the outcome of the match. I think the same applies for Glicko. There are so many surprising things happening in the rankings (some of which are obviously bugs, like the league-upgrade from platinum to platinum with point penalty) that I am pretty sure the algorithm is either considering other variables or is just full of bugs.  In any case I don't think people will figure it out without the help of Blizzard. | ||
|
Maynard
United States889 Posts
| ||
|
oberon
United States1320 Posts
On February 24 2010 05:09 Asta wrote: After reading a bit on wikipedia it seems to me that the non-trivial part of the TrueSkill system lies within the fact that players can join different groups to play together. As this doesn't happen in SC2 (every 2on2 team you play in gets its own ranking... or actually, how is it for random teams? doesn't matter...) the rest of the system seems very simple. However there is no mention of considering other variables than just the outcome of the match. I think the same applies for Glicko. There are so many surprising things happening in the rankings (some of which are obviously bugs, like the league-upgrade from platinum to platinum with point penalty) that I am pretty sure the algorithm is either considering other variables or is just full of bugs. In any case I don't think people will figure it out without the help of Blizzard. Glicko2 considers both players' ratings, deviations, and consistencies*, where the deviation is a measure of how accurate the rating is (a high deviation means the rating is less accurate) and the consistency measures how consistent the player's results have been (a low consistency indicates the player's true rating has likely changed in some way, and his assigned rating does not yet reflect this). The algorithm can be modified to handle teams -- multiple companies, including Microsoft, have done this. It is possible that Blizzard is using other variables, such as maps and races. This seems unlikely, however, because no "TvZ" rating is available -- just one overall rating. --oberon *"Consistency" is actually not used -- its inverse, volatility is. I felt this was easier to explain | ||
|
Foxhawk
United States41 Posts
| ||
|
starcraft911
Korea (South)1263 Posts
| ||
| ||
