EPTCode S Groups Released - Page 8
| Forum Index > SC2 General |
|
AssyrianKing
Australia2118 Posts
| ||
spena
Canada116 Posts
| ||
|
Wintex
Norway16849 Posts
| ||
|
TriZ
Belgium379 Posts
Group B: RorO & Happy Group C: Hero & Curious Group D: Rain & ? Group E: Mvp & Creator Group F: Marineking & Seed Group G: Leenock & KeeN Group H: Life & Ryung Kinda what I want and expect at the same time.. Really can't decide on second one of group D | ||
|
gruff
Sweden2276 Posts
On October 27 2012 20:58 spena wrote: Rain will walkover Group C so hard.. It would be impressive if he could crush two groups at the same time. | ||
Shellshock
United States97277 Posts
On October 27 2012 19:24 TheBB wrote: These are the probabilities for each player to advance, based on their current Elo ratings. Note that the ratings could change before the groups are played (as will certainly be the case with DRG and Rain for example). Group F is certainly the most even group, with the most and least likely players not even 4% away from a pure coinflip. Group D is also very close. Group C, G and H are the most uneven. The biggest favourites to advance are Life, Hero, Squirtle, Hero, Curious, Taeja and Leenock. The biggest favourites to drop out are Soulkey, Bogus, Bbyong and Baby. (Note that some Kespa players are still underrated.) + Show Spoiler [Group A] + Group A TaeJa: 77.553% PartinG: 49.076% Polt: 43.950% finale: 29.421% + Show Spoiler [Group B] + Group B Symbol: 66.833% RorO: 60.818% Hack: 41.677% IMHappy: 30.672% + Show Spoiler [Group C] + Group C HerO: 85.972% Curious: 79.079% Maru: 26.324% Neo.G_Soulkey: 8.625% + Show Spoiler [Group D] + Group D By.Sun: 61.782% GuMiho: 49.263% HyuN: 45.195% YoDa: 43.760% + Show Spoiler [Group E] + Group E Creator: 69.455% Mvp: 69.032% DongRaeGu: 50.428% Bbyong: 11.085% + Show Spoiler [Group F] + Group F MarineKing: 53.273% sHy: 51.546% Sniper: 48.617% Seed: 46.565% + Show Spoiler [Group G] + Group G Squirtle: 85.393% Leenock: 76.368% KeeN: 28.836% Bogus: 9.403% + Show Spoiler [Group H] + Group H Life: 94.610% Vampire: 46.012% Ryung: 41.083% BaBy: 18.295% I like how you were doing this for the up and down groups and it seems relatively accurate, but I have a question about how you come up with the percentages. Is the program based on assuming they all play each other? I guess I'm just having trouble seeing how it would predict a group that isn't using a round robin format. I could see how it's all working with the up and down matches but do you have it set up so it takes into account who their first opponent is + assuming who the second opponent is, etc? Or is it just in general? Very interesting stuff, though. | ||
opterown
Australia54784 Posts
On October 27 2012 21:24 Shellshock1122 wrote: I like how you were doing this for the up and down groups and it seems relatively accurate, but I have a question about how you come up with the percentages. Is the program based on assuming they all play each other? I guess I'm just having trouble seeing how it would predict a group that isn't using a round robin format. I could see how it's all working with the up and down matches but do you have it set up so it takes into account who their first opponent is + assuming who the second opponent is, etc? Or is it just in general? Very interesting stuff, though. i think he runs a thousand simulations, which would take into account the dual tournament nature? | ||
|
Larkin
United Kingdom7161 Posts
| ||
|
ES.Genie
Germany1370 Posts
On October 27 2012 17:13 X3GoldDot wrote: mana got seeded into ups and downs, and he proved that he could qualify to code S from there....... Yeah and then he managed to get to Code B right that season. Great achievement... | ||
|
Mattchew
United States5684 Posts
| ||
|
TheBB
Switzerland5133 Posts
On October 27 2012 21:24 Shellshock1122 wrote: I like how you were doing this for the up and down groups and it seems relatively accurate, but I have a question about how you come up with the percentages. Is the program based on assuming they all play each other? I guess I'm just having trouble seeing how it would predict a group that isn't using a round robin format. I could see how it's all working with the up and down matches but do you have it set up so it takes into account who their first opponent is + assuming who the second opponent is, etc? Or is it just in general? Very interesting stuff, though. For a round robin group I run 100,000 simulations and count the frequency of each outcome. (Opterown's suggestion of 1,000 is really not enough to get accurate numbers in a 6-player group, for instance. Ideally I would like to do even more.) For simpler tournament formats (a single elimination bracket for example), it's actually pretty easy to calculate probabilities exactly. You just start at the earliest round and calculate the probability of each player advancing based on his or her probability of winning a BoN against each of the possible opponents. Since the two sides of a bracket are completely independent, this is pretty straightforward. (If you know your maths.) For a double elimination bracket, it's much, much more difficult, and once I get around to implementing it I will probably resort to the Monte Carlo method again (as with round robin groups). The complexity stems from the non-independence of different outcomes, mixing losers and winners, etc. However, the MSL style groups that the GSL uses are relatively simple special cases of a double elimination bracket where the list of possible outcomes isn't so large that you can't compute them exactly. Basically, what I do is this: for each set of possible outcomes in the first two matches (4), compute the probability of that outcome happening, then for each possible outcome in the winner's and loser's match (also 4), compute the probability of those happening, and then just add each possible outcome from the final match. Sum up and you're done. Ultimately everything is based on the probability of a player winning a BoN match against another player, which is again based on (a) combinatorics and (b) the probability of a player winning a single game. And that's where Elo ratings and TLPD come in. Once I feel like my code is user friendly enough (and possibly once I have implemented general double elimination brackets), I'd like to write a blog post or something similar. Until then, if you feel like braving my undocumented code (and sometimes poor error handling), you can try it out yourself here (requires Python, I use 3.2.x but it might work on 2.7.x too): https://github.com/TheBB/simul PM me if you're interested in a brief user guide or something.... Thanks for the thumbs up, by the way. It's nice that it gets noticed.  | ||
|
X3GoldDot
Malaysia3840 Posts
On October 27 2012 21:37 ES.Genie wrote: Yeah and then he managed to get to Code B right that season. Great achievement... if he was that bad he wouldve not got into code S in the first place. | ||
|
ACrow
Germany6583 Posts
On October 27 2012 18:39 azzih wrote: Where are the foreign guys seated: TLO, Naniwa,Stephano... arent they all still in korea? They still could get a Code A seat, assuming they weren't all promised to Kespa as well. Stephano is out of the picture though, as he has a ton of tournaments to attend and made it pretty clear that he has no interest in competing in GSL. | ||
|
Shellshock
United States97277 Posts
On October 27 2012 21:54 TheBB wrote: For a round robin group I run 100,000 simulations and count the frequency of each outcome. (Opterown's suggestion of 1,000 is really not enough to get accurate numbers.) For simpler tournament formats (a single elimination bracket for example), it's actually pretty easy to calculate probabilities exactly. You just start at the earliest round and calculate the probability of each player advancing based on his or her probability of winning a BoN against each of the possible opponents. Since the two sides of a bracket are completely independent, this is pretty straightforward. (If you know your maths.) For a double elimination bracket, it's much, much more difficult, and once get around to implementing it I will probably resort to the Monte Carlo method again (as with round robin groups). The complexity stems from the non-independence of different outcomes, mixing losers and winners, etc. However, the MSL style groups that the GSL uses are relatively simple special cases of a double elimination bracket where the list of possible outcomes isn't so large that you can't compute them exactly. Basically, what I do is this: for each set of possible outcomes in the first two matches (4), compute the probability of that outcome happening, then for each possible outcome in the winner's and loser's match (also 4), compute the probability of those happening, and then just add each possible outcome from the final match. Sum up and you're done. Ultimately everything is based on the probability of a player winning a BoN match against another player, which is again based on (a) combinatorics and (b) the probability of a player winning a single game. And that's where Elo ratings and TLPD come in. Once I feel like my code is user friendly enough (and possibly once I have implemented general double elimination brackets), I'd like to write a blog post or something similar. Until then, if you feel like braving my undocumented code (and sometimes poor error handling), you can try it out yourself here (requires Python, I use 3.2.x but it might work on 2.7.x too): https://github.com/TheBB/simul PM me if you're interested in a brief user guide or something.... Ok cool yea I'll definitely try it out. I was telling my friend about it and he's really into coding so he wanted to see how you did it. Thanks! | ||
|
PVJ
Hungary5221 Posts
Let's see My heart wants: + Show Spoiler + Grp A: TaeJa Finale Polt PartinG Grp B: Symbol RorO Hack Happy Grp C: HerO Soulkey Maru Curious Grp D: By.Rain HyuN YoDa GuMiho Grp E: Mvp Bbyong Creator DongRaeGu Grp F: MarineKing sHy Sniper Seed Grp G: Leenock Bogus KeeN Squirtle Grp H: Life BaBy Vampire Ryung While my head says: + Show Spoiler + Grp A: TaeJa Finale Polt PartinG Grp B: Symbol RorO Hack Happy Grp C: HerO Soulkey Maru Curious Grp D: By.Rain HyuN YoDa GuMiho Grp E: Mvp Bbyong Creator DongRaeGu Grp F: MarineKing sHy Sniper Seed Grp G: Leenock Bogus KeeN Squirtle Grp H: Life BaBy Vampire Ryung Also while I think it's definitely the group with Mvp, Creator & DRG which seems to be the deadliest, with PartinG's group as a close second, the Life - BaBy - Vampire - Ryung, and Leenock - Bogus - KeeN - Squirtle, are frightening because of how even all the players seem to be skill-wise. This season is just stacked. I can't wait for it to start. EDIT: Now that I re-read what I wrote I really feel like basically all the groups are pretty even, maybe only YoDa, Maru, Hack, Sniper and Vampire are the ones that I feel just aren't on the same level, as the others. EDIT2: Hahaha, poor finale, he's just so grey I completely forgot to even mention him :D He's totally out of his league, I'd say. | ||
|
ACrow
Germany6583 Posts
A: Parting, Taeja close group, Polt could easily make it as well B: Symbol, Happy C: Hero, Curious D: Rain, Gumiho Hyun didn't impress me with his 1a syndrome in the up&downs, this has a lot of potential to be exploited by a prepared Rain and the gifted multitasker Gumiho E: MVP, DRG I know Creator is gifted and all, but I don't think he really will become a Code S mainstay, he just lacks ... something, don't know why or what. F: Seed, MKP G: Leenock, Keen Squirtle might make it, but he didn't look too good lately, Keen on the other hand finally seems to have realized that Nada is gone. H: Life, Ryung Overall Code S champion predicition: Heart: HerO Brain: Rain Edit: I think I might underestimate some Kespa players here, if so it's simply because I didn't see enough games of them yet. | ||
|
Melliflue
United Kingdom1389 Posts
On October 27 2012 21:42 Mattchew wrote: Back to no foreigners and only liquid as a foreign team even What about FXO? Are they considered Korean or foreign? I guess it would be considered as the Korean part of FXO. I'm not sure what makes a team a foreign team or a Korean team so I never know whether people think of FXO as Korean or not. The concept of a "foreign team" is weird to me anyway, when some "foreign" teams field all-Korean line ups in tournaments outside of Korea. I think Liquid and Fnatic have both done this in the EG Masters' Cup. I just looked it up on Liquipedia and they did it against each other, the Liquid v Fnatic match featured no foreign players. Anyway, on-topic; Mvp, Creator and DRG in one group is unlucky for whoever falls out. The other player in that group was hand-picked by Mvp, which makes it all the more unlikely that he would end up in such a tough group. | ||
|
blacksheepwall
China1530 Posts
Gonna be a great season. Taeja this time (but I was wrong last season). | ||
|
zOverdose
Korea (South)8 Posts
| ||
|
zkeller51
United States25 Posts
| ||
| ||
