On October 11 2013 09:47 Shellshock wrote:
Oh no! Maru has dropped to 11/300k not qualifying instead of 1/300k! He's doomed!
Oh no! Maru has dropped to 11/300k not qualifying instead of 1/300k! He's doomed!
are you making fun of me

Forum Index > SC2 General |
Dodgin
Canada39254 Posts
On October 11 2013 09:47 Shellshock wrote: Oh no! Maru has dropped to 11/300k not qualifying instead of 1/300k! He's doomed! are you making fun of me ![]() | ||
Die4Ever
United States17588 Posts
On October 11 2013 09:47 Shellshock wrote: Oh no! Maru has dropped to 11/300k not qualifying instead of 1/300k! He's doomed! It's out of 3 million now, not 300k, I changed that last night after much optimization | ||
![]()
Shellshock
United States97274 Posts
On October 11 2013 09:49 Die4Ever wrote: Show nested quote + On October 11 2013 09:47 Shellshock wrote: Oh no! Maru has dropped to 11/300k not qualifying instead of 1/300k! He's doomed! It's out of 3 million now, not 300k, I changed that last night after much optimization oh phew. he's still ok On October 11 2013 09:48 Dodgin wrote: Show nested quote + On October 11 2013 09:47 Shellshock wrote: Oh no! Maru has dropped to 11/300k not qualifying instead of 1/300k! He's doomed! are you making fun of me ![]() no D: | ||
Die4Ever
United States17588 Posts
| ||
5c0rp10n
Germany342 Posts
On October 10 2013 16:27 Die4Ever wrote: Show nested quote + On October 10 2013 16:24 5c0rp10n wrote: Hey, nice thread! One question i want to know. Which program do you use to simulate that stuff? Or which coding language do you use? I wrote it all myself (it uses the help of aligulac ratings) in C++ Wow, that is huge. Didn't know that this is possible with C++ and i program with it too. :D Would you mind sending me the code? Is it big? I dont wanna use it, im just interested in how it works. A flowchart would be enough tho, if you created something like that. ![]() | ||
Marcinko
South Africa1014 Posts
On October 11 2013 04:46 Dragoonstorm7 wrote: Show nested quote + On October 11 2013 02:57 Die4Ever wrote: On October 11 2013 02:41 Marchinko wrote: Maybe add a link to the liquipedia Blizzcon page? You mean this one? http://wiki.teamliquid.net/starcraft2/2013_StarCraft_II_World_Championship_Series A direct link to the current Standings page is probably more useful for this thread. http://wiki.teamliquid.net/starcraft2/2013_StarCraft_II_World_Championship_Series/Standings (i know getting to standings from your link is easy as pie, but sometimes people overlook the most obvious things; i do so often ![]() Yeah, this one. | ||
yuliuswin
Indonesia24 Posts
| ||
Die4Ever
United States17588 Posts
On October 11 2013 18:44 5c0rp10n wrote: Show nested quote + On October 10 2013 16:27 Die4Ever wrote: On October 10 2013 16:24 5c0rp10n wrote: Hey, nice thread! One question i want to know. Which program do you use to simulate that stuff? Or which coding language do you use? I wrote it all myself (it uses the help of aligulac ratings) in C++ Wow, that is huge. Didn't know that this is possible with C++ and i program with it too. :D Would you mind sending me the code? Is it big? I dont wanna use it, im just interested in how it works. A flowchart would be enough tho, if you created something like that. ![]() Haha thanks. My code is very messy at the moment, maybe I'll clean it up, or I might just release the source code for the 2014 version. The basic idea of it is + Show Spoiler [pseudo code] + class Player; class Event { int player_id; int tournament_id; int placing; } class Sim { vector<Player> players;//their state at the end of the simulation vector<Event> events; } class Match { vector<Player> players; void simulate(Sim &sim); }; class Round { vector<Match> matches; int points_for_4th,points_for_3rd,points_for_2nd,points_for_1st; void simulate(Sim &sim) { /* Here I add the points to the player in the sim object depending on their placement in the match/group I also push them to the next round if they placed high enough to advance. If they don't advance in this match then I add an event in the sim object for their placement in the tournament. */} }; class Tournament { vector<Round> rounds; void simulate(Sim &sim); } And then I loop through simulating the tournaments. for(int i=0; i<num_sims; i++) { for(int t=0; t<tournaments.size();t++) { tournaments[t].simulate(simulations[i]); } } Now I have a ton of simulation runs each with a ton of events. So then I do something like this for(int i=0;i<players.size();i++) { find_big_events_for_player_and_group_them(); /*this finds events that help/hurt the players' chances and groups them together if they exist in the same simulation at least once*/ count_eventgroups_hits(); /*This counts how many times each event/eventgroup that was found in the previous function happens It counts hits and misses separately. A hit is when the event/eventgroup happens AND the player qualifies for blizzcon. A miss is when the event/eventgroup happens AND the player does not qualify for blizzcon. */ sort_eventgroups_for_player(); show_next_match_for_player(); show_top15_eventgroups_for_player(); } edit:better formatting ![]() | ||
Cyrik
Germany50 Posts
On October 12 2013 00:24 Die4Ever wrote: Show nested quote + On October 11 2013 18:44 5c0rp10n wrote: On October 10 2013 16:27 Die4Ever wrote: On October 10 2013 16:24 5c0rp10n wrote: Hey, nice thread! One question i want to know. Which program do you use to simulate that stuff? Or which coding language do you use? I wrote it all myself (it uses the help of aligulac ratings) in C++ Wow, that is huge. Didn't know that this is possible with C++ and i program with it too. :D Would you mind sending me the code? Is it big? I dont wanna use it, im just interested in how it works. A flowchart would be enough tho, if you created something like that. ![]() Haha thanks. My code is very messy at the moment, maybe I'll clean it up, or I might just release the source code for the 2014 version. The basic idea of it is + Show Spoiler [pseudo code] + class Player; class Event { int player_id; int tournament_id; int placing; } class Sim { vector<Player> players;//their state at the end of the simulation vector<Event> events; } class Match { vector<Player> players; void simulate(Sim &sim); }; class Round { vector<Match> matches; int points_for_4th,points_for_3rd,points_for_2nd,points_for_1st; void simulate(Sim &sim) {/* Here I add the points to the player in the sim object depending on their placement in the match/group, I also push them to the next round if they placed high enough to advance. If they don't advance in this match then I add an event in the sim object for their placement in the tournament. */} }; class Tournament { vector<Round> rounds; void simulate(Sim &sim); } And then I loop through simulating the tournaments. for(int i=0; i<num_sims; i++) { for(int t=0; t<tournaments.size();t++) { tournaments[t].simulate(simulations[i]); } } Now I have a ton of simulation runs each with a ton of events. So then I do something like this for(int i=0;i<players.size();i++) { find_big_events_for_player_and_group_them();/*this finds events that help/hurt the players' chances, and groups them together if they exist in the same simulation at least once*/ count_eventgroups_hits();/*This counts how many times each event/eventgroup that was found in the previous function happens It counts hits and misses separately. A hit is when the event/eventgroup happens AND the player qualifies for blizzcon. A miss is when the event/eventgroup happens AND the player does not qualify for blizzcon. */ sort_eventgroups_for_player(); show_next_match_for_player(); show_top15_eventgroups_for_player(); } its funny how unreadable c++ is, even when its in pseudo code ![]() | ||
Die4Ever
United States17588 Posts
On October 12 2013 03:03 Cyrik wrote: Show nested quote + On October 12 2013 00:24 Die4Ever wrote: On October 11 2013 18:44 5c0rp10n wrote: On October 10 2013 16:27 Die4Ever wrote: On October 10 2013 16:24 5c0rp10n wrote: Hey, nice thread! One question i want to know. Which program do you use to simulate that stuff? Or which coding language do you use? I wrote it all myself (it uses the help of aligulac ratings) in C++ Wow, that is huge. Didn't know that this is possible with C++ and i program with it too. :D Would you mind sending me the code? Is it big? I dont wanna use it, im just interested in how it works. A flowchart would be enough tho, if you created something like that. ![]() Haha thanks. My code is very messy at the moment, maybe I'll clean it up, or I might just release the source code for the 2014 version. The basic idea of it is + Show Spoiler [pseudo code] + class Player; class Event { int player_id; int tournament_id; int placing; } class Sim { vector<Player> players;//their state at the end of the simulation vector<Event> events; } class Match { vector<Player> players; void simulate(Sim &sim); }; class Round { vector<Match> matches; int points_for_4th,points_for_3rd,points_for_2nd,points_for_1st; void simulate(Sim &sim) {/* Here I add the points to the player in the sim object depending on their placement in the match/group, I also push them to the next round if they placed high enough to advance. If they don't advance in this match then I add an event in the sim object for their placement in the tournament. */} }; class Tournament { vector<Round> rounds; void simulate(Sim &sim); } And then I loop through simulating the tournaments. for(int i=0; i<num_sims; i++) { for(int t=0; t<tournaments.size();t++) { tournaments[t].simulate(simulations[i]); } } Now I have a ton of simulation runs each with a ton of events. So then I do something like this for(int i=0;i<players.size();i++) { find_big_events_for_player_and_group_them();/*this finds events that help/hurt the players' chances, and groups them together if they exist in the same simulation at least once*/ count_eventgroups_hits();/*This counts how many times each event/eventgroup that was found in the previous function happens It counts hits and misses separately. A hit is when the event/eventgroup happens AND the player qualifies for blizzcon. A miss is when the event/eventgroup happens AND the player does not qualify for blizzcon. */ sort_eventgroups_for_player(); show_next_match_for_player(); show_top15_eventgroups_for_player(); } its funny how unreadable c++ is, even when its in pseudo code ![]() Partly due to my formatting lol. | ||
Gatesleeper
Canada300 Posts
2.81% of the time 3000 WCS points is enough to qualify 2.95% of the time 3050 WCS points is enough to qualify 5.37% of the time 3100 WCS points is enough to qualify 12.41% of the time 3125 WCS points is enough to qualify 12.61% of the time 3150 WCS points is enough to qualify 25.40% of the time 3200 WCS points is enough to qualify 37.30% of the time 3300 WCS points is enough to qualify 63.70% of the time 3400 WCS points is enough to qualify 66.59% of the time 3500 WCS points is enough to qualify 97.14% of the time 3600 WCS points is enough to qualify I've been casually following this thread since its inception, and from the start I was curious about this particular calculation. Back then, Revival's 3000 points made him look pretty darn safe, he was at like 80% to qualify. Some people, here and on reddit, were speculating what the points cutoff would be come Blizzcon. We did this becauses Mvp had just been knocked out of Premier league, and we were wondering whether his 3575 points would be enough to stay in the top 16. I believe it was your calculations that said he was 99.9998~% safe. I think the general consensus was that the points cutoff would be somewhere between 3000 and 3100. But looking at the numbers today, it seems the point cutoff will be much higher, somewhere in the range of 3300 to 3500. If Revival hadn't gone to IEM and done well for himself, he would've been dead in the rankings, where before we thought he was 80% safe. What accounts for this shift? I mean, at the beginning, we knew exactly how many points were available to be earned, and by which players. It's not like there was a surprise surplus of points being given out that raised that cutoff number unexpectedly. Why were our original estimations so much lower than the current reality? | ||
Die4Ever
United States17588 Posts
On October 12 2013 14:05 Gatesleeper wrote: Show nested quote + 2.81% of the time 3000 WCS points is enough to qualify 2.95% of the time 3050 WCS points is enough to qualify 5.37% of the time 3100 WCS points is enough to qualify 12.41% of the time 3125 WCS points is enough to qualify 12.61% of the time 3150 WCS points is enough to qualify 25.40% of the time 3200 WCS points is enough to qualify 37.30% of the time 3300 WCS points is enough to qualify 63.70% of the time 3400 WCS points is enough to qualify 66.59% of the time 3500 WCS points is enough to qualify 97.14% of the time 3600 WCS points is enough to qualify I've been casually following this thread since its inception, and from the start I was curious about this particular calculation. Back then, Revival's 3000 points made him look pretty darn safe, he was at like 80% to qualify. Some people, here and on reddit, were speculating what the points cutoff would be come Blizzcon. We did this becauses Mvp had just been knocked out of Premier league, and we were wondering whether his 3575 points would be enough to stay in the top 16. I believe it was your calculations that said he was 99.9998~% safe. I think the general consensus was that the points cutoff would be somewhere between 3000 and 3100. But looking at the numbers today, it seems the point cutoff will be much higher, somewhere in the range of 3300 to 3500. If Revival hadn't gone to IEM and done well for himself, he would've been dead in the rankings, where before we thought he was 80% safe. What accounts for this shift? I mean, at the beginning, we knew exactly how many points were available to be earned, and by which players. It's not like there was a surprise surplus of points being given out that raised that cutoff number unexpectedly. Why were our original estimations so much lower than the current reality? I think originally it was calculated without Dreamhack and IEM. Also some player replacements did increase the estimated points cutoff a little bit. | ||
bludragen88
United States527 Posts
On October 12 2013 14:05 Gatesleeper wrote: Show nested quote + 2.81% of the time 3000 WCS points is enough to qualify 2.95% of the time 3050 WCS points is enough to qualify 5.37% of the time 3100 WCS points is enough to qualify 12.41% of the time 3125 WCS points is enough to qualify 12.61% of the time 3150 WCS points is enough to qualify 25.40% of the time 3200 WCS points is enough to qualify 37.30% of the time 3300 WCS points is enough to qualify 63.70% of the time 3400 WCS points is enough to qualify 66.59% of the time 3500 WCS points is enough to qualify 97.14% of the time 3600 WCS points is enough to qualify I've been casually following this thread since its inception, and from the start I was curious about this particular calculation. Back then, Revival's 3000 points made him look pretty darn safe, he was at like 80% to qualify. Some people, here and on reddit, were speculating what the points cutoff would be come Blizzcon. We did this becauses Mvp had just been knocked out of Premier league, and we were wondering whether his 3575 points would be enough to stay in the top 16. I believe it was your calculations that said he was 99.9998~% safe. I think the general consensus was that the points cutoff would be somewhere between 3000 and 3100. But looking at the numbers today, it seems the point cutoff will be much higher, somewhere in the range of 3300 to 3500. If Revival hadn't gone to IEM and done well for himself, he would've been dead in the rankings, where before we thought he was 80% safe. What accounts for this shift? I mean, at the beginning, we knew exactly how many points were available to be earned, and by which players. It's not like there was a surprise surplus of points being given out that raised that cutoff number unexpectedly. Why were our original estimations so much lower than the current reality? Back when Revival was 80%, he had not played his WCS AM RO32 matches. He was about 50% chance to advance from there and 50% to get eliminated. Since he was eliminated, half of his chances to qualify disappeared (roughly), which dropped him from 80% down to 40%, so it makes sense. Overall, the points cutoff is a bit higher than the average expected one from before, but that means that more people that did well previously continued to do well (which is not too surprising - Aligulac will always lag behind "real" changes in skill, so it would underestimate those people's chances to gain more points). | ||
Die4Ever
United States17588 Posts
On October 12 2013 14:29 bludragen88 wrote: Show nested quote + On October 12 2013 14:05 Gatesleeper wrote: 2.81% of the time 3000 WCS points is enough to qualify 2.95% of the time 3050 WCS points is enough to qualify 5.37% of the time 3100 WCS points is enough to qualify 12.41% of the time 3125 WCS points is enough to qualify 12.61% of the time 3150 WCS points is enough to qualify 25.40% of the time 3200 WCS points is enough to qualify 37.30% of the time 3300 WCS points is enough to qualify 63.70% of the time 3400 WCS points is enough to qualify 66.59% of the time 3500 WCS points is enough to qualify 97.14% of the time 3600 WCS points is enough to qualify I've been casually following this thread since its inception, and from the start I was curious about this particular calculation. Back then, Revival's 3000 points made him look pretty darn safe, he was at like 80% to qualify. Some people, here and on reddit, were speculating what the points cutoff would be come Blizzcon. We did this becauses Mvp had just been knocked out of Premier league, and we were wondering whether his 3575 points would be enough to stay in the top 16. I believe it was your calculations that said he was 99.9998~% safe. I think the general consensus was that the points cutoff would be somewhere between 3000 and 3100. But looking at the numbers today, it seems the point cutoff will be much higher, somewhere in the range of 3300 to 3500. If Revival hadn't gone to IEM and done well for himself, he would've been dead in the rankings, where before we thought he was 80% safe. What accounts for this shift? I mean, at the beginning, we knew exactly how many points were available to be earned, and by which players. It's not like there was a surprise surplus of points being given out that raised that cutoff number unexpectedly. Why were our original estimations so much lower than the current reality? Back when Revival was 80%, he had not played his WCS AM RO32 matches. He was about 50% chance to advance from there and 50% to get eliminated. Since he was eliminated, half of his chances to qualify disappeared (roughly), which dropped him from 80% down to 40%, so it makes sense. Overall, the points cutoff is a bit higher than the average expected one from before, but that means that more people that did well previously continued to do well (which is not too surprising - Aligulac will always lag behind "real" changes in skill, so it would underestimate those people's chances to gain more points). This too. | ||
Sandermatt
Switzerland1365 Posts
On October 12 2013 14:43 Die4Ever wrote: Show nested quote + On October 12 2013 14:29 bludragen88 wrote: On October 12 2013 14:05 Gatesleeper wrote: 2.81% of the time 3000 WCS points is enough to qualify 2.95% of the time 3050 WCS points is enough to qualify 5.37% of the time 3100 WCS points is enough to qualify 12.41% of the time 3125 WCS points is enough to qualify 12.61% of the time 3150 WCS points is enough to qualify 25.40% of the time 3200 WCS points is enough to qualify 37.30% of the time 3300 WCS points is enough to qualify 63.70% of the time 3400 WCS points is enough to qualify 66.59% of the time 3500 WCS points is enough to qualify 97.14% of the time 3600 WCS points is enough to qualify I've been casually following this thread since its inception, and from the start I was curious about this particular calculation. Back then, Revival's 3000 points made him look pretty darn safe, he was at like 80% to qualify. Some people, here and on reddit, were speculating what the points cutoff would be come Blizzcon. We did this becauses Mvp had just been knocked out of Premier league, and we were wondering whether his 3575 points would be enough to stay in the top 16. I believe it was your calculations that said he was 99.9998~% safe. I think the general consensus was that the points cutoff would be somewhere between 3000 and 3100. But looking at the numbers today, it seems the point cutoff will be much higher, somewhere in the range of 3300 to 3500. If Revival hadn't gone to IEM and done well for himself, he would've been dead in the rankings, where before we thought he was 80% safe. What accounts for this shift? I mean, at the beginning, we knew exactly how many points were available to be earned, and by which players. It's not like there was a surprise surplus of points being given out that raised that cutoff number unexpectedly. Why were our original estimations so much lower than the current reality? Back when Revival was 80%, he had not played his WCS AM RO32 matches. He was about 50% chance to advance from there and 50% to get eliminated. Since he was eliminated, half of his chances to qualify disappeared (roughly), which dropped him from 80% down to 40%, so it makes sense. Overall, the points cutoff is a bit higher than the average expected one from before, but that means that more people that did well previously continued to do well (which is not too surprising - Aligulac will always lag behind "real" changes in skill, so it would underestimate those people's chances to gain more points). This too. When will you run the next simulation? | ||
Die4Ever
United States17588 Posts
On October 13 2013 00:20 Sandermatt wrote: Show nested quote + On October 12 2013 14:43 Die4Ever wrote: On October 12 2013 14:29 bludragen88 wrote: On October 12 2013 14:05 Gatesleeper wrote: 2.81% of the time 3000 WCS points is enough to qualify 2.95% of the time 3050 WCS points is enough to qualify 5.37% of the time 3100 WCS points is enough to qualify 12.41% of the time 3125 WCS points is enough to qualify 12.61% of the time 3150 WCS points is enough to qualify 25.40% of the time 3200 WCS points is enough to qualify 37.30% of the time 3300 WCS points is enough to qualify 63.70% of the time 3400 WCS points is enough to qualify 66.59% of the time 3500 WCS points is enough to qualify 97.14% of the time 3600 WCS points is enough to qualify I've been casually following this thread since its inception, and from the start I was curious about this particular calculation. Back then, Revival's 3000 points made him look pretty darn safe, he was at like 80% to qualify. Some people, here and on reddit, were speculating what the points cutoff would be come Blizzcon. We did this becauses Mvp had just been knocked out of Premier league, and we were wondering whether his 3575 points would be enough to stay in the top 16. I believe it was your calculations that said he was 99.9998~% safe. I think the general consensus was that the points cutoff would be somewhere between 3000 and 3100. But looking at the numbers today, it seems the point cutoff will be much higher, somewhere in the range of 3300 to 3500. If Revival hadn't gone to IEM and done well for himself, he would've been dead in the rankings, where before we thought he was 80% safe. What accounts for this shift? I mean, at the beginning, we knew exactly how many points were available to be earned, and by which players. It's not like there was a surprise surplus of points being given out that raised that cutoff number unexpectedly. Why were our original estimations so much lower than the current reality? Back when Revival was 80%, he had not played his WCS AM RO32 matches. He was about 50% chance to advance from there and 50% to get eliminated. Since he was eliminated, half of his chances to qualify disappeared (roughly), which dropped him from 80% down to 40%, so it makes sense. Overall, the points cutoff is a bit higher than the average expected one from before, but that means that more people that did well previously continued to do well (which is not too surprising - Aligulac will always lag behind "real" changes in skill, so it would underestimate those people's chances to gain more points). This too. When will you run the next simulation? Once the IEM ro8 is set. | ||
bludragen88
United States527 Posts
| ||
Sandermatt
Switzerland1365 Posts
On October 13 2013 01:09 bludragen88 wrote: Well that closes the door on Revival and TLO I guess... we're down to basically people in the top 16 plus anyone alive in WCS AM/Season 3 finals. The current statistic says revival has a 32% chance if he looses. | ||
Die4Ever
United States17588 Posts
| ||
Die4Ever
United States17588 Posts
On October 13 2013 01:32 Sandermatt wrote: Show nested quote + On October 13 2013 01:09 bludragen88 wrote: Well that closes the door on Revival and TLO I guess... we're down to basically people in the top 16 plus anyone alive in WCS AM/Season 3 finals. The current statistic says revival has a 32% chance if he looses. Unfortunately for Revival, Hyun and Naniwa both advanced. So now it's 27.4%. | ||
| ||
![]() StarCraft 2 Stormgate Dota 2 League of Legends Counter-Strike Super Smash Bros Heroes of the Storm Other Games Grubby8533 summit1g6176 ScreaM3935 Dendi1888 fl0m1772 Beastyqt1497 B2W.Neo997 FrodaN856 XBOCT383 ArmadaUGS276 Trikslyr56 Dewaltoss44 ViBE35 JuggernautJason24 Organizations Other Games StarCraft 2 StarCraft: Brood War
StarCraft 2 • StrangeGG StarCraft: Brood War![]() • davetesta17 • Kozan • Migwel ![]() • AfreecaTV YouTube • intothetv ![]() • sooper7s • IndyKCrew ![]() • LaughNgamezSOOP Dota 2 Other Games |
Replay Cast
Wardi Open
Monday Night Weeklies
PiGosaur Monday
The PondCast
Replay Cast
SOOP StarCraft League
CranKy Ducklings
[BSL 2025] Weekly
SOOP StarCraft League
[ Show More ] Sparkling Tuna Cup
|
|