EPTTLnet Poll - Do you prefer the new ESL Swiss group format,…
Do you prefer the new ESL Swiss group format, or the previous round-robin format? :
| Forum Index > SC2 General |
TL.net Bot
TL.net133 Posts
| ||
|
Blargh
United States2103 Posts
| ||
|
LaughNgamez
Canada525 Posts
| ||
|
dbRic1203
Germany2655 Posts
Map score seem to not matter at all and it looks pretty random in general. In the rule book it is stated that they would use map score (but they do not). So I would love to see some clarification on how they decide the matches. It seems like the casters are left in the dark as well, wich doesn t help.. | ||
|
sc2rush
2 Posts
Other than that I like the Swiss | ||
|
Balnazza
Germany1244 Posts
| ||
|
MJG
United Kingdom1373 Posts
| ||
|
Philippe
354 Posts
| ||
|
tigera6
3444 Posts
| ||
|
batatm
Israel116 Posts
| ||
|
mark_lenders
74 Posts
| ||
hjpalpha
Germany339 Posts
pseudo-"swiss" is a massive pain in the ... it causes lots of issues with the automations we implemented over the last years and hence cause additional work for contributors (and that is after i had to invest several hours to get the swissTableLeague module up to its current capabilities) as a clarification: it is no swiss format, it is a triple-elimination bracket with very very very bad (and unpublished) re-seeding rules personal opinion: i very much dislike the pseudo-"swiss" yes it cuts down on the number of irrelevant matches but with such bad (re-)seeding rules and esl not even announcing the re-seeding rules it just is a total mess the stuff they show on stream doesn't help people in understanding what actually is going on either the display we have on lp is already not optimal (imo displaying it as a bracket would be better but for that the seeding rules would have to be logical ...) but what they show on stream is just plain bad imo a clean gsl style group stage is the way better format i personally would even cut the number of players to 16 for eu and 8 for the other 2 regions so we actually get high(er) lvl of play all around btw. you can expect HSC to have a better format than ept  | ||
|
jy_9876543210
265 Posts
On May 14 2023 00:07 hjpalpha wrote: from Liquipedia perspective: pseudo-"swiss" is a massive pain in the ... it causes lots of issues with the automations we implemented over the last years and hence cause additional work for contributors (and that is after i had to invest several hours to get the swissTableLeague module up to its current capabilities) as a clarification: it is no swiss format, it is a triple-elimination bracket with very very very bad (and unpublished) re-seeding rules personal opinion: i very much dislike the pseudo-"swiss" yes it cuts down on the number of irrelevant matches but with such bad (re-)seeding rules and esl not even announcing the re-seeding rules it just is a total mess the stuff they show on stream doesn't help people in understanding what actually is going on either the display we have on lp is already not optimal (imo displaying it as a bracket would be better but for that the seeding rules would have to be logical ...) but what they show on stream is just plain bad imo a clean gsl style group stage is the way better format i personally would even cut the number of players to 16 for eu and 8 for the other 2 regions so we actually get high(er) lvl of play all around btw. you can expect HSC to have a better format than ept So it looks like a swiss format but just much worse? So why don't they just use swiss format, which CS GO major has been doing for years? | ||
|
hjpalpha
Germany339 Posts
On May 14 2023 00:51 jy_9876543210 wrote: So it looks like a swiss format but just much worse? So why don't they just use swiss format, which CS GO major has been doing for years? cs doesn't use swiss either it is pseudo-swiss too to quote wikipedia: A Swiss-system tournament is a non-eliminating tournament format that features a fixed number of rounds of competition, but considerably fewer than for a round-robin tournament; thus each competitor (team or individual) does not play all the other competitors. Competitors meet one-on-one in each round and are paired using a set of rules designed to ensure that each competitor plays opponents with a similar running score, but does not play the same opponent more than once. The winner is the competitor with the highest aggregate points earned in all rounds. With an even number of participants, all competitors play in each round. esl (and other esport TOs) wrongly call it swiss even though it clearly is no swiss format i will probably push something on lp later today explaining this bad format better than esl did | ||
|
jy_9876543210
265 Posts
On May 14 2023 00:58 hjpalpha wrote: cs doesn't use swiss either it is pseudo-swiss too to quote wikipedia: esl (and other esport TOs) wrongly call it swiss even though it clearly is no swiss format i will probably push something on lp later today explaining this bad format better than esl did That's interesting... I've thought about this problem "what's the most efficient algorithm for finding the largest k elements from n by comparing (in parallel)". k=1 is simple, just use single elimination, but it becomes way more complicated after that. Double elimination / GSL format works for k=2, but they don't prevent rematches. I tried to find a way to avoid rematches in swiss format, and I had limited success with binary labeling but there are cases that there are odd number of players with the same score. Also, I noticed that the CSGO swiss format doesn't even guarantee that the top 2 teams with 3-0 are actually top 8 if we don't have any presumed knowledge about them, because #9 team may get really lucky and advance with a 3-0. So anyways, it's a very complicated problem. I wonder how the seeding actually works. | ||
|
hjpalpha
Germany339 Posts
 Pseudo-Swiss format | ||
|
jy_9876543210
265 Posts
On May 14 2023 04:25 hjpalpha wrote: small format explanation: Pseudo-Swiss formatI wonder if you could just modify the csgo module and use it here? It should be easier than writing one from scratch, no? Is it because the seeding rules are different? BTW This is my explanation of the pseudo-swiss format in simple language: You have 16 players and you want to select the top 8, here's what you do: Pair them up, so you have 8 winners (1:0) and 8 losers (0:1); Those who won would face each other, and who lost face each other, the result would be 4 with score (2:0), 8 with (1:1), 4 with (0:2); Pair them up within each score group, so you get 2 players with score (3:0), 2+4=6 with score(2:1), 4+2=6 with score (1:2) and 2 with score (0:3). The two top players advance, the two bottom player are eliminated, Pair them up within each score group, so you get 3 with score (3:1) and 3+3=6 with score (2:2) and 3 with score (1:3), those won 3 games advance, and those lost 3 games are eliminated; Finally the last 6 players in the middle face each other, those who won would advance and those who lost would be eliminated. You get your top 8 players. The only non-trivial problem is how to avoid rematches. (It's like a triple elimination but not really, because it doesn't guarantee that the top 3 selected are actually the best 3. In fact, if the seed is random, rank 9 may advance as top 1 and rank 8 may be eliminated as bottom 1, though the probability is small. But of course, any other common format has this issue too. I think this format is a good balance between complexity and accuracy. E.g, on the wiki page, if the ranks of players I~P are 9~16, player #9 would win 3:0 and advance, although they're not in top 8.) | ||
|
covetousrat
2109 Posts
| ||
|
hjpalpha
Germany339 Posts
On May 14 2023 07:51 jy_9876543210 wrote: I wonder if you could just modify the csgo module and use it here? It should be easier than writing one from scratch, no? Is it because the seeding rules are different? The csgo module is outdated af, has some issues with wrong calculations, only supports teams not players, doesn't support several functionality we want to have on sc2 and uses the old match systems data instead of the new match systems data. And it also has the same limitations the current sc2 version has. So not really an option to use that. And i didn't write one from scratch i had to make an existing one useable for our automations. It still cost me several hours. BTW This is my explanation of the pseudo-swiss format in simple language: You have 16 players and you want to select the top 8, here's what you do: Pair them up, so you have 8 winners (1:0) and 8 losers (0:1); Those who won would face each other, and who lost face each other, the result would be 4 with score (2:0), 8 with (1:1), 4 with (0:2); Pair them up within each score group, so you get 2 players with score (3:0), 2+4=6 with score(2:1), 4+2=6 with score (1:2) and 2 with score (0:3). The two top players advance, the two bottom player are eliminated, Pair them up within each score group, so you get 3 with score (3:1) and 3+3=6 with score (2:2) and 3 with score (1:3), those won 3 games advance, and those lost 3 games are eliminated; Finally the last 6 players in the middle face each other, those who won would advance and those who lost would be eliminated. You get your top 8 players. The only non-trivial problem is how to avoid rematches. or in short: In this format 16 participants play until they have won 3 matches or lost 3 matches. The participants that win 3 matches advance to the next tournament stage while the players that lost 3 matches are eliminated. In each round participants are matched versus an opponent with the same match score. Usually rematches are not permitted. (It's like a triple elimination but not really, because it doesn't guarantee that the top 3 selected are actually the best 3. Triple elimination doesn't guarantee that either. It all depends on the seeding and the results in the matches. There are events that use rnd seeding in brackets. In fact, if the seed is random, rank 9 may advance as top 1 and rank 8 may be eliminated as bottom 1, though the probability is small. But of course, any other common format has this issue too. I think this format is a good balance between complexity and accuracy. E.g, on the wiki page, if the ranks of players I~P are 9~16, player #9 would win 3:0 and advance, although they're not in top 8.) I still heavily prefer gsl style groups (technically a 3 round pseudo swiss but with this one the seeding is obvious and hence no confusion and you directly know the next rounds matchups) over the 5 round pseudo swiss. And ESL does a very bad job at explaining the format. No explanation at all on their website nor in the rules, which makes the rule book very useless this year as it factually is just plain wrong in big chunks. And on stream the graphics are pretty useless too. I think the way we show it on LP with the swissTableLeague is better. I would love to use a triple elim bracket. but for that the seeding would have to be logical and fixed instead of the unknown (and as of now arbitrary and in contradiction to the rule book) seeding they do. | ||
|
Balnazza
Germany1244 Posts
Speaking purely as a user, I don't really see what the problem with the (pseudo)Swiss system is? You can explain it in two sentences and you can see all matches in one table. Sure, the seeding isn't known beforehand for the next rounds, but who really cares about that in a groupstage? Compared to that, I find even the CS:GO ESL Pro League triple-elim bracket confusing and they only use eight teams per group, not sixteen. I had to think rather long about it and I still feel the seeding doesn't make that much sense. Swiss (or "swiss") I did understand in like 45sec top... | ||
| ||
