Can a "group stage" tournament cut straight to a final?

[BananaHarambe]

BGA gets a lot of critique for their "try no to prevent" pairing policy for Swiss, but it's actually quite difficult to solve for 3+p games.

There are already MANY perfect pairing codes out there for 2 player card game and chess tournaments currently in use and there have been several papers published on algorithms. How to do it is more or less "solved" if you do some research. Based on my knowledge I could code it in 2p without too much issue, but it would still take some serious time to get correct.

To my knowledge no one has ever written a paper on how to do Swiss pairings for 3+ player games and no one has proposed an algorithm to do it before. BGA might be the very first.

The order of magnitude of difficulty to code for 3 players is so much higher! Each table has "3 pairings" and the code to shuffle players around to check their "pairing agreement" needs to be so much more complex. The code would be 4-9x as complicated.

Now also consider that BGA supports tournaments for up to 8 player games (correct me if I'm wrong) and they need a code that's universal for all player numbers and Swiss groups can also have uneven tables (4-4-4-4-3-3 for example) and the complexity skyrockets.

I understand why it works the way it does with its limitations and that they warn organizers not to use it for small groups.

[Mathew5000]

BGA gets a lot of critique for their "try no to prevent" pairing policy for Swiss, but it's actually quite difficult to solve for 3+p games.

Granted. But I wish the default at tournament creation was to not try to prevent repeat matchups. I put it in a formal proposal just now: https://boardgamearena.com/bug?id=170168

I understand why it works the way it does with its limitations and that they warn organizers not to use it for small groups.

Yes but it's not just the group size, but the number of rounds in the tournament relative to log of number of players. For a 2p game, if you have 60 players and 6 rounds it's okay to check "try to prevent repeats" but not if you have 60 players and 10 rounds.

And for 3 (or more) players in each game, the "try to prevent" option will inevitably lead to unfairness, where after a few rounds there are some tables with a high-ranked player against low- and medium-ranked players, and other tables with one low-ranked player against some high-ranked players. Am I wrong about that?

[BananaHarambe]

And for 3 (or more) players in each game, the "try to prevent" option will inevitably lead to unfairness, where after a few rounds there are some tables with a high-ranked player against low- and medium-ranked players, and other tables with one low-ranked player against some high-ranked players. Am I wrong about that?

You are correct that if the tournament is too long for the number of players the match quality will continue to drop until the matches become way off. Current example with my friend Andrewsmile here: https://boardgamearena.com/tournament?id=270995

4 player swiss, try not to prevent, only 50 players (down to 49 now) and up to 20 matches, currently on 13. The 1st ranked player in standings in playing against the 32nd, 48th (2nd to last) and 49th (last) in standings. How are players supposed to catch up to the winner in a Swiss if they keep getting easier and easier pairings?

Swiss tournaments generally pair from the top in standings down so when pairings become poor it's usually the top player paired against multiple weak opponents as seen in this example.

My general tournament guideline for Swiss with try to prevent: Players should play less than half of the total players in the tournament.

Applying that here, this tournament will give everyone 60 different opponents. To avoid the top players from needing to pair down below about the middle point, there should be about 120 players or opponents to ensure somewhat reasonable pairings, but there is only 50/49 (not even enough to give all the matches, tournament will break before the end).

Going the other way, this tournament has 49 opponents total so players should play at most 24/25 of them. Based on my guidelines this should have been a 8 round tournament. After 8 rounds the pairing quality drops off quite a bit.

[Meeplelowda]

To my knowledge no one has ever written a paper on how to do Swiss pairings for 3+ player games and no one has proposed an algorithm to do it before. BGA might be the very first.

If you like reading academic papers on game-related topics, check out my post in this thread.