rec.games.trading-cards.jyhad

Formula for calculating how many people will meet each other twice in a tourney?

17 messages from 6 participants · 25 February 2009 – 27 February 2009
original thread on Google Groups

Archibald Zimonyi

Hi all, been working for quite some time on a Java version of the Archon. I want to implement the optimal seating that spans across the whole playing field and I have now finally gotten it to work for 16 and 25+ players, so that if you choose the optimal seating randomization variant it will make sure that two people do not meet each other twice in a tournament (this is for a tournament with up to three qualifying rounds). Howeber, for a tournament with say 12 players, or 17 players (and three qualifying rounds), someone will always meet someone else twice, simply due to the fact that you can only move so many players between the rounds in order to avoid this. So I am trying to figure out some kind of formula that allows me to know when I need to allow for meeting conflicts without continuing to try to do my optimization. Does anyone have any idea on how I should proceed? Thanks, Archie

Kevin M.

[ quoted text not captured ] This may not be helpful, but why not just use the seating chart given in the Archon? Why the need to "further" optimize the chart? Kevin M., Prince of Las Vegas "Know your enemy, and know yourself; in one-thousand battles you shall never be in peril." -- Sun Tzu, *The Art of War* "Contentment... Complacency... Catastrophe!" -- Joseph Chevalier Please visit VTESville daily! http://vtesville.myminicity.com/ Las Vegas NAQ 2009! http://members.cox.net/vtesinlv/

Archibald Zimonyi

[ quoted text not captured ] > Las Vegas NAQ 2009!http://members.cox.net/vtesinlv/- Dölj citerad text - > > - Visa citerad text - Because the current chart only goes up to a certain amount of player (albeit a large amount). The optimal seating is (unless it has been changed) also grouped to include batches of 50 people, and you won't meet people outside your batch, my optimized seating seats players across the whole playing field. I also include the option to have true randomization (meaning only the prey/predator rule is in effect) which is supported by the tournament rules (not sure anyone ever plays with that but it still supported). Archie

carn...@gmail.com

[ quoted text not captured ] As far as "won't be at the same table": It's impossible to do with a three-table tournament and three rounds, which rules out 15 or less players. It's also impossible to do with four tables if any of the tables has >4 people, which rules out 17-20 - unless you force the 20 to be five 4-man tables. > Archie -John Flournoy

Archibald Zimonyi

[ quoted text not captured ] > -John Flournoy- Dölj citerad text - > > - Visa citerad text - Yes, hence my first post. I am hoping someone will help me find a simple formula for knowing this, so I can allow for conflicts. Archie

LSJ

[ quoted text not captured ] I think the biggest benefit to an automated "optimal" seating chart comes from the ability to quickly handle drops and adds. And to handle marathon tournaments (4+F or 5+F or more). The rest is probably insufficient (on its own) to justify the effort. Anyway, to the original question: use a cost function. Then choose the seating with the lowest cost (or: stop looking when improvements the cost level off). Choose costs for violations of the optimal criteria based on their relative priority.

carn...@gmail.com

On Feb 25, 2:59 pm, carne...@gmail.com wrote: > On Feb 25, 2:07 pm, Archibald Zimonyi <a...@aranzo.netg.se> wrote: > > > > > Hi all, > > > been working for quite some time on a Java version of the Archon. I > > want to implement the optimal seating that spans across the whole > > playing field and I have now finally gotten it to work for 16 and 25+ > > players, so that if you choose the optimal seating randomization > > variant it will make sure that two people do not meet each other twice > > in a tournament (this is for a tournament with up to three qualifying > > rounds). Additionally, since you it working for 25+ players, you by default have it working for all cases 50 and above; you can group them into optimized batches of 25+ and swap between the batches as you so desire . For instance, in a 50-man tournament, you create two batches of 25. After the first round, swap roughly 1/3 of the people between batches while keeping their distribution order the same (i.e swapping the 'first person table A' from one batch with 'first person table A' from the second.) And then after round two swap a different third. And so forth for larger sizes and batches. 25-49 should be doable as a single batch easily enough. So: 1-15: impossible. 16: solved 17-20: impossible (possible exception: 20 if you insist on 4-man tables) 21-24: unsolved 25+: solved (or at least easily solvable based on what you've already got) > > Archie -John Flournoy

carn...@gmail.com

[ quoted text not captured ] There's not really a need for a formula; unless you have <15 or 16-20, there isn't a need for any conflicts. The other trick is making sure that the distribution you use doesn't put people at a 4-person table instead of a 5-person more often than it has to. > Archie -John Flournoy

Archibald Zimonyi

[ quoted text not captured ] > >> Las Vegas NAQ 2009!http://members.cox.net/vtesinlv/-Dölj citerad text - > > >> - Visa citerad text - > > Because the current chart only goes up to a certain amount of player > > (albeit a large amount). The optimal seating is (unless it has been > > changed) also grouped to include batches of 50 people, and you won't > > meet people outside your batch, my optimized seating seats players > > across the whole playing field. I also include the option to have true > > randomization (meaning only the prey/predator rule is in effect) which > > is supported by the tournament rules (not sure anyone ever plays with > > that but it still supported). > > I think the biggest benefit to an automated "optimal" seating chart comes from > the ability to quickly handle drops and adds. > > And to handle marathon tournaments (4+F or 5+F or more). > > The rest is probably insufficient (on its own) to justify the effort. > > Anyway, to the original question: use a cost function. Then choose the seating > with the lowest cost (or: stop looking when improvements the cost level off). > Choose costs for violations of the optimal criteria based on their relative > priority.- Dölj citerad text - > > - Visa citerad text - Yes, adding removing players will be a function and a benefit for larger tournaments. The others are merely done for my own coding pleasure (which is also why I have been working on this for the last three-four years). Thanks, will look into the cost function and how to implement that. Archie

Archibald Zimonyi

[ quoted text not captured ] No, 1-15 and 17-20 is not impossible. I know I have to allow conflicts here since otherwise my tool would be useless, I want to know just how many conflicts depending on number of players and rounds. The same goes for 21-24. Also, 20 players has to be four table tournament according to the tournament rules (as many 5-player tables as possible). Archie

carn...@gmail.com

[ quoted text not captured ] In fact, in some ways that's the key issue - which is more important? Keeping the equity between 4-man and 5-man tables, or not seeing the same opponent twice? The two may not be mutually achievable. > > Archie -John Flournoy

carn...@gmail.com

[ quoted text not captured ] I have no idea how to fashion formulas to automatically cover variable numbers of rounds and players, honestly. 21-24 players in a 3-round setting doesn't have to have any conflicts, btw, unless you care about evening out appearances at 4 and 5-man tables (and might not even then). It's not hard to just manually determine a possible non-conflicting seating for each of those four cases and work from there. > Also, 20 players has to be four table tournament according to the > tournament rules (as many 5-player tables as possible). Yep, just being completist by noting that 20 is doable with 5 tables. It's obviously got to have conflicts with only 4. > Archie -John Flournoy

Archibald Zimonyi

[ quoted text not captured ] I prefer to handle one problem at a time. Right now I just want a tournament with 17 players to actually end (instead of currently trying to come up with a zero conflict solution forever). LSJ also mentioned players should not (in his opinion) have to be seated at position 1 more then once either, also something that I plan on adding later on as an option. Archie

JH

On Feb 25, 11:25 pm, Archibald Zimonyi <a...@aranzo.netg.se> wrote: > LSJ also mentioned players should not (in his opinion) have to be > seated at position 1 more then once either, also something that I plan > on adding later on as an option. > > Archie Here are the complete seating rules revisited. Unless Scott has changed them within the last few years. 1. No pair of players repeat their predator-prey relationship. This is mandatory, by the VEKN rules. 2. No pair of players share a table through all three rounds, when possible. 3. Available VPs are equitably distributed. 4. No pair of players share a table more often than necessary. 5. A player doesn't sit in the fifth seat more than once. 6. No pair of players repeat the same relative position[*], when possible. 7. A player doesn't play in the same seat position, if possible. 8. Starting transfers are equitably distributed. [NOAL] 9. No pair of players repeat the same relative position group[^], when possible. * relative position: prey, predator, cross-table (4-player table), grand-prey, grand-predator (5-player table) ^ relative position group: adjacent or non-adjacent

Archibald Zimonyi

[ quoted text not captured ] No those look like the ones Scott gave me. Need to throw out my current formula and start from scratch again then. See you in four more years :) Archie

Johannes Walch

LSJ schrieb: [ quoted text not captured ] Also note that any Archon not implemented in Excel could offer a proper export format (e.g XML) that could be automatically processed by a White Wolf rating site. Importing Excel files with multiple tabs in anything not Microsoft is a pain. -- If playing against Cock all you need to remember is: Don´t get caught by Cock.

Archibald Zimonyi

[ quoted text not captured ] > >>> Las Vegas NAQ 2009!http://members.cox.net/vtesinlv/-Dölj citerad text - > > >>> - Visa citerad text - > >> Because the current chart only goes up to a certain amount of player > >> (albeit a large amount). The optimal seating is (unless it has been > >> changed) also grouped to include batches of 50 people, and you won't > >> meet people outside your batch, my optimized seating seats players > >> across the whole playing field. I also include the option to have true > >> randomization (meaning only the prey/predator rule is in effect) which > >> is supported by the tournament rules (not sure anyone ever plays with > >> that but it still supported). > > > I think the biggest benefit to an automated "optimal" seating chart > > comes from the ability to quickly handle drops and adds. > > > And to handle marathon tournaments (4+F or 5+F or more). > > > The rest is probably insufficient (on its own) to justify the effort. > > Also note that any Archon not implemented in Excel could offer a proper > export format (e.g XML) that could be automatically processed by a White > Wolf rating site. Importing Excel files with multiple tabs in anything > not Microsoft is a pain. > > -- > If playing against Cock all you need to > remember is: Don´t get caught by Cock. Of course. I am planning to include a lot of nice features (one for example is my hope that organizers need only enter VEKN number and can then fill in by searching player registry for name/city). As LSJ pointed out, dropping/adding players between rounds is a huge plus. As you point out, having a standardized format for exporting tournament results would be a huge help to organizers too (and probably WW as well). Archie