[gordontd]

Free, on 2010-October-26, 05:47, said:

So why don't you explain it and help the OP instead of targetting someone's (in your eyes) inferior method?

Sure. You get the bottom three teams to play two half-matches, using an appropriate VP scale that gives them up to 10 VPs for each half. The movement within the three tables is a mini American Whist.

Imagine you are playing 8 board matches. The three teams in the triple will have 12 boards in play - four on each table. After playing four boards against one of their opposing teams, the EW pairs take the completed boards back to their team-mates (for them to play), and then go on to play at the third table. After playing both half-matches they return to their team-mates to score up.

You need to have spare boards available for the extra half-set, and it's a good idea for them to be very high numbers so they don't get confused with other boards in play.

[RMB1]

I don't read posts with large bold letters in, or the posts they refer to, so I might be repeating what has already been said.

You would do well to get an extra team, or play an extra round and play a complete movement. You can play six rounds with seven teams, where matches are completed after one or two rounds.

AvB/BvC/CvA DvE/EvD FvG/GvF
AvC/BvA/CvB DvF/FvD EvG/GvE

AvD/DvG/GvA BvE/EvB CvF/FvC
AvG/DvA/GvD BvF/FvB CvE/EvC

AvE/EvF/FvA BvD/DvB CvG/GvC
AvF/EvA/FvE BvG/GvB CvD/DvC

Playing 5 rounds with 7 teams with 3-way matches involves 15 matches as part of the 3-ways, leaving only 6 assignments for the rest of the matches, so this will not work. You need to play some "long triangles" with 3-way matches spread over two rounds, and indeed you must play two long triangles during four of the five rounds. This is almost the schedule above for six rounds: one team must play in the long triangle for the first four rounds. You can then play a "short triangle" 8 board 3-way for the final round. You can try to preserve some of the Swiss element, but it will be a sham.

So in round 1: three teams start a long triangle, two head-to-head matches. Round 2: complete first long triangle, winners and losers of head-to-head play another head-to-head match.

For rounds 3/4, one team stays in the triangle, joined by two teams who have not played yet; the other four teams play their possible unplayed matches (two head-to-head matches in each round).

In round 5, the team who stayed in the triangle before must play in the short triangle with the two teams who have not played in the triangle. The remaining two head-to-head matches can be assigned on Swiss basis.

Alternatively, round 5 could be a Danish round. 1st v 2nd, 3rd v 4th and remainder in the short triangle, ignoring previous assignments.

[Free]

gordontd, on 2010-October-26, 07:03, said:

Sure. You get the bottom three teams to play two half-matches, using an appropriate VP scale that gives them up to 10 VPs for each half. The movement within the three tables is a mini American Whist.

Imagine you are playing 8 board matches. The three teams in the triple will have 12 boards in play - four on each table. After playing four boards against one of their opposing teams, the EW pairs take the completed boards back to their team-mates (for them to play), and then go on to play at the third table. After playing both half-matches they return to their team-mates to score up.

You need to have spare boards available for the extra half-set, and it's a good idea for them to be very high numbers so they don't get confused with other boards in play.

If I understand this correctly, it looks like you're using the same principle as I do for seating, and swap around after 4 boards so you can compare the teams against each other based on 4 boards...

The difference is that you compare AvsB, BvsC and AvsC based on 4 boards and need to change VP scales or divide the total by 2, while my method just calculates the scores team per team based on 8 boards and needs no rescaling or different VP scales.

[gordontd]

Free, on 2010-October-26, 10:48, said:

If I understand this correctly, it looks like you're using the same principle as I do for seating, and swap around after 4 boards so you can compare the teams against each other based on 4 boards...

The difference is that you compare AvsB, BvsC and AvsC based on 4 boards and need to change VP scales or divide the total by 2, while my method just calculates the scores team per team based on 8 boards and needs no rescaling or different VP scales.

Yes, but for me it's not teams if you don't play the same boards against the same team as your team-mates.

I thought this process of having a short triple was universal. I'm surprised that you appear not to know it. It's not complicated when you're prepared for it - we always carry three table cards, with 0-10 VP scale on them, to any Swiss Teams we run.

[Cascade]

It should be possible to do one of the following:

1. Play a three-team match with two half matches for each team against each of the other two teams

2. Similar to 1. but over two rounds play a full match against each other team. This method has obvious problems if there is scheduled an odd number of rounds. Perhaps you can play the half matches at the end.

Either of these methods can be programmed so that when you play the three-way matches you are playing in 2. two other teams that you have not played and in 1. two other teams that you have played at most for a half match in a previous three-way.

The first method for small fields has the potential to break down quickly if you want to avoid replays.

As an aside does anyone know of a good discussion of algorithms to make a Swiss Draw? I am particularly interested in any tricks to make an algorithm more efficient when making a draw around the half way point - round n for 2n teams - which is where algorithms seem to have problems.

[Gerardo]

Cascade, on 2010-October-26, 19:27, said:

It should be possible to do one of the following:

As an aside does anyone know of a good discussion of algorithms to make a Swiss Draw? I am particularly interested in any tricks to make an algorithm more efficient when making a draw around the half way point - round n for 2n teams - which is where algorithms seem to have problems.

Backtracking seems a good way to find a pairing set.

Had seen some chess literature, they had added complexity though (alternate colors, dense scoring, floating players).

[Cascade]

Gerardo, on 2010-October-26, 21:44, said:

Backtracking seems a good way to find a pairing set.

Had seen some chess literature, they had added complexity though (alternate colors, dense scoring, floating players).

In my experience brute force backtracking can take a very long time in the worst (but relatively) common cases around the middle of the draw.

I was wondering if there was anything better.

[benlessard]

If you dont have a break time I would make a round robin without a bye.

put team 1 table 1,team 2 table 2, etc.

tell the EW to go 3 table (not sure about this) shuffle and play.

The first 3 table to have finished get an extra boards. (7 instead of 6)
After each round board go down 2 table and pair go up 3 table (not 100% sure about this)

after 6 round every teams has played 6-7 boards vs all other teams.

If you need a break in between, putting 3 boards per tables will be annnoying because timing will suck a bit.

[Bbradley62]

Bottom line: There are many good answers, but 5 rounds of Swiss is not among them.