[nige1]

Many BW editorials and a recent BBO post address the dumping phenomenon. Dumping is OK unless local regulations prohibit it. Anyway it's hard to prove and, In most cases, the fault is with the tournament organizers. Careful seeding might alleviate the problem. Nevertheless, experience at all levels has shown that competitions, involving RR and KO, invite dumping behaviour.

RR+KO championships have other defects. The winner is decided by sudden-death matches that throw out almost all relevant data. Often, the lead changes sides, several times, during the final. Hence, who wins depends on when the music stops. e.g. the latest World-championships.

A problem with relying only on a complete RR, is that matches are short. Rabbit-killing skills are over-emphasized. The best teams are inadequately tested against each other. Admittedly, RR does have the advantage of using all the data.

In contrast, a pure KO format generates half as much data and loses most it. It's advantage is that it allows much longer matches. A major disadvantage of KO format is that, even were all matches won by the best team, the second best team could be eliminated in the first round. The chance element in Bridge exacerbates this problem. With bad luck, even the best team can lose a match against a poor team. A nuisance of KOs is that they leave losing teams at a loose-end, because each round eliminates half the field.

An improvement is double-elimination (e.g. the EBU Spring 4s). It retains long matches, A team must suffer two defeats, before being eliminated, thus reducing the effect of chance.

Taking double-elimination to its logical extreme produces the Swiss format. Easily the best compromise between KO and RR, it allows long matches (like a KO), matching winning teams against each other. All the teams are placed in a pecking order. The result retains most of the data. Swiss is like KO with continuous repêchage. IMO, the main arguments against it are prejudice and tradition.

[barmar]

nige1, on 2019-October-02, 12:43, said:

In contrast, a pure KO format generates half as much data and loses most it. It's advantage is that it allows much longer matches. A major disadvantage of KO format is that, even were all matches won[size="2"] by the best team, the second best team could be eliminated in the first round.

That's why important events are practically always seeded. If everyone runs true to form, the top 2 teams won't meet until the final. Unless there are an unusual number of upsets, the final will pit teams near the top of the ranking.

Double elimination makes this less likely, but even a good team can have two bad sessions in a row.

[pescetom]

You are kicking down an open door as far as I am concerned. Although I've seen Swiss interpreted both with and without a "locked down" last round where the winner at the top table on the last round is the overall winner whatever happens at other tables. Not sure if that is considered implicit or not, but I much prefer it in any case. It's hard enough facing the probable best opponents in the last round without the near certainty that if you are close to a draw then the winner among those you have both already beaten will overtake you.

[sfi]

In Australia, almost all of the tournaments (pairs, teams, national, weekend event, club championship, you name it) are Swiss events. It is clear that the format is dreadful at reliably giving you the top X teams (4, 8 or whatever number you want) and not much better at identifying the "correct" winner.

The reason for this is that your final placing is heavily dependent on who you play in the last round. Say you have 100 teams and take the top 8 for a KO. Before the last round, the teams running 5th-16th are likely to all be in the hunt for a KO spot. Some of them will be highly seeded teams, some will be teams who have just had a big win and are out of their depth, and there will be a range in between. You might be ranked between the number 2 seed and the number 45 seed and your chances vary enormously which one you are drawn against.

It's better if you're choosing one winner, since often one or a small number of teams pull away from the field. But even then there is inherent randomness not present in other formats.

The fundamental problem with a Swiss event is that the matches are of unequal importance. I have a standing offer in any of our major events for the organisers to give me any score they like in the first round in exchange for me being able to sleep in longer. The first round doesn't make any meaningful contribution to our finishing position, and the next couple of rounds not much more.

I think it's important that each round have similar importance, and a Swiss fails this criterion badly. What the Swiss format does do well is provide teams with competitive matches most of the time and hands out masterpoints across the field. These are important factors for healthy tournaments, but not for fair ones.

[sfi]

As an example, take the Bermuda Bowl round robin results and use them for a Swiss ranking. I've started with the actual first round and run from there. Assuming my calculations are correct, here are the rankings after a few rounds. How would you feel if you were Chile and had to play the leaders in round 7 while watching USA 1 thrash Canada? It gets worse for Chile. Here are the teams each of them played in the first 7 rounds:

Chile: Singapore, England, New Zealand, Sweden, Italy, Hong Kong, China
USA 1: Egypt, India, Netherlands, Bangladesh, Morocco, Australia, Canada

Note there are no opponents in common after 7 rounds, and I know which set of opponents I would rather play.

You can find other hard luck stories as well. Hong Kong was in the top 8 the entire way before getting beaten up by Poland. All these fluctuations go away in a full round robin.

Round 6					Round 7
1	China		84.5		1	China		102.9
2	Sweden		84.39		2	Sweden		102.4
3	England		77.63		3	Poland		93.66
4	Poland		76.21		4	USA 1		84.02
5	Chile		75.23		5	Indonesia	81.9
6	Norway		73.89		6	Norway		80.85
7	Indonesia	68.86		7	England		79.66
8	Hong Kong	67.96		8	Israel		77.17
9	Italy		67.66		9	Chile		76.79
10	Israel		64.88		10	Italy		75.37
11	USA 1		64.86		11	Hong Kong	70.51
12	Canada		58.08		12	New Zealand	68.16
13	New Zealand	57.55		13	Netherlands	65.69
14	Netherlands	56.3		14	Guadeloupe	64.99
15	Argentina	56.01		15	Russia		63.69
16	India		55.34		16	India		63.29
17	Australia	50.16		17	Argentina	63.21
18	USA 2		48.34		18	Australia	62.96
19	Guadeloupe	48.11		19	Egypt		62.24
20	Egypt		47.24		20	USA 2		60.39
21	Singapore	46.58		21	Canada		58.92
22	Bangladesh	46.27		22	Singapore	51.58
23	Russia		46.24		23	Bangladesh	49.39
24	Morocco		17.71		24	Morocco		20.26