Number of rounds in swiss?

For official USCF tournaments, they set the number of rounds based on the number of days there are in the tournament rather than the number of players.  For a typical weekend tournament (Friday night plus weekend), this will typically be a 5 round Swiss.  If there are short time controls (e.g. G/60 or G/45) or it is a longer tournament (e.g., Thursday through Sunday, Friday through Monday), then they might go to 7 or 9 rounds.  For one-day tournaments, they usually have short time controls and will have only 4 rounds or possibly 3 rounds.  I played a 5 round tournament where there were almost 60 players (high 50s as I recall).  In theory, there could have been a tie for first place but I think that there was actually a clear winner in that tournament.

Of course having more rounds than strictly needed is a good thing. It just make the results more reliable, expecially for those in the middle of the standings.

Are any of you familiar with the McMahon variant of the Swiss system?  If more rounds are always better, this seems to be an excellent method that basically creates a virtual extra round.  In normal Swiss systems, the first round has the biggest rating mismatches since the upper half is paired against the lower half.  If the range of ratings of the participants is wide, then this causes some huge mismatches.  Most of the results are predictable (aside from a few big blunders).  So with the McMahon system, this first round is basically presumed and the upper half is given a point to begin with.  Tournament play then basically begins in the second round.  The difference with McMahon and "acceleration" is that the acceleration point is later taken away, while in the McMahon it's kept.  I don't think giving and then taking away acceleration points does anything really useful.  The San Diego Chess Club has used McMahon effectively for years and I think it's the standard system for the Go tournaments in Europe.  I'm surprised it's not more frequently used or recommended for chess.

If you have N players and you want first k places to be fair you need sqrt(N+2k) rounds for that. Another formula is log2(N)+log2(k), where both logs are rounded to the nearest integer.

Do not forget tiebreaker. Else those last a long time i mean a long long time.

2 players: 1 round
3-4 players: 2 rounds
5-8 players: 3 rounds
9-16 players: 4 rounds
17-32 players: 5 rounds
33-64 players: 6 rounds
65-128 players: 7 rounds
129-256 players: 8 rounds
257-512 players: 9 rounds
513-1024 players: 10 rounds
1025-2048 players: 11 rounds

Not enough rounds leaves a tie for first place.
Too many rounds may lead to impossible pairings.
An odd number of rounds is preferred to maintain white/black balance for all.

a tie for first is still possible

impossible pairings arent really a problem with like 100 people or more

and an odd number of rounds guarantees everyone has unequal amounts of white/black

#30
A tie is still possible with the right # of rounds, but with too few rounds a tie is expected.
With 100 people and 11 rounds you are sure to get into problems.
With 9 rounds everybody gets 4 or 5 white.
With 10 rounds many get 4 white and 6 black.

QUOTE:
It may be said, then, that most Swiss tournaments are divided into two stages, as follows:

1. The first stage is the "knockout" period ... If the number of rounds is too small for the size of the field, the first stage may never be completed.

2. In the second stage, the "winner" of the first stage begins to meet the strong contenders he (sic) missed on the way up, and the real competition for the title progresses until the tournament ends. ...

The accuracy of the final standings depends on the duration of the second stage of the tournament. ... How long should the second stage last? ... For all practical purposes, therefore, it may be said that the second stage should be "as long as possible". If ideal conditions can be arranged, the total number of rounds for both stages should be about half the number of contestants. Fairly good results can be obtained if the players-to-rounds ratio is not more than 4 to 1.

--Kenneth Harkness (1956) The Official Blue Book and Encyclopedia of Chess, pages 176-177

#32
That was 1956. We now have elo ratings and computer pairings for swiss tournaments.

@tygxc - I don't understand your point.

First, Harkness invented the use of ratings for sorting the scoregroups when doing pairings. In fact he advocated using tournament performance rating for sorting the later rounds, so he was quite sophisticated in his understanding of all the implications.

Second, computerized pairings use an algorithm that replicates the same pairings an arbiter would make by following the rulebook. If the computer does otherwise then it fails to meet requirements. The only advantages to using a computer are speed and accuracy. In other respects computer pairings are in no way "better" than human pairings.

#34
Before computer pairings there were not that many organisers capable of timely producing correct Swiss pairings.

Before elo ratings first round Swiss pairings were random, nowadays pairings are accelerated based on ratings. With 2N players e.g. player 1 plays N+1 with white, player 2 plays N+2 with black, player 3 plays N+3 with white, player 4 plays N+4 with black. This avoids many problems.

You are absolutely correct!

The thing about Swiss system is that the main aim is to let everybody play, gain experience instead of finding the strongest  X players . It is not a good format for a qualification tournament.

Players = 2^rounds finds only the strongest and weakest player. So a 5 round 32 player Swiss will find only the strongest and weakest. The placement of the rest of 30 players is close to random. 

The second stage that @yetanotheraoc mentioned is important but you rarely see a second stage that is as long as the knockout stage. Frankly I have never seen it although know about it.

Another formula is to add 2 times extra placements you want to find out. Bozidar Kazic mentioned this IIRC. So to find top and bottom 3 players in a 32 tournament you need 5 (knockout phase) + 2 x (3-1) = 9 rounds. The second stage is 4 rounds which is close to knockout stage.

So Swiss system is a nice system to let everyone play the game and hopefully enjoy the game but do not expect it to find the strongest X number of players in it. For that nothing beats a double round-robin or better yet quadruple round-robin

I can schedule single or double round robins: rrpair.php but even that isn't perfect. Someone who is (far) out of the running late in the event might not be fighting as hard as would have been the case early.

Having run 4 round Swiss systems at my chess club, I can attest to what happens when the number of players is like 10 or 12. You start pairing by matching the top board with his most likely opponent and then maybe the next and then you see that everybody else has played everybody else. So you end up pairing someone with three points against someone with half a point or some such nonsense, to allow the lower boards to shift. Then there is the problem of color allocation...A total mess.

This is a super point. There are different kinds of tournaments and different reasons for tournaments. We can view a swiss as a modified knockout where nobody gets knocked out. Everybody gets to play all the rounds, even the last place finisher.

For this scenario, Harkness recommended pairing the top score first, then pairing the bottom score group second, then back to the top, then the bottom, and so forth. Also in the top half you give the higher ranked player their due color, but in the bottom half you give the lower ranked player their due color. This way the problem colors get pushed to the middle of the field. The middle score group gets paired last. Since this is also the largest score group, you have the most flexibility there to swap players around.