Ok, maybe the dumbest question ever, and not directly chess-related
(but who knows swiss system better than chess players?), and written in
bad English... But I'll be grateful if someone will answer.

I have a tournament that *must* have 38 rounds, and I expect more than
38 players (but less than 2^38...). Let's say 50 players. I'd like to
use the swiss system. But, with the normal rules (shared by all the
variants), in the last round the player in position 1 will go against a
player in position 30 or similar: a match not so significant. I'd like
to have matches that are more significant, even at the end of the
tournament.

The questions a
- Could I allow two players playing against each other *more than
once*? which are the downsides?
- And what if I'd allow any number of repetitions (i.e. the first ever
against the second, even if they've played against each other three
times)?

writes:

Ok, maybe the dumbest question ever, and not directly chess-related
(but who knows swiss system better than chess players?), and written in
bad English... But I'll be grateful if someone will answer.

I have a tournament that *must* have 38 rounds, and I expect more than
38 players (but less than 2^38...). Let's say 50 players. I'd like to
use the swiss system. But, with the normal rules (shared by all the
variants), in the last round the player in position 1 will go against a
player in position 30 or similar: a match not so significant. I'd like
to have matches that are more significant, even at the end of the
tournament.

The questions a
- Could I allow two players playing against each other *more than
once*? which are the downsides?

You *could* - but you can avoid it.

- And what if I'd allow any number of repetitions (i.e. the first ever
against the second, even if they've played against each other three
times)?


If I were running a 38-round event, with 50 players, I would use a
RR-guided Swiss. This is explained in detail in the USCF rulebook, but
here are the basics:

a) find a Round Robin table for a 50 player event
b) pair the first round normally, and assign pairing numbers AFTER you
pair. Assign pairing numbers so that the pairings for the first round
correspond to the pairings in the first round of the 50-player Round
Robin table.
c) on subsequent rounds, pair the top score group to determine the
correct pairing for the top-ranked player. Then find that pairing in
the 50-player RR table. Play that round, as specified in the table
(except that you should balance colors).

You will end up playing 38 of the rounds listed in the 50-player RR
table (so...no one will play the same player twice) - but you will use
Swiss system ideas to select which rounds to actually play. Pairings
for the top players will be very much like normal Swiss pairings - but
pairings for the lower ranked players will be a bit strange. One
advantage is that you will get many "strange" pairings early on - but
far fewer impossible pairing problems in later rounds.

The hard part will be finding a 50-player RR

Frankly, it would also work (perhaps better, and certainly easier) to
simply assign pairing numbers by lot and play the first 38 lines in the
RR-table. I have often done this (on a much smaller scale) when I run
Quads and the bottom section is a Hex and we only have time for 3
rounds. Rather than go through the effort of doing Swiss pairings, or
even RR-guided Swiss, I simply post the first 3 lines from the 6-player
RR table. If we have time for a 4th round, we just play line 4, too.
You might do the same thing by pretending to run a 50-player (or however
many show up) RR, and then stop after "only" 38 rounds. A major
advantage of doing it this way is that you can post the pairings for all
38 rounds at the beginning of the event, and not take time for pairing
in between rounds.


--
Kenneth Sloan
Computer and Information Sciences (205) 934-2213
University of Alabama at Birmingham FAX (205) 934-5473
Birmingham, AL 35294-1170 http://www.cis.uab.edu/sloan/

If I were running a 38-round event, with 50 players, I would use a
RR-guided Swiss.

Thanks for the answer! I din't know this system. It seems clever. But..

Pairings
for the top players will be very much like normal Swiss pairings - but
pairings for the lower ranked players will be a bit strange.

(Probably another dumb question) Doesn't we have just *one* correct
pairing (the top player's match) and twenty-four random ones?

The hard part will be finding a 50-player RR

I've found an easy way to build one in Wikipedia:
http://en.wikipedia.org/wiki/Round-r...ling_Algorithm

Frankly, it would also work (perhaps better, and certainly easier) to
simply assign pairing numbers by lot and play the first 38 lines in the
RR-table.

Sure. But I'd like to have: (1) every player, every match, playing
against a near opponent, so, if he'll win, he'll go up some positions,
and, if he'll lose, he'll go down; (2) not every time the same pairings
(but I'll accept, let's say, that two players meet 4 times).

Maybe a solution is to use the swiss system, dividing the tournament in
mini-tournaments:
- In every mini-tournament, two players can't fight each other more
than once; but, when a new tournament begins, I reset this counter.
- For the first round of the second (third, fourth...) mini-tournament
begins, the players are paired according to the points they conquered
in the previous mini-tournaments.
- For the next rounds, I have two choices: I can pair the players
either (1) according to the total points thay gained in the entire
tournament or (2) according to the points they gained in this
mini-tournament (in this case, players in better form are paired one
against the other).
- The number of rounds in every mini-tournament (except the last...)
could be the square root of the number players (i.e. every
mini-tournament would be good for a normal swiss system).

Kenneth Sloan
http://www.cis.uab.edu/sloan/

It seems that you study interesting topics!

- The number of rounds in every mini-tournament (except the last...)
could be the square root of the number players (i.e. every
mini-tournament would be good for a normal swiss system).

Bad error... not the square root! I mean that the number of players is
2^r (r = number of rounds).

writes:

....
Pairings
for the top players will be very much like normal Swiss pairings - but
pairings for the lower ranked players will be a bit strange.

(Probably another dumb question) Doesn't we have just *one* correct
pairing (the top player's match) and twenty-four random ones?

"Correct" is a slippery term. If you want to run an event which can't
be done as a normal Swiss, and can't be done as a normal Round Robin,
then you have to make adjustments.

One common error made by TDs is to assume that the standard procedures
for a normal Swiss are, for some reason, the *definition* of "proper
pairings". Not at all. You need to consider the goals of your event,
and pair to meet those goals. If you have 50 players and 38 rounds,
normal Swiss pairings will *not* get the job done. Things will look OK
for awhile, but then you will be painted into a corner and have great
difficulty making valid pairings.

One (usually) important criterion is: no one plays the same player
twice. Using the RR table assures that this criterion will be met.

A second (Swiss) criterion is that players should always play opponents
with the same score in the event, so far. Well, round robin tournaments
violate that criterion all the time - because they *must*. A 50-player,
38-round event is more like a Round Robin than a Swiss.

Not all events should be Swiss events!



The hard part will be finding a 50-player RR

I've found an easy way to build one in Wikipedia:
http://en.wikipedia.org/wiki/Round-r...ling_Algorithm

Frankly, it would also work (perhaps better, and certainly easier) to
simply assign pairing numbers by lot and play the first 38 lines in the
RR-table.

Sure. But I'd like to have: (1) every player, every match, playing
against a near opponent, so, if he'll win, he'll go up some positions,
and, if he'll lose, he'll go down; (2) not every time the same pairings
(but I'll accept, let's say, that two players meet 4 times).

Can't be done. Your only real choice is whether the "bad" pairings
(from a Swiss point of view) happen all clumped together at the end (and
possibly force repeat pairings) or they happen uniformlly scattered
throughout the event.

Also, note that Swiss pairings do not guarantee matches between players
of near *strength*. In fact, Swiss pairings *resist* pairing players of
the same strength until the few rounds.




Maybe a solution is to use the swiss system, dividing the tournament in
mini-tournaments:

Well, yes. Now you are talking about smaller events


- In every mini-tournament, two players can't fight each other more
than once; but, when a new tournament begins, I reset this counter.
- For the first round of the second (third, fourth...) mini-tournament
begins, the players are paired according to the points they conquered
in the previous mini-tournaments.
- For the next rounds, I have two choices: I can pair the players
either (1) according to the total points thay gained in the entire
tournament or (2) according to the points they gained in this
mini-tournament (in this case, players in better form are paired one
against the other).
- The number of rounds in every mini-tournament (except the last...)
could be the square root of the number players (i.e. every
mini-tournament would be good for a normal swiss system).

But...you haven't said how you'll deal with repeat pairings.

You started by saying you wanted a 50-player, 38-round event, paired by
Swiss rules, with no repeat pairings. My answer is you can't do that -
you must try something else.

You need to decide the *order* of your constraints. I suggest that
"Swiss pairings" can be fairly low down on the list.

Any organizer can invent novel structures, so just about any structure
of mini-events *can* be done.

So...please order these criteria:

a) tournament uses Swiss pairings, by the book
b) tournament uses an accepted structure, in the book
c) players play opponents of the same strength
d) players play opponents with the same score
e) players play a given opponent only once
f) critical games are played in the later rounds

The point is: you must compromise on one or more of these goals. You
must decide what's most important first - only then can you design your
event.

--
Kenneth Sloan
Computer and Information Sciences (205) 934-2213
University of Alabama at Birmingham FAX (205) 934-5473
Birmingham, AL 35294-1170 http://www.cis.uab.edu/sloan/

If you want to run an event which can't
be done as a normal Swiss, and can't be done as a normal Round Robin,
then you have to make adjustments.

One common error made by TDs is to assume that the standard procedures
for a normal Swiss are, for some reason, the *definition* of "proper
pairings". Not at all. You need to consider the goals of your event,
and pair to meet those goals.

Two clever considerations.

One (usually) important criterion is: no one plays the same player
twice.

I know. In fact, my initial answer was: it's important, but *why* is it
important? what happens if one *can* play the same player twice?

Also, note that Swiss pairings do not guarantee matches between players
of near *strength*. In fact, Swiss pairings *resist* pairing players of
the same strength until the few rounds.

I know. In fact, I didn't write about balanced matches, but about
meaningful matches: I'd like every player playing every match with a
strong motivation, i.e. to go up or down one position (or more).

Maybe a solution is to use the swiss system, dividing the tournament in
mini-tournaments:

Well, yes. Now you are talking about smaller events

And what if there's a rule like "if A and B play against at round x,
they'll can't play against until round x+k"?

- In every mini-tournament, two players can't fight each other more
than once; but, when a new tournament begins, I reset this counter.

But...you haven't said how you'll deal with repeat pairings.

Maybe I wasn't clear: a pairing can't be repeated in the same
mini-tournament, but can be repeated in different mini-tournaments (or,
maybe, I haven't understood your question).

You started by saying you wanted a 50-player, 38-round event, paired by
Swiss rules, with no repeat pairings. My answer is you can't do that -
you must try something else.

Again, I wasn't clear: I want 50 players, 38 rounds, with the Swiss
principle of pairing players with similar points. And I asked: is it
necessary to use the other Swiss principle, "no repeat pairings"? and,
if yes, why?

In every case, we agree: I must find another formula - nor a pure
round-robin, neither a pure Swiss -.

So...please order these criteria:

Ordered, they would be:
d) players play opponents with the same score
f) critical games are played in the later rounds
c) players play opponents of the same strength
b) tournament uses an accepted structure, in the book
a) tournament uses Swiss pairings, by the book
e) players play a given opponent only once

The point is: you must compromise on one or more of these goals. You
must decide what's most important first - only then can you design your
event.

Thank you for the attention, for the answers, for the questions, for
the suggestions.

On Sun, 20 Aug 2006 02:55:48 -0500, Kenneth Sloan
wrote:

The hard part will be finding a 50-player RR

That's not hard. Consider a long table with 25 boards. Player #1
sits at board #1. Initially #2 is opposite him. #3 is next to #2,
then #4, ... #26 on the end, then it wraps around with #27 opposite
#26, then 28, ... and 50 is on board 2 next to #1. Player #1 stays
stationary. On each round, #2 takes #50's place, #3 takes #2's place,
#4 takes #3's place, etc.
---
Replace you know what by j to email

On 20 Aug 2006 00:29:59 -0700, wrote:

I have a tournament that *must* have 38 rounds,

It must have exactly 38 round, no more, no less? Can you explain why?
---
Replace you know what by j to email

Consider a long table with 25 boards.

The same way explained in the Wikipedia article I cited. An elegant
idea.

I have a tournament that *must* have 38 rounds,

It must have exactly 38 round, no more, no less? Can you explain why?

Su it's a fantasy soccer tournament, and Italian soccer tournament
have 38 rounds.

On 20 Aug 2006 23:09:37 -0700, wrote:

Su it's a fantasy soccer tournament, and Italian soccer tournament
have 38 rounds.

Is there any limit to the number of games that can be played during
one round?
---
Replace you know what by j to email