Author: Mike Byrne
Date: 20:46:16 02/22/03
Go up one level in this thread
On February 22, 2003 at 17:27:17, Uri Blass wrote: >On February 22, 2003 at 16:58:12, Robert Hyatt wrote: > >>On February 22, 2003 at 09:29:47, Rolf Tueschen wrote: >> >>>On February 22, 2003 at 09:25:16, Robert Hyatt wrote: >>> >>>>On February 22, 2003 at 08:54:30, Rolf Tueschen wrote: >>>> >>>>>I read 6: Ikarus 3.5 / 6 2b= 13w+ 1w= 4b= 3w- 14b+. >>>>> >>>>>So that means that Ikarus although playing the placed 1, 2, 3, 4 progs, it could >>>>>get full points against the last and pre-last. Placed om 14 and 13. >>>>>Is this ok? Something seems to be wrong or biased. Point is that a game against >>>>>14 is a SURE win. That is as if a top program after a loss or two draws got a >>>>>point for free. Note Ikarus had 2,5 pts before playing Matador with 0.5 pts. >>>>> >>>>>Could some expert explain why such things are still possible? >>>>> >>>>>Rolf Tueschen >>>> >>>> >>>>Too few programs, too many rounds. >>> >>>Also, many progs get a "good" result and that is also a good side-effect we >>>should consider. What would be the optimal number of rounds for 14 participants? >>> >>>Rolf Tueschen >> >>optimal number of rounds is log2(N) rounded up to the next integer value. >>In this case 4 rounds, is optimal, 5 acceptable. > >I do not agree. > >more rounds give bigger chance that the winner is not by luck. >I do not agree with a theory that claims that less rounds is better. > >Uri The issue becomes one of fairness. The ideal situation is a Round Robin. Anything less than that has an element of luck -- that is not all partipants will play all other particpants. So who one plays does an have an element of chance in a swiss. Why people such as myself favor the log2(N) rounded up to the next integer value and at most log2(N)+1 is that by the end of the tournament all the top players would have played each other and they would have played each other either in the last or second to last round -- so aesthetically it has the most natural built in suspense. Once you get beyond the log2(N)+1 round, the top programs in the tournament are no longer playing each other, they are playing weaker opponents -- and the farther you get from the log2(N)+1 round the weaker the opponents the top players are playing and sometimes one opponent is much weaker than what another participant is playing (it’s a forgone conclusion that the stronger team will win -- ) -- thus you may have one participant play a substantially weaker opposition than another opponent, secondly - any of the natural suspense and excitement that a tournament would obtain with the top participants playing in the last round for the championship is not there - because most of the time, the top players would have played against each other in a far earlier round. So for my $.02, I would take log2(N) rounded up to the next integer value and at most log2(N)+1 over any tournament that is log2(N)+2 rounded up any day. I think this is of personal preference as opposed to which is correct. The truly only correct ( Correct is this sense meaning totally fair) tournament is a double RR with each side getting white and black and playing all opponents.
This page took 0 seconds to execute
Last modified: Thu, 15 Apr 21 08:11:13 -0700
Current Computer Chess Club Forums at Talkchess. This site by Sean Mintz.