Author: blass uri
Date: 08:50:17 07/13/98
Go up one level in this thread
On July 13, 1998 at 09:54:54, Robert Hyatt wrote: >On July 13, 1998 at 03:40:28, Peter McKenzie wrote: > >>>>Do you really think the Computer Chess World would be best served by >>>>a 16 player tournament? This seems like a dated concept - especially >>>>to the *many* programmers who would likely be excluded :-). >>>>Surely the aim of such a tournament is not only to establish the champion >>>>program/machine combination, but to stimulate activity in the field. >>>> >>> >>> >>>my response would be that it *is* the WCCC. Can you or I get into the >>>human world chess championship cycle? No chance. The WMCCC is ideal to let >>>everyone in. For the WCCC I'd much prefer to see a reduced field with fewer >>>rounds to make large machines possible again. >> >>Comparisons with the world human championship are of limited use. >>If you're going to start comparisons with the human world chess championship, >>then why don't you start asking why we only have a 5 round swiss? >>If you proposed a 5 round swiss between the top 16 players for the next >>world human championship, you would not be taken seriously. >> > >Just for fun, I'd like to see a world championship event (human) where the >winner is decided by some 5 minute blitz games if the regular (Swiss) event >ends in a tie. Think that would also not be taken seriously? :) > >(hint: look at the last WCC event won by Karpov, and look at how many rounds >he played at a regular time control, and then what happened after that was a >draw.) :) I do not take seriously the last WCC event I see kasparov as the world champion. > > > > >>>>The best solution is to make it a large tournament, with a sufficiently >>>>large number of rounds (eg. 9) to make it a decent tournament. A >>>>slightly reduced time control is a very small price to pay for the >>>>advantages of this format. >>>> >>> >>> >>> >>>the math for 9 rounds is hopeless... IE there are two good ways to run >>>a tournament, as the "humans" have found out: >>> >>>1. a swiss where rounds <= log2(players). >>> >>>2. a round-robin >> >>I dispute this. I've played in many good tournaments where the above >>does not hold. > >can you give an example cross-table? I've *never* played in a swiss, human >or computer, where log2(entries) was way less than the number of rounds >actually played. There is definite math involved in choosing the right number >of rounds for a Swiss to work properly. Too few and you end up with ties in >all the top positions. Too many and the same thing happens... > > > >> >>> >>>a swiss with rounds > log2(players) is a waste of time after a couple of >>>extra rounds. All you are doing is just playing games, because you have >>>already seen the best 3-4 programs play each other by the time you get to >>>log2(players). If you make the stupid mistake Jaap made at the last WMCCC >>>and use accelerated pairings, you make this worse, and not better. But >>>simply stated, too many rounds is no better than too few, unless too many >>>becomes a round robin... >> >>Surely more rounds adds greater statistical confidence to the results, even >>in a swiss. > > >Actually, for the winner, maybe. But it also adds random chance due to the >pairings. Which is why a round-robin is the fairest scheme as pairings don't >affect the outcome with everyone playing everyone. The last WMCCC event shows >what can happen however, in that all the top programs had played by round 5 or >6, so that the only question left was would a top-rated program make that one >bad move and lose against a lower rated program or not? At every WCCC and ACM >event I've attended, with one exception in 1984, the last round determined the >champion. I am against this. I do not like the idea of determining the champion by 1 game I want to know that the champion is the best. Uri
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.