Author: Robert Hyatt
Date: 20:41:35 06/08/99
Go up one level in this thread
On June 08, 1999 at 23:09:07, Dann Corbit wrote: >The problem with accelerated pairings is that it assumes you know the ELO of the >opponents. In this case, for each and every entrant, it is not known. >Therefore, the judges choice of which programs are "top programs" and which are >not is not only arbitrary, but detrimental. > >If the contest were for SSDF programs on SSDF hardware, it would make sense. >For this type of contest, it clearly shows that the tournament director does not >understand the mathematics behind it. Period. This is one of _two_ problems. problem 1 is the previously mentioned one in that if seeding is right, by round 7 the top two programs have already played and are 2 rounds beyond that 'final' game. Problem _two_ is exactly as you mention. Seeding is arbitrary, and I have even seen the utterly stupid idea of seeding then flipping a coin to decide whether to reverse 1 and 2, then 3 and 4, before pairings. Nonsense, utter nonsense, because being the 1 seed is a significant advantage, particularly if you have the better program, because you get better pairings and better tie-break. And if you are 'close' to being the best, being seeded 1 is even better, because in the case of a tie, _you_ will likely win on tiebreak. The best way for this event is to simply do a normal Swiss pairing. Because I'll bet that a program in the bottom half of the seedings finishes in the top four. and if that happens, the entire concept is flawed.
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.