Author: Dave Gomboc
Date: 08:02:25 06/20/99
Go up one level in this thread
On June 20, 1999 at 09:39:59, Robert Hyatt wrote: >On June 19, 1999 at 22:39:55, Roger D Davis wrote: > >>Seems that the Buchholz valuation is really an index of how tough your opponents >>were, then it seems that to the extent that the index is not correlated with the >>ranking (the winner should have the highest, right?), the programs weren't >>seeded correctly, and that extra rounds should be run until the valuation and >>the rankings can be brought to a certain level of agreement. >> >>Yes? No? Insane? >> >>Roger > > > > >the seedings were definitely done wrong. IE Rebel was seeded 3rd based on >the results of the Anand match. Which had nothing to do with anything at all >since the WCCC was 40/2 and rebel only managed 1 draw against Anand at that >time control. And then Shredder was seeded low even after winning the 1996 >WMCCC event and playing well ever since. Small sample, but scoring 0.5/2 against Anand, and considering the quality of the two games, it was a reasonable decision. Shredder was, in hindsight, obviously underranked. >So the sum-of-opponents-scores is pretty random... SOS is not that random unless players who you are comparing pretty much won everything except drew each other. If Ferret had drawn Hiarcs and drawn another higher program instead of losing to Hiarcs and beating SOS then we might have seen it have an equal or better tiebreak. IMO, tiebreak is a sucky way to resolve this stuff in general. No reason why they can't play g/30, then g/5 to decide. Dave
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.