Author: Vine Smith
Date: 17:33:23 03/29/02
Go up one level in this thread
On March 29, 2002 at 15:35:19, Uri Blass wrote: >On March 29, 2002 at 15:13:10, Vine Smith wrote: > >>On March 29, 2002 at 12:43:03, Uri Blass wrote: >> >>>On March 29, 2002 at 12:38:02, Shay Bushinsky wrote: >>> >>>>I agree with your scheme and in fact this is the deal for the next competition >>>>with a member of the 2700 Elo club >>> >>>I believe that it is going to do the match more interesting but it does not mean >>>that it is going to encourage the GM to do the best. >>> >>>Uri >> >>If paying GMs based on results won't work, what will? Perhaps the old expedient >>of life-or-death stakes might do the trick? Or would that make them too >>nervous...maybe just nasty electrical shocks administered after each loss. >> >>Regards, >>Vine Smith > >see my explanation in http://www.talkchess.com/forums/1/message.html?220390 > >The problem that I see is simply that they may take unnecessary risks when the >money is not proprtional to the result. > >It may do the games more interesting because the GM's may try to play more for a >win and not to be careful by playing for a draw and trying to win only if they >get an opportunity to do it with no risks. > >Uri I don't believe GMs evaluate positions in the manner your referenced post suggests. That is, I don't believe any GM looks at a move, and thinks "Hmm...e5 gives me a 30% chance to win, but a 40% chance to lose, while Nc3 gives me a 10% chance to win, yet only a 10% chance to lose." Instead, based on many GM annotations I have read, it is more like, "Hmm...e5 might win, but it could also lose, while after Nc3, I cannot possibly lose, and I even keep some slight winning chances." This is sometimes called "playing with the draw in hand", or "playing for two results" (win or draw, whereas "playing for three results", i.e. playing speculatively, can lead to win, draw, or loss). I don't believe it would matter to a typical GM if the first, dangerous, move yielded 30% winning and 40% losing chances, or the other way around -- the second, safe, move would normally be chosen unless it was a "must win" situation. Whether this prevailing attitude represents the "best" chess GMs have to offer is a matter of some debate. There have been many schemes to dissuade GMs from drawing with each other, and none have seemed to really work. It is also unclear whether "better" chess was played in events where draws had to be replayed or counted only for 0.25 points or whatever other mechanisms have been employed in the past. However, paying only for wins in a GM - computer program match will at least avoid some of the typical stratagems used by the GMs to shirk combat, such as quick liquidations of most of the material on the board, or complete closure of the position followed by "shuffle chess". While this will certainly be to the computer program's advantage, since it does not care at all about the result, and just dutifully follows its search and evaluation routines whatever the terms of the match, it will also certainly make the match more interesting, which is what the sponsor (almost always the program's developer or owner) is paying for. Because nobody is interested in the question of whether GMs can draw against top computer programs (yet), the recent Gulko vs. The Computers match would have been "better" all around if the 3-5 outcome had consisted entirely of decisive games. In this case there would have been 5, not 3, games that the program manufacturers could use as examples when promoting their product, and proponents of the human side would have had 3 games, instead of just 1, to examine the remaining defects of software-generated chess. It does not seem wrong to discourage draws in this context. Regards, Vine Smith
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.