Author: Howard Exner
Date: 11:22:05 07/19/98
Go up one level in this thread
On July 19, 1998 at 11:38:49, Don Dailey wrote: >On July 18, 1998 at 20:28:44, Howard Exner wrote: > >>On July 18, 1998 at 03:52:14, blass uri wrote: >> >>> >>>On July 17, 1998 at 14:02:14, Howard Exner wrote: >>> >>> >>>>Here's what I'm getting in this position: >>>>1. Rxe6 fxe6 2. Bh6 a6 3. Bxa6 Qc7 4. Nxe6 Qc6 5. Bxb7 Qxb7 6. Re1 Qc6 >>>>7. Qf3 Ne5 8. Ng7+ Kd8 9. Qe4 Bf6 10. Nc5 >>> >>>after 7.Qf3 my programs play N7f6 >> >>This move, N7f6 is better than what I proposed. I realize that most the >>computer programs will consider this quite hopeless for white now. But >>just to satisfy my curiousity do you mind if we continue on a bit. >>Does the next set of moves look reasonable to you. >> >>7 ...N7f6 8.Nac5 Rg8 9. Nd3 >> >>We've come 9 moves from the original position and part of my assessment on this >>requence of moves is pretty well in line with what Bruce Moreland posted a while >>back that he spoke to Im's and one Gm and they said it was a complicated >>position. So Rxe6 does stir up a hornet's nest. But despite the high eval >>computers give to black for the resulting position the position does have >>that element of an uncastled black king. It is well sheilded by by its pieces >>but how will black undertake anything, given its pieces being preoccupied >>with sheilding the king? That's my reasoning for playing on a bit. >> >>After white plays Nac5 how would your programs autoplay this game out? >>Would they manage to keep building on Black's big material advantage? >> >>If the computers are correctly assessing the position now with the >>black king in the middle then I'll join Bob in saying the move, while >>complicating things, is nevertheless unsound. Ditto for agreeing with >>fca who started this thread. >> >>Also has anyone played the game out using Bh6 or Qa3 as the first move? >>After 9 moves of play in these lines what is the eval? That could save us time >>too as the original intent of this thread was to see if Rxe6 was any >>worse than other first choices. If other first choices were "losing" >>also then perhaps the entire example is the wrong choice, as fca >>originally suggested. >> >>Either way the verdict falls I'm enjoying analysing this position. Puzzles >>have always intiqued me and chess seems to have an unlimited supply of them. > >There are a couple of issues. One has do with whether the line is >ultimately sound and the other has to do with its value as an Anti-GM >move. I think the second question is more interesting, the first >is just a interesting puzzle. > >It occurs to me that if you're willing to make your program take more >risks, then the consequences are that it will sometimes lose some >games. That is the whole idea of risk, take a chance to get a chance. This reminds me of how some GM authors have described the play of Tal. That many of his combinational victories contained an unsound choice. The trade off though was that because his imagination was so unique,the ensuing positions were so unexpected and complicated that not many could survive. As if he were saying "welcome to my world". He too said of his play that he tossed his pieces in the direction of the king and trusted that something would develop. These comments I always found to be refreshingly honest in compared to "yea, I saw the 30 mover clearly from start to finish". >I am assuming that is a big part of anti-GM play, giving the program >some judgement or bias toward risk but I don't really know what Ed >has is mind here. It doesn't always have to be extra risk. That's my thinking or guess on what Ed is attempting also - to get positions that compliment the computers strength in open, tactical positions. Even Anand said that computers are better in some areas than anyone else. It would be nice in a follow up interview for someone to explore in more detail his thoughts on computers. > >I would like my program to a little more opportunistic about when >to sacrafice material. Unless I'm extremely conservative about >this I am assuming that when I have it about right, it will error >on both sides occasionally, sac unsoundly and not sac when the >opportunity exists. I would hope that even the unsound sacrafices >don't all lead to a loss against imperfect opponents. > >I think it is possible that a good anti-GM implementation could >make your program play more moves that are technically losing >but also cause it to get better results in practical play. It's >also possible that you implement Anti-GM strategies without >weakening the program at all. > >- Don
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.