Author: Robert Hyatt
Date: 10:28:18 08/26/05
Go up one level in this thread
On August 26, 2005 at 13:05:15, Gian-Carlo Pascutto wrote: >On August 26, 2005 at 12:48:30, Robert Hyatt wrote: > >>On August 26, 2005 at 09:04:06, Gian-Carlo Pascutto wrote: >> >>>On August 26, 2005 at 08:36:40, Bernhard Bauer wrote: >>> >>>>Here is what I get from Crafty19.20 for your second position. >>>>A zugzwang position in a real game! >>> >>>> Rb8 >>>> 8 0.34 -1 1. Rb1 >>>> 8 0.36 -3.53 1. Rb1 Rxd3 2. Qxd3 Qg4 3. Rb8+ Kc7 >>>> 4. Qc4+ Bc5 5. Rc8+ Kxc8 6. Qg8+ Kb7 >>>> 7. Qxg7+ >>> >>>Some bug here?? The analysis is bogus, the line given is an easy draw, not a win >>>for black. >> >>I suggest you look again. This is _not_ an "easy" draw. This is an easy mate >>in 3 for black... >> >>Not sure who has the bug, but it isn't me. :) >> >>I spotted this easily as a human, so even though Crafty doesn't see it on the >>end of that shallow search, -3.5 is certainly good enough to convince it that >>black is better. _much_ better.. >> >>BTW, after Qxg7, Kb6 and white can't avoid Qxe4#. He can put it off by a spite >>check tossing his queen, but that's it... > >ROTFL > >It's a stalemate. Those are draws where I live. The *goal* is exactly to get rid >of all pieces "with spite checks tossing the queen". Note that it isn't a stalemate if you follow the entire PV. And that is what the thing evaluated... If the search ends on a ply where black or white has no legal moves, then it isn't going to see that. But another ply deeper (next iteration) will reveal the fact. Checks in q-search won't solve that. What if there _are_ no checks, you just reach the q-search stalemated? You return the static eval, which is wrong, and hopefully fix this the next iteration where you will still have a ply of real search to do after that position is reached... > >The last move in Crafty's variation is wrong, white can just play moves like >Qc8+ etc... > >As I said, I forgot Crafty can't understand that because of how you do qsearch. > >-- >GCP You miss the point. By not doing checks in the q-search, I get far deeper more quickly. So I don't have to live with a .3 second 8 ply search, I would have gotten a 17 ply search here instead...
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.