Author: Robert Hyatt
Date: 17:43:48 11/22/00
Go up one level in this thread
On November 22, 2000 at 20:28:50, Andrew Dados wrote: >On November 22, 2000 at 14:06:22, Robert Hyatt wrote: > >>On November 22, 2000 at 12:32:47, Bruce Moreland wrote: >> >>>On November 22, 2000 at 03:55:13, Jeremiah Penery wrote: >>> >>>>On November 22, 2000 at 00:02:34, Bruce Moreland wrote: >>>> >>>>>On November 21, 2000 at 20:24:31, Jeremiah Penery wrote: >>>>> >>>>>>On November 21, 2000 at 20:11:10, Bruce Moreland wrote: >>>>>> >>>>>>>What does Crafty think of the Reti study: >>>>>>> >>>>>>>[D]7K/8/k1P5/7p/8/8/8/8 w - - 0 1 >>>>>>> >>>>>>>White to move draws. I might have this slightly wrong, because I'm constructing >>>>>>>it from memory. >>>>>>> >>>>>>>bruce >>>>>> >>>>>>I get a drawscore at ply 9 from Crafty in this position. >>>>>> >>>>>>This is the last line I got: >>>>>> >>>>>>22-> 1:01 -0.50 1. Kg7 Kb6 2. Kf6 Kxc6 3. Kg5 h4 4. >>>>>> Kxh4 Kc7 5. Kh5 Kb6 6. Kh6 Ka5 7. Kg6 >>>>>> Kb6 8. Kh5 Ka5 9. Kh6 Kb6 10. Kg6 Ka5 >>>>>> 11. Kh5 Kb6 >>>>> >>>>>What interests me would be the scores prior to that ply, where Crafty things >>>>>white is going to hit a home run. >>>>> >>>>>bruce >>> >>>> 7 0.19 -0.50 1. c7 Kb7 2. c8=Q+ Kxc8 3. Kh7 Kb7 >>>> 4. Kh6 Kc6 5. Kxh5 >>>> 7-> 0.21 -0.50 1. c7 Kb7 2. c8=Q+ Kxc8 3. Kh7 Kb7 >>>> 4. Kh6 Kc6 5. Kxh5 >>> >>>>Crafty never thinks white is winning in this position. >>> >>>It should think black is winning, because black's pawn can't be stopped, but >>>white's can. I snipped out a lot of the other lines here, but does anyone else >>>think that the line I kept is completely ridiculous? >>> >>>Bob, what the heck is crafty doing there? >>> >>>bruce >> >>There are several possible issues here. My PV may well be silly on the end. >>When I get a hash hit, with an EXACT score, the PV is obviously cut short at >>that point. Later, I try to probe the hash table to add more moves to the >>end. And on occasion, those moves might not make any sense, and might not >>even be part of the set of moves that leads to the score it gave... >> >>Here is a non-egtb output from crafty (400mhz machine, one cpu): >> >> time surplus 0.00 time limit 30.00 (3:00) >> nss depth time score variation (1) >> 1 0.00 -3.50 1. Kg8 >> 1-> 0.00 -3.50 1. Kg8 >> 2 0.00 -- 1. Kg8 >> 2 0.00 -3.91 1. Kg8 h4 >> 2 0.00 ++ 1. c7!! >> 2 0.00 0.00 1. c7 Kb7 2. c8=Q+ Kxc8 > >[snip] > >Bob.. how can you justify zero score here...? Good question... I am looking... There is one potential explanation for old crafty's... someone sent me a change to the rook pawn + wrong bishop code that worked in every case except one... no bishops and the opponent has enough pawns to win, while it only had a rook pawn that couldn't promote. The change would say "draw" which was wrong. I "thought" this was fixed. In looking at the position, I might have either lost the fix, or else my fix had a hole I didn't notice. More in a little while... OK... my "fix" got lost. Not sure how... probably meant to copy from my notebook (with fix) to my quad xeon, but must have done it backward losing the fix. I will re-do the fix and rerun the test. It is nonsense at present...
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.