Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Crafty-IsiChess,CCT4,r11 ==> A move to avoid?

Author: Thomas Mayer

Date: 17:03:42 01/29/02

Go up one level in this thread


Hi Uri,

>I guess that Fritz and Quark do not use the same null move algorithm

definitely, I think that Fritz still does not do any or much null move
recheck... Quark does it, but according to an idea of my own which is very depth
dependant. For that reason, Quark needs sometimes very much time to discover
null move problems if they are some plys away, but on the other hand my null
move recheck does not cost very much...

In fact I have implemented it, because in Leiden last year Quark lost a game
against Patzer because Quark does not recheck null moves... So it is a good
feeling to have this recheck in it...

>Note that the more interesting position is the position 2 plies before Ne5

And I will show you that Quarks system does work much worser in that position...

[D]8/8/1p1r1k2/p1pPN1p1/P3KnP1/1P6/8/3R4 b - - 0 1

Analysis by Quark 1.68alpha on Athlon 1400:

1...Td6-d8 2.Se5-c4 Td8-b8 3.d5-d6
  ±  (0.90)   Tiefe: 4   00:00:00
1...Sf4-e2 2.Td1-f1+ Kf6-e7 3.Tf1-f7+ Ke7-e8 4.Tf7-h7
  ²  (0.67)   Tiefe: 4   00:00:00
1...Sf4-e2 2.Td1-f1+ Kf6-e7 3.Tf1-f7+ Ke7-e8 4.Tf7-h7
  ²  (0.67)   Tiefe: 4   00:00:00
1...Td6-d8 2.Td1-h1 Kf6-g7 3.Ke4-f5 Td8-f8+ 4.Kf5xg5 Sf4xd5
  ²  (0.30)   Tiefe: 6   00:00:00  65kN
1...Sf4-e2 2.Td1-d3 Se2-f4 3.Td3-d2 Kf6-g7 4.Se5-f3 Kg7-f6
  ²  (0.40)   Tiefe: 7   00:00:00  164kN
1...Sf4-e2 2.Td1-f1+ Kf6-e7 3.Tf1-f7+ Ke7-e8 4.Tf7-f3 Ke8-e7 5.Tf3-h3 Td6-f6
6.Th3-h8
  ²  (0.46)   Tiefe: 8   00:00:00  333kN
1...Sf4-e2 2.Td1-d3 Se2-d4 3.Td3-h3 Kf6-g7 4.Se5-c4 Td6-g6 5.d5-d6 Tg6-e6+
6.Ke4-d5
  ²  (0.56)   Tiefe: 9   00:00:01  570kN
1...Sf4-e2 2.Td1-f1+ Kf6-e7 3.Se5-c4 Se2-g3+ 4.Ke4-e5 Td6xd5+ 5.Ke5xd5 Sg3xf1
6.Kd5-c6 Sf1-h2 7.Kc6xb6 Sh2xg4
  ²  (0.31)   Tiefe: 10   00:00:02  1365kN
1...Sf4-e2 2.Td1-f1+ Kf6-e7 3.Tf1-f7+ Ke7-e8 4.Tf7-h7 Se2-c3+ 5.Ke4-f5 Sc3xd5
6.Kf5xg5 Sd5-f6 7.Th7-h8+ Ke8-e7 8.Kg5-f5 Td6-d1
  ²  (0.38)   Tiefe: 11   00:00:04  2113kN
1...Sf4-e2
  ²  (0.68)   Tiefe: 12   00:00:08  5782kN
1...Sf4-e2 2.Td1-f1+ Kf6-e7 3.Se5-c4 Se2-c3+ 4.Ke4-d3 Sc3xd5 5.Sc4xd6 Ke7xd6
6.Tf1-f5 Sd5-f4+ 7.Kd3-e4 Sf4-e6 8.Tf5-d5+ Kd6-c6 9.Td5-d1
  ²  (0.69)   Tiefe: 12   00:00:10  6481kN
1...Sf4-e2 2.Td1-f1+ Kf6-e7 3.Se5-c4 Se2-c3+ 4.Ke4-d3 Sc3xd5 5.Sc4xd6 Ke7xd6
6.Kd3-e4 Sd5-f4 7.Ke4-f5 Sf4-e6 8.Tf1-d1+ Kd6-e7 9.Kf5-e5
  ±  (0.73)   Tiefe: 13   00:00:18  12404kN
1...Sf4-e2 2.Td1-f1+ Kf6-e7 3.Se5-c4 Se2-c3+ 4.Ke4-d3 Sc3xd5 5.Sc4xd6 Ke7xd6
6.Tf1-f5 Sd5-f4+ 7.Kd3-e4 Sf4-h3 8.Tf5-f6+ Kd6-c7 9.Ke4-f5 Kc7-b7 10.Kf5-e5
  ±  (0.96)   Tiefe: 14   00:00:37  24879kN
1...Sf4xd5
  ±  (0.95)   Tiefe: 14   00:00:51  34999kN
1...Sf4-e2 2.Td1-f1+ Kf6-e7 3.Se5-c4 Se2-c3+ 4.Ke4-d3 Sc3xd5 5.Sc4xd6 Ke7xd6
6.Tf1-f5 Sd5-f4+ 7.Kd3-e4 Sf4-h3 8.Tf5-f6+ Kd6-c7 9.Ke4-f5 Kc7-b7 10.Kf5-e5
  ±  (0.96)   Tiefe: 14   00:00:51  34999kN
1...Sf4-e2 2.Td1-f1+ Kf6-e7 3.Se5-c4 Se2-c3+ 4.Ke4-d3 Sc3xd5 5.Sc4xd6 Ke7xd6
6.Kd3-e4 Kd6-e6 7.Tf1-f5 Sd5-f6+ 8.Ke4-f3 Sf6-h7 9.Kf3-e3 Ke6-d6 10.Tf5-f7
  ±  (0.94)   Tiefe: 15   00:01:21  56103kN
1...Sf4xd5
  ±  (0.93)   Tiefe: 15   00:01:24  58178kN
1...Sf4xd5 2.Td1xd5 Td6-e6 3.Ke4-f3 Te6xe5 4.Td5-d6+ Te5-e6 5.Td6xe6+ Kf6xe6
6.Kf3-e4 Ke6-d7 7.Ke4-d5 Kd7-c7 8.Kd5-e5 Kc7-c6 9.Ke5-f5 c5-c4
  =  (-0.25)   Tiefe: 15   00:03:56  158712kN
1...Sf4xd5
  ³  (-0.55)   Tiefe: 16   00:04:39  187531kN
1...Sf4xd5 2.Td1xd5 Td6-e6 3.Ke4-d3 Te6xe5 4.Td5-d6+ Te5-e6 5.Td6-d5 Kf6-g6
6.Kd3-d2 Kg6-h6 7.Td5-d8 Te6-f6
  µ  (-1.14)   Tiefe: 16   00:06:21  253455kN
1...Sf4xd5 2.Td1xd5 Td6-e6 3.Ke4-d3 Te6xe5 4.Kd3-c4 Te5-e4+ 5.Kc4-c3 Te4-b4
6.Td5-f5+ Kf6-g6 7.Tf5-e5 Kg6-h6 8.Te5-d5 Kh6-g6 9.Td5-e5
  µ  (-0.89)   Tiefe: 17   00:07:16  289337kN

you can see that Sxd5 blinks first time in ply 14 but was for sure refused by
null move... one ply later Quark finds the move... besides, after a hint from
John Stanbeck and some checks by my own, it seems Quarks implementation of null
move rechecks was the problem in the game against Zarkov - together with
adaptive null move it seems to result in the major blunder 43. Rb4+ ... on the
other hand, I believe that it is simply not possible to fully avoid
"Zugzwang"-problems when using null move... and the win of null move is that way
high, that those problems must be accepted somehow...

Greets, Thomas

P.S.: My idea of null move recheck goes back to a hint from Dieter Bürßner with
some modifications by myself



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.