Computer Chess Club Archives


Search

Terms

Messages

Subject: nunn match fritz5 - Junior4.6(1000 seconds per move).

Author: blass uri

Date: 11:15:15 05/15/98


The result in the nunn match is now 5:1 for fritz5
the game I thought junior win was a draw
and I post this game(game 5 fritz5 white in the 3rd nunn position)
1.e4 e6
2.d4 d5
3.Nc3 Bb4
4.e5 c5
5.a3 Bxc3+
6.bxc3 Qc7
7.Nf3 Ne7
8.a4 b6
9.Bb5+ Bd7
10.Bd3 Nbc6
11.0-0 (the first nunn position)
11...c4 0.00/17
12.Be2 -0.22/15   Na5
13.Ba3 Nf5
14.Qd2 Qd8
15.Bb4 Nb7
16.a5 Nxa5
17.Bxa5 bxa5
18.Rfb1 0-0
19.g4 Nh6 -1.11/19
20.Qf4 -1.03/14 a4
21.Qg3 Rb8
22.Nd2 Rxb1+ -1.64/18
23.Rxb1 -1.50/13 a3
24.Ra1 Qa5
25.Ra2 Rb8 -2.07/18
26.Qf3 Be8
27.Bd1 Rb2
28.Ra1 a2
29.Qe3 Ba4
30.h3 -3.00/14      Kh8 -3.39/19
after Isaw these evaluations I felt sure to post a win to black
but the game ended in draw
31.Nf1 -3.09/14    Rb1 -3.78/19
32.Rxa2 -2.69/17 Rxd1 -3.77/18
33.Kg2 Qb5
black has a bishop adventage but both programs do not
understand its king is in danger.
34.Ng3 Rg1+
35.Kh2 Rb1
36.Qf4 Rd1
37.Kg2 a5
38.g5 Ng8
39.Qxf7 Qe8
40.Qc7 Bd7 -1.68/17
41.Rb2 a4
42.Qa7 Ne7
43.Rb7 Nc8
44.Qa5 Kg8
45.h4 Rd2
46.h5 -2.47/14     Kf8 -2.39/16
47.Qb4+ Ne7
48.Qd6 Bc8
49.Ra7 -1.94/16 g6 -2.20/18 (fritz was expecting Rxc2)
junior rejected Rxc2 because of Nf5 but did not devote enought
time to this move after g6(a bug of Junior)

50.Nf5 exf5(gxf5 even losing)
51.Qf6+ Qf7
52.Qh8+ Qg8
53.Qf6+ and draw with perpetual check

game 3(nunn position 2 fritz white) was a long draw of 135 moves
fritz5 got an adventage  in a queen ending
(there were also bishops most of the time)
when fritz's pawn was at a7.

Junior evaluate the ending most of the time as equality
and did not understand that Black is the only one who can lose
(maybe because the white queen most of the
time did not control a8 (after white queen went to b8 both programs
increased fritz5 advantage and fritz even
evaluated its position as almost 3 pawns for white
and it is not logical because it was clear that Junior has a
chance of draw because of the queens)

Junior did a draw but for one of its move it found the right
move only after almost an hour if I believe its evaluation.

I suggest that there will be a rule in the evaluation function
that if there is a promoted passed pawn in a queen ending
it decreases the value of all the other pawns.

Junior should be better in evaluating unclear positions
for example if Junior see position that one pawn is in d7 and threat to
be a queen and the other side has a big material adventage
(less than a queen)
the static evaluation should be near equality
and the computer should analyse it if it can
(sometimes it cannot because there are a lot of checks)

maybe ideas like this will cause Junior to see less nodes
per second but I believe it is better to use them
because otherwise Junior can be wrong in its evaluation
function for a long time

I hope the next version of Junior will be better at evaluating
king's danger
Junior's main problem in the nunn match is evaluating
king's danger.



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.