Computer Chess Club Archives


Search

Terms

Messages

Subject: Re: Endgame (again!) test position: Check out Delfi 4.4

Author: Jorge Pichard

Date: 11:36:49 12/04/03

Go up one level in this thread


On December 04, 2003 at 11:03:43, Uri Blass wrote:

>On December 04, 2003 at 10:55:06, Jaime Benito de Valle Ruiz wrote:
>
>>(ALL these tests are done with 3,4 and some 5 tablebases in an Athlon 1900+ with
>>512 Mb)
>>
>>This position was recently posted by Joachin Rang, and no one really suggested
>>why is it so hard for commercial engines:
>>
>>[D]8/5R2/2p5/8/7P/1K2b3/P3kp2/8 w - - 0 1

Delfi  WB2 27 MB:
 2   00:00 -2.16   Kc4 f1=Q. Rxf1 Kxf1.
 3   00:00 -2.16   Kc4 f1=Q. Rxf1 Kxf1.
 4   00:00 -2.16   Kc4 f1=Q. Rxf1 Kxf1.
 5   00:00 -1.99   Kc4 c5. a4 f1=Q. Rxf1 Kxf1.
 6   00:00 -2.20   Kc4 f1=Q. Rxf1 Kxf1. a4 Ke2.
 7   00:00 -1.98   Kc4 f1=Q. Rxf1 Kxf1. a4 Ke2. a5
 8   00:00 -2.07   Kc4 f1=Q. Rxf1 Kxf1. a4 Ke2. a5 c5.
 9   00:00 -1.80   Kc4 f1=Q. Rxf1 Kxf1. a4 Ke2. a5 Kf3. a6
 9   00:00 -1.70   Rxf2+...
 9   00:00 -1.80   Kc4 f1=Q. Rxf1 Kxf1. a4 Ke2. a5 Kf3. a6
10   00:00 -1.97   Kc4 f1=Q. Rxf1 Kxf1. a4 Kf2. a5 Kg3. h5 Kf4.
10   00:00 -1.81   Rxf2+...
10   00:00 -1.70   Rxf2+ Bxf2. Kc4 Kf3. h5 Ke4. h6 Be3. h7 Bd4. a4 Bh8.
10   00:00 -1.60   h5...
10   00:00 -1.70   Rxf2+ Bxf2. Kc4 Kf3. h5 Ke4. h6 Be3. h7 Bd4. a4 Bh8.
11   00:00 -1.48   Rxf2+ Bxf2. Kc4 Kf3. h5 Ke4. h6 Be3. h7 Bd4. a4 Bh8. a5
12   00:01 -1.44   Rxf2+ Bxf2. Kc4 Kf3. h5 Ke4. h6 Be3. h7 Bd4. a4 Bh8. a5 Kf5.
12   00:01 -1.34   h5...
12   00:02 -0.26   h5! f1=Q. Rxf1 Kxf1. Kc4 Ke2. h6 Bxh6. Kc5 Kf3. Kxc6 Ke4.
12   00:02 -0.16   Kc4...!
12   00:02 -0.26   h5! f1=Q. Rxf1 Kxf1. Kc4 Ke2. h6 Bxh6. Kc5 Kf3. Kxc6 Ke4.
13   00:03 -0.15   h5 f1=Q. Rxf1 Kxf1. Kc4 Ke2. h6 Bxh6. Kc5 Be3+. Kxc6 Kd3. a4
Kd4.
14   00:04 -0.16   h5 f1=Q. Rxf1 Kxf1. Kc4 Ke2. h6 Bxh6. Kc5 Kd3. Kxc6 Bf4. a4
Kd4.
15   00:19  0.38   h5! f1=Q. Rxf1 Kxf1. a4 Bd2. Kc4 Ke2. Kc5 Kd3. Kxc6 Kc4. a5
Bxa5. h6
16   00:43  0.63   h5 f1=Q. Rxf1 Kxf1. a4 Bd2. Kc4 Ke2. a5 Bxa5. h6 Bc7. Kc5
Kd3. h7 Be5. Kxc6
17   01:01  0.49   h5 f1=Q. Rxf1 Kxf1. a4 Bd2. Kc4 Ke2. a5 Bxa5. h6 Bc7. Kc5
Kd3. h7 Be5. Kxc6 Bh8.
18   01:53  0.53   h5 f1=Q. Rxf1 Kxf1. a4 Bd2. Kc4 Ke2. a5 Bxa5. h6 Bc7. Kc5
Kd3. h7 Be5. Kxc6 Kd4. Kd7
Delfi

>>
>>My guess is that some engines' evaluation are not "aware" of the danger of two
>>passed pawns being stopped just by a bishop, and because of pruning and/or
>>Null-move, they need to reach a great depth to find the move.
>>
>>For example, in this position (4 plies after the initial one)...
>>
>>[d]8/8/2p5/8/7P/4b3/P1K5/5k2 w - - 0 3
>>
>>Shredder 7 needs a depth of 17 to give it a positive score (over +4.00), and
>>realize it's winning.
>>Fritz 8 needs only 14 to come up with a positive score(+0.44), but it takes
>>longer than for Shredder. Later, at depth 17 the score is a convincing +5.75.
>>Junior 8 needs to reach depth 21 (fast, though).
>>
>>[D]8/6p1/P1b1pp2/2p1p3/1k4P1/3PP3/1PK5/5B2 w - -
>>
>>In this other position, the problem appears to be the same: After several plies
>>we have:
>>
>>[d]8/P5p1/8/2p1p3/1k1P4/4p3/1PK3b1/8 w - - 0 6

Delfi  WB2 27 MB:
 2   00:00  0.83   a7 Ba8.
 3   00:00  0.89   a7 Ba8. Be2
 4   00:00  0.83   a7 Ba8. Be2 g5.
 5   00:00  0.79   a7 Ba8. Kd2 Kb3. Kc1
 6   00:00  0.53   a7 Kb5. d4+ Kb6. dxc5+ Kxa7. Bc4 Bd5.
 7   00:00  0.68   a7 e4. Kd2 exd3. Bxd3 c4. Bg6
 8   00:00  0.57   a7 e4. b3 Bd5. dxe4 Bxe4+. Bd3 Ba8. e4
 9   00:00  0.55   a7 e4. dxe4 Bxe4+. Bd3 Ba8. b3 e5. Bf5 Kb5.
10   00:01  0.62   a7 e4. dxe4 Bxe4+. Bd3 Ba8. b3 e5. Bc4 Be4+. Kd2 Ba8.
11   00:01  0.28   a7 e4. dxe4 Ka5. Bc4 Kb6. e5 fxe5. Kd3 g5. Bxe6 Kxa7.
12   00:02  0.11   a7 e4. dxe4 Bxe4+. Bd3 Ba8. Kd2 Ka5. Bc4 Kb6. Bxe6 Kxa7. Kd3
13   00:05  0.03   a7 e4. d4 cxd4. exd4 Ka5. Bc4 Kb6. Bxe6 Kxa7. d5 Ba4+. b3
Bb5.
14   00:08  0.00   a7 e4. dxe4 Bxe4+. Bd3 Ba8. Bg6 Ka5. Kd3 Kb6. Bf7 Kxa7. Bxe6
Kb6. Kc4
15   00:17 -0.12   a7 e4. dxe4 Bxe4+. Bd3 Ba8. Bg6 Ka5. Kd3 Kb6. Bf7 Kxa7. Bxe6
Kb6. Kc4 Be4.
16   00:28 -0.10   a7 e4. dxe4 Bxe4+. Bd3 Ba8. Bg6 Ka5. Kd3 Kb6. Bf7 Kxa7. Bxe6
Kb6. Kc4 Be4. Bd5
17   00:51 -0.16   a7 e4. dxe4 Bxe4+. Bd3 Ba8. Bg6 Ka5. Kd3 Kb6. Bf7 Kxa7. Bxe6
Kb6. Kc4 Bf3. Bd7 g6.
18   01:36 -0.14   a7 e4. dxe4 Bxe4+. Bd3 Ba8. Bg6 Ka5. Kd3 Kb6. Bf7 Kxa7. Bxe6
Kb6. Kc4 Bf3. Bd7 g6. Be6
18   03:10 -0.04   Bg2...
18   04:06  2.65   Bg2! Bxg2. e4 c4. a7 cxd3+. Kxd3 f5. a8=Q Bxe4+. Qxe4+ fxe4+.
Kxe4 Kb3. g5 Kc4. Kxe5 g6. Kxe6 Kd3. Kf7 Ke4. Kxg6
Delfi


>>
>>Both Fritz and Shredder need depth 12 from here to give a positive score! Any
>>(bad) player would realize almost instantly that white are winning here!
>>
>>Does it make sense what I'm saying?
>>
>>Regards,
>>
>>  Jaime
>
>Of course
>
>programs are basically stupid in detecting unstopable passed pawns.
>
>The practical importance of this problem was simply not big enough for the
>commercial programmers to care about fixing the problem until today.
>
>Uri



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.