When the endgame goes bad...

Discussion of anything and everything relating to chess playing software and machines.

Moderator: Ras

Golem

When the endgame goes bad...

Post by Golem »

Below some endgame played in my blitz tournament. All engines played without tablebase but I think it is not an excuse for the very bad play in the last 2 games...

Endgame 1
White : Scorpio 2.0.2
Black : Thinker 5.3B (Passive)
Score : 0-1
Position :
7K/8/4nk2/5N2/6R1/2r5/8/8 w - - 68 105

Continuation : 105. Nh4?? Rc8+ 106. Kh7 Ng5+ 107. Rxg5 and white resign few moves later.

Note : Which engine can show a mate score after Nh4 ?? (without TB)


Endgame 2
White : Rybka 2.2n2
Black : Naum 2.0
Score : 1/2-1/2

6k1/8/7b/8/4BBK1/8/8/8 w - - 1 152

Rybka plays 152. Bxh6 and draw de game by 50 moves rule. The King on g4 will never move (he seems pinned to his square by invisible forces ;-)).


Endgame 3
White : Toga 1.4beta5c
Black : Stockfish 1.01
Score : 1/2-1/2

8/8/5K2/3q4/8/8/3N1k2/8 b - - 3 95

After 95. ... Qxd2, Stockfish announce "Mat in 9" but the games will end in a draw...
Note : Stockfish has plenty of time but does not use it, Toga is under time pressure, how is it possible that this endgame ends in a draw ?

The games continue like this :

96. Ke5 {-10.41/8 1} Ke3 {+M9/6 0}
97. Ke6 {-M6/13 0} Qd4 {+M9/6 0} 98. Ke7 {-10.41/7 1} Qg7+ {+M8/6 0} 99.
Kd6 {-10.33/7 0} Kf4 {+M7/6 0} 100. Kc5 {-10.23/7 0} Qb7 {+M9/6 0} 101. Kc4
{-10.39/8 1} Qb6 {+M11/7 0} 102. Kd3 {-10.32/7 0} Qe3+ {+M9/7 0} 103. Kc4
{-10.37/8 1} Qe2+ {+M11/8 0} 104. Kc5 {-10.34/8 1} Qb2 {+M10/6 0} 105. Kc4
{-10.35/8 0} Ke4 {+M9/6 0} 106. Kc5 {-10.23/4 0} Kf3 {+M11/6 0} 107. Kc4
{-10.26/8 1} Ke2 {+M9/6 0} 108. Kd5 {-10.23/7 0} Qf6 {+M7/6 0} 109. Kc4
{-10.16/7 0} Qb6 {+M10/7 0} 110. Kd5 {-10.33/8 1} Kd3 {+M7/6 0} 111. Ke5
{-10.08/4 0} Ke3 {+M8/6 0} 112. Kd5 {-10.41/8 1} Qd8+ {+M6/6 0} 113. Kc5
{-10.35/8 2} Qb8 {+M10/9 0} 114. Kc4 {-10.33/8 1} Kd2 {+M8/6 0} 115. Kd4
{-10.19/7 0} Qf4+ {+M10/6 0} 116. Kd5 {-10.23/8 1} Kd1 {+M9/6 0} 117. Kc5
{-10.23/8 1} Qc7+ {+M14/6 0} 118. Kd4 {-10.01/7 0} Qc1 {+M9/6 0} 119. Ke5
{-10.03/7 0} Qc4 {+M8/6 0} 120. Kd6 {-10.18/9 1} Qc8 {+M9/7 0} 121. Ke5
{-10.14/8 1} Qa6 {+M9/6 0} 122. Kd4 {-10.21/9 1} Qc6 {+M10/7 0} 123. Ke3
{-10.34/9 0} Qf6 {+M9/6 0} 124. Ke4 {-10.33/9 0} Qf2 {+M9/6 0} 125. Kd5
{-10.23/9 1} Qh4 {+M10/6 0} 126. Ke5 {-10.23/8 0} Qh8+ {+M9/6 0} 127. Kd5
{-10.23/9 1} Qc3 {+M8/6 0} 128. Ke4 {-10.20/9 1} Kc2 {+M8/6 0} 129. Kf5
{-10.28/8 1} Qh3+ {+M11/7 0} 130. Kf4 {-10.23/8 1} Kd2 {+M10/6 0} 131. Ke5
{-10.23/8 1} Ke1 {+M9/6 0} 132. Kd4 {-10.08/8 1} Qg3 {+M11/6 0} 133. Ke4
{-10.20/8 1} Qg8 {+M12/6 0} 134. Ke5 {-10.20/8 0} Kf2 {+M10/6 0} 135. Kd4
{-10.14/7 0} Qg7+ {+M11/6 0} 136. Kc4 {-10.23/7 1} Qb2 {+M12/6 0} 137. Kd5
{-10.23/7 0} Qc3 {+M11/6 0} 138. Kd6 {-10.23/7 1} Kg3 {+M12/6 0} 139. Ke6
{-10.26/7 0} Qc4+ {+M10/6 0} 140. Ke5 {-10.43/8 1} Kf3 {+M8/7 0} 141. Kf6
{-10.43/9 1} Qg8 {+M7/6 0} 142. Ke5 {0.00/18 0} Qd8 {+M9/6 0} 143. Ke6
{0.00/31 0} Qf8 {0.00/100 0} 144. Ke5 {0.00/51 0} Ke3 {0.00/100 0} 145. Kd5
{0.00/63 0} Qf3+ {0.00/100 0 50 moves rule} 1/2-1/2

PS : If someone can post the position, I don't find the tag to do it...
Dann Corbit
Posts: 12803
Joined: Wed Mar 08, 2006 8:57 pm
Location: Redmond, WA USA

Re: When the endgame goes bad...

Post by Dann Corbit »

Strange, I can't see the diagrams in preview mode either.
This should do it:
[d]7K/8/4nk2/5N2/6R1/2r5/8/8 w - -
[d]6k1/8/7b/8/4BBK1/8/8/8 w - -
[d]8/8/5K2/3q4/8/8/3N1k2/8 b - -
User avatar
Eelco de Groot
Posts: 4681
Joined: Sun Mar 12, 2006 2:40 am
Full name:   Eelco de Groot

Re: When the endgame goes bad...

Post by Eelco de Groot »

Golem wrote:Below some endgame played in my blitz tournament. All engines played without tablebase but I think it is not an excuse for the very bad play in the last 2 games...

Endgame 1
White : Scorpio 2.0.2
Black : Thinker 5.3B (Passive)
Score : 0-1
Position :
7K/8/4nk2/5N2/6R1/2r5/8/8 w - - 68 105

Continuation : 105. Nh4?? Rc8+ 106. Kh7 Ng5+ 107. Rxg5 and white resign few moves later.

Note : Which engine can show a mate score after Nh4 ?? (without TB)


105. Nh4 is still a draw says the database at Shredderchess.com, but White is lost after 106. Kh7?:

106. Kh7 is -M23
106. Rg8 is a draw (0)
(Online 6 men databases www.shredderchess.com, thanks Stefan for this resource!)

Toga does not find the mate (even) with bitbases but at least it plays the winning Ng5+:

[d]2r5/7K/4nk2/8/6RN/8/8/8 b - -

Engine: Toga Mara Beta 6 Build 82 (Athlon 2009 MHz, 256 MB)
by Fabien Letouzey, Thomas Gaksch and Tam

1/05 0:00 +0.35 2...Ng5+ 3.Rxg5 Kxg5 4.Nf3+ Kf4 (14)

2/11 0:00 +48.80 2...Ng5+ 3.Rxg5 Kxg5 (2.240)
.
.
30/48 0:37 +48.90 2...Ng5+ 3.Rxg5 (23.184.242) 618

31/50 0:54 +48.90 2...Ng5+ 3.Rxg5 (33.536.480) 612

32/52 1:18 +48.90 2...Ng5+ 3.Rxg5 (47.716.325) 604

33/52 1:53 +48.90 2...Ng5+ 3.Rxg5 (67.794.926) 595

34/52 2:40 +48.90 2...Ng5+ 3.Rxg5 (93.974.445) 584

35/52 3:43 +48.90 2...Ng5+ 3.Rxg5 (128.449.274) 575

36/52 5:09 +48.90 2...Ng5+ 3.Rxg5 (175.061.773) 564

best move: Ne6-g5 time: 5:36.704 min n/s: 560.938 CPU 100.0% n/s(1CPU): 560.938 nodes: 188.870.000

Probably Togas without bitbases will find a mate, but I have not tried that.
The experimental Build 83 also does not find mate but it does go very deep compared to Build 82 here, where Build 82 goes to 36 plies in the same time 83 does 44. It's a bit risky, not very sophisticated, it does 'brute pruning' :) :


[d]2r5/7K/4nk2/8/6RN/8/8/8 b - -

Engine: Toga Mara Beta 6 Build 83 (Athlon 2009 MHz, 256 MB)
by Fabien Letouzey, Thomas Gaksch and Tam

25/38 0:02 +48.90 2...Ng5+ 3.Rxg5 (2.232.659) 855

26/40 0:03 +48.90 2...Ng5+ 3.Rxg5 (2.730.584) 812

27/40 0:04 +48.90 2...Ng5+ 3.Rxg5 (3.344.025) 786

28/42 0:05 +48.90 2...Ng5+ 3.Rxg5 (4.407.857) 754

29/42 0:07 +48.90 2...Ng5+ 3.Rxg5 (5.496.305) 731

30/44 0:10 +48.90 2...Ng5+ 3.Rxg5 (7.694.345) 718

31/44 0:14 +48.90 2...Ng5+ 3.Rxg5 (10.363.896) 692

32/50 0:20 +48.90 2...Ng5+ 3.Rxg5 (13.848.052) 692

33/50 0:26 +48.90 2...Ng5+ 3.Rxg5 (17.352.886) 666

34/50 0:31 +48.90 2...Ng5+ 3.Rxg5 (20.880.506) 656

35/50 0:44 +48.90 2...Ng5+ 3.Rxg5 (28.037.949) 634

36/52 0:50 +48.90 2...Ng5+ 3.Rxg5 (31.617.626) 626

37/60 1:06 +48.90 2...Ng5+ 3.Rxg5 (41.557.378) 628

38/60 1:18 +48.90 2...Ng5+ 3.Rxg5 (48.475.561) 616

39/60 1:33 +48.90 2...Ng5+ 3.Rxg5 (56.664.647) 605

40/60 2:00 +48.90 2...Ng5+ 3.Rxg5 (70.972.195) 589

41/60 2:25 +48.90 2...Ng5+ 3.Rxg5 (85.084.623) 584

42/66 3:31 +48.90 2...Ng5+ 3.Rxg5 (123.676.869) 584

43/66 4:09 +48.90 2...Ng5+ 3.Rxg5 (142.875.929) 572

44/66 5:13 +48.90 2...Ng5+ 3.Rxg5 (175.344.876) 559

best move: Ne6-g5 time: 5:27.203 min n/s: 556.963 CPU 100.0% n/s(1CPU): 556.963 nodes: 182.240.000

Eelco
Debugging is twice as hard as writing the code in the first
place. Therefore, if you write the code as cleverly as possible, you
are, by definition, not smart enough to debug it.
-- Brian W. Kernighan