Historical bugs(1) - who made what wrong?

[Guenther Simon]

Here is an old debug excerpt from the past. At the time being I adjourned the game as a draw
because a 3 time rep was already on board after 20. Ne4, but for my current bugbase I still would like
to know whose bug it really was? AFAIK KKFChess practically played the same move twice which
is illegal of course, but was HokusPokus completely correct too(I guess simply ignoring the draw offer
is ok as in real chess?)?
Draw offering is probably still a dark zone for some older engines and in the past at least
led to strange behaviour not only in the programs which were offered them;-)
BTW I consider it very annoying if some programs in lost positions permanently offer draws
may be in the hope to crash the opponent(not meaning KKFChess here, but I saw at least one
case in one of my recent rating tourneys!)

Note that the WB version still was plain 4.27 at the time of that game:

Code: Select all

799229 >second: usermove 799229 >second: Ng3
799500 <second:  1   55   0   0   19...Qf6 20.c4 Nf4 21.Qe3 O-O 22.Qd4 [MG]
799500 <second:  5   22   26   720   19...Qf6 [Fail #1: Low <-1.10,+0.72>]
799520 <second:  5   -48   28   10853   19...Qf6 20.c4 Nf4 21.Qd2 O-O 22.Ne4 Qf5 [MG]
799550 <second:  6   -15   32   26709   19...Qf6 20.c4 Nf4 21.Qd2 O-O 22.Ne4 Qf5 [MG]  [Fail #1: High <-0.65,+1.17>]
799590 <second:  6   -13   35   41518   19...Qf6 20.c4 Nf4 21.Ne4 Ne2+ 22.Kh1 Qxf3 23.gxf3 Nxc1 24.Raxc1 [MG]
799660 <second:  7   -20   42   77043   19...Qf6 20.c4 Nf4 21.Qe3 O-O-O 22.Qxa7 Qxb2 23.Rfd1 [MG]
801072 <second:  8   -27   183   642426   19...Qf6 20.c4 Nf4 21.Qe3 O-O 22.Qe5 Rae8 [MG]
805969 <second:  9   -24   673   3050947   19...Qf6 20.c4 Nf4 21.Qe3 O-O 22.Qe5 Rac8 23.Qxf6 Rxf6 [MG]
818777 <second: 10   -16   1955   9986510   19...Qf6 20.c4 Nf4 21.Qd2 O-O 22.Ne4 Qf5 23.Qe3 [MG]
826719 <second: 11   -16   2749   13939172   19...Qf6 20.c4 Nf4 21.Qd2 O-O 22.Ne4 Qf5 23.Qe3 Rad8 [MG]
826719 <second: move Qf6
Interrupting first
826719 >first : time 52050
otim 46542
826719 >first : f5f6
826759 <first : 3 97 0 1077 c2c4 d5f4 f1d1
826779 <first : 3 97 0 3056 c2c4 d5f4 f1d1
826779 <first : 4 52 0 5206 c2c4 d5f4 f1d1 e8g8
826789 <first : 4 74 0 8485 g3e4 f6f5 e4g3 f5f8
826859 <first : 4 74 0 25083 g3e4 f6f5 e4g3 f5f8
826899 <first : 5 86 0 34146 g3e4 f6f5 e4g3 f5f8 f1d1
827159 <first : 5 86 0 137828 g3e4 f6f5 e4g3 f5f8 f1d1
827320 <first : 6 83 100 191949 g3e4 f6f5 e4g3 f5f4 f1e1 f4c1 a1c1
829923 <first : 6 83 300 957651 g3e4 f6f5 e4g3 f5f4 f1e1 f4c1 a1c1
831025 <first : 7 91 400 1337323 g3e4 f6f5 e4g3 f5f4 c1f4 d5f4 a1d1
848310 <first : 7 91 2200 7835648 g3e4 f6f5 e4g3 f5f4 c1f4 d5f4 a1d1
848310 <first : move g3e4
848310 >second: time 46542
otim 49890
848310 >second: usermove 848310 >second: Ne4
848320 <second: offer draw
Interrupting first
848320 >first : draw
848570 <second: offer draw
848570 >first : draw
848570 <second: move Qf6
GameEnds(31, Forfeit due to illegal move, 4)
848570 >first : result 1-0 {Forfeit due to illegal move}
848570 >second: result 1-0 {Forfeit due to illegal move}
848570 >first : force
848570 >second: force
848570 >second: ping 2
Exiting: Match HokusPokus_043 vs. KKFChess v2.6.1pmx: final score 1-0-0
GameEnds(0, (null), 2)
848570 >first : quit
848570 >second: quit


Guenther

[Guenther Simon]

In the position below Black played Rb5
and Blikskottel_07 captured axb6ep

[diag]6k1/pr4p1/8/P1p2p2/5B2/2K3P1/2P5/4R3 b - - 0 55[/diag]
[D]6k1/pr4p1/8/P1p2p2/5B2/2K3P1/2P5/4R3 b - - 0 55[D]

...if I figure out how to post diagrams again I will correct this post



Thanks to Fonzy who remembered this service at:
http://webchess.freehostia.com

Guenther

[H.G.Muller]

My gues is that KKFChess cannot search from a position that is a 3-fold rep, perhaps because it gets a hash hit in the root that says: draw to infinite depth. Th search then returns wwithout having found a best move, and thus the previous move is still in the move buffer, and it prints it again.

In the WB 4.2.7 system plus WBEC requirement that engines claim, KKFChess should have send the RESULT command here (1/2-1/2 {3-fold rep}), and not offer a draw.

In WinBoard 4.3.15 the offer draw in the presence of a draw condition before or after the following move would be treated as a draw claim, and lead to imemdiate termination of the game. Note that in 4.3.15 it is also not possible to distract your opponent throgh draw offers: if you make a draw offer, the GUI holds it back from the opponent until it also has your next move, and then (if it cannot treat the offer as a claim, even after that move) sends the offer + mov in immediate succession. So in fact you cannot even disturb the opponent's pondering.