|(1) Posted by Siegfried Hornecker [Sunday, Apr 17, 2011 11:38]|
The advent of 7 piece EGTB
Kirill Kryukov (Russia/Japan) shared the following links with me to distribute:
In short, the bounty is open. The first finely working 7 piece EGTB generator can win a money prize, but the real feat would be to probably gain entrance into encyclopaedias for his programming skills. ;-)
|(2) Posted by Torsten Linß [Wednesday, Apr 27, 2011 23:49]|
There are plenty of 7 piece TB generated chess problems around already...
|(3) Posted by Olaf Jenkner [Friday, Apr 29, 2011 19:25]|
Yes, most of them are selfmates, see
None of them is cooked.
|(4) Posted by Olaf Jenkner [Friday, Apr 29, 2011 19:27]|
Sorry, the link doesn't work this way, please use the clipboard.
|(5) Posted by Administrator [Saturday, Apr 30, 2011 02:39]|
Neither clipboard works for me (invalid conversion from UTF-8 'ß'); the following should do the job:
|(6) Posted by Olaf Jenkner [Saturday, Apr 30, 2011 09:17]|
You get 7 problems by Lindgren and Lindberg more.
ß is a strange letter...
|(7) Posted by Geoff Foster [Monday, May 2, 2011 02:36]|
How is it possible to use EGTB to compose selfmates? The EGTB only shows what happens with best play from both sides.
|(8) Posted by Olaf Jenkner [Monday, May 2, 2011 18:21]|
These tablebases are generated with worst moves of both sides.
|(9) Posted by Geoff Foster [Tuesday, May 3, 2011 00:50]; edited by Geoff Foster [11-05-03]|
Have tablebases been generated especially for selfmates? Is it possible that someone has been living amongst us for years (on the top floor of a specially-built compound), generating these tablebases without being detected?
|(10) Posted by Olaf Jenkner [Tuesday, May 3, 2011 20:54]|
Ask Torsten, it is in Germany no secret.
It is clear when you look at his unbelievable miniatures shown in the link above.
|(11) Posted by Geoff Foster [Wednesday, May 4, 2011 02:28]|
This is remarkable news. One of the steps in generating a Depth To Mate (DTM) tablebase for the given material is to perform "retrograde analysis" from all checkmated positions. To generate a Depth To Selfmate tablebase, the retrograde analysis need only be performed from positions in which Black must checkmate White. Apparently this narrows down the possibilities to such an extent that even 7-piece problems can be found!
All of the problems in the link had just two black pieces (including the black king). Also, Torsten was not necessarily searching for the "ultimate truth", so he could terminate the retrograde analysis at any point (although there was one s#90). However, it is remarkable that these problems date from as early as 1996!
I'd like to know more. Torsten would only have been interested in dual-free White play, so how many potential problems had to be trawled through in order to find the nuggets? I'm still struggling to understand the process -- aside from the fact that the retrograde analysis must be done from selfmate positions (not checkmate positions), is there any other difference in the method used to generate selfmate tablebases?
No more posts
MatPlus.Net Forum Internet and Computing The advent of 7 piece EGTB