Hey, I am sure someone here has done this already. I would pay for the download and the hard work you did. It is ok if the promotions are only queen.
Also, KQPPKQP, and other pieces would be valuable if you have them.
KRPPKPP tablebase purchase
-
- Posts: 11
- Joined: Thu Dec 02, 2010 5:49 pm
- Sign-up code: 10159
- jshriver
- Posts: 298
- Joined: Tue Jan 31, 2006 5:59 am
- Sign-up code: 0
- Location: Toledo, OH, USA
- Contact:
Re: KRPPKPP tablebase purchase
If I could I would. 7men, has been on my dream list for the past several years but alas no code to generate these. If I had it within my means I'd generate and give away for free. Good luck.
-Josh
-Josh
-
- Posts: 489
- Joined: Sat Jan 21, 2006 10:43 am
- Sign-up code: 10159
- Location: Reading, UK
- Contact:
Re: KRPPKPP tablebase purchase
I presume you meant KRPPKRP rather than KRPPKPP.
Yakov K and Marc B wrote an article about their sub-9-man EGT Generation, using Yakov's program. It is in "64", 2009/7 in Russian. Their previous MB/YK article on 7-man P-less EGTs was in EG Vol XI pp 493-510.
They did all sub-7-man EGTs afresh (to the DTC metric) with no restrictions on P-promotion. They have compressed EGTs which are 50% the size of Nalimov's. This is even better than the size of the MB/JTamplin 'Nalimov' DTC EGTs, so they have achieved some compression improvements there.
Generation-performance is also much better because of (a) Pentium Assembler programming and (b) use of 'unmoves' as well as 'moves'. KQPKQQ DTC EGT generated in 1 hour, 15 hours for KQPKRB.
They point out that it is possible to hasten (and simplify) the business of creating EGTs for P-ful endgames if promotion is constrained to P=Q or P=Q/N. Indeed they have run 5-man EGT generation with various constraints on P-promotion, and thereby discovered a scenario which, in different lines, features promotion to R, N or Q. Consider their position Kc8,Qf4,g7/Kh5,Qh1 - White to win![Smile :-)](./images/smilies/icon_smile.gif)
It appears that the P-ful 7-man endgames being investigated are the ones which feature most otb and that KRPPKRP was first. KXPPKXP is being investigated for all X and MB may comment on otb errors which have been discovered by the new ground truth.
They also mention KBBPKNN where P(b) is generally won, and P(other) is drawn if the defender's King confronts the Pawn.
These EGTs are not, afaik, Nalimov format, and I doubt if Nalimov's current EGT-access code would access them if they were.
g
Yakov K and Marc B wrote an article about their sub-9-man EGT Generation, using Yakov's program. It is in "64", 2009/7 in Russian. Their previous MB/YK article on 7-man P-less EGTs was in EG Vol XI pp 493-510.
They did all sub-7-man EGTs afresh (to the DTC metric) with no restrictions on P-promotion. They have compressed EGTs which are 50% the size of Nalimov's. This is even better than the size of the MB/JTamplin 'Nalimov' DTC EGTs, so they have achieved some compression improvements there.
Generation-performance is also much better because of (a) Pentium Assembler programming and (b) use of 'unmoves' as well as 'moves'. KQPKQQ DTC EGT generated in 1 hour, 15 hours for KQPKRB.
They point out that it is possible to hasten (and simplify) the business of creating EGTs for P-ful endgames if promotion is constrained to P=Q or P=Q/N. Indeed they have run 5-man EGT generation with various constraints on P-promotion, and thereby discovered a scenario which, in different lines, features promotion to R, N or Q. Consider their position Kc8,Qf4,g7/Kh5,Qh1 - White to win
![Smile :-)](./images/smilies/icon_smile.gif)
It appears that the P-ful 7-man endgames being investigated are the ones which feature most otb and that KRPPKRP was first. KXPPKXP is being investigated for all X and MB may comment on otb errors which have been discovered by the new ground truth.
They also mention KBBPKNN where P(b) is generally won, and P(other) is drawn if the defender's King confronts the Pawn.
These EGTs are not, afaik, Nalimov format, and I doubt if Nalimov's current EGT-access code would access them if they were.
g