Yes, I'm thinking about compiling a 64-bit TSCP (Sherwin Bitboard Version). Maybe it could win WBEC 5th div
Yes Master Jim, TSCP has some small benefits when compiled on 64-bits, especially Michael Sherwin's fixed versions. I'm also practicing this on the other open-sources. I heard before that Daniel Shawul made a version with bitbases. That's one to recon with. TSCP needs some additional upgrades to increase it's standings.
Fimbulwinter, Freyr, Needle, GK, Alibaba, Chessrikus, EXchess, Samchess, Simontacchi, Smallpotato, Scidlet, and even Amy are some engines that could try the 64-bit region. Faile and Phalanx-Reborn compiles in 64-bit, but haven't tried yet. There should be a Special tournament for 64-bit Old-timer Revivals.
ChessZone.org wrote:Did someone had success to beat it? I can't make even a draw (~5 games), my ELO is ~2250
Slowchess is a strong engine, it always did well in the analysis positions from Michael Gurevich in the CSS forum, but you do not often see it as a participant in the computer tournaments that people play. And, this version is said to be considerably stronger than the previous 2.82. I'm curious which is the best version at longer time controls than Blitz, is the Blitz version stronger overall, at long timecontrols too?
Eelco
I can't read German, are the analysis positions from Michael Gurevich available for download? Can anyone provide a link?
Thanks
Dave
Hello David,
There exists not I believe not a file consisting of all the positions that Michael has posted. The positions were meant as candidates for testpositions, and they were not all shown to be 'correct' testpositions, not all positions have a single solution for instance. Michael has made several selections of these postions that are available as tests, a link to the last collection of 40 postions:
Usually the testposions Michael posted are difficult and a program may need more than twenty minutes per position to solve some of them.
But I find myself most interesting that often the positions were tested by a big variety of programs and on different hardware, so you can compare many programs in a single position with often extensive output. If you have to translate German that sometimes will be tricky I admit but the PVs at least you can follow I think. The last list by Michael Gurevich of posted testpositions in the CSS forum is from July;
ChessZone.org wrote:Did someone had success to beat it? I can't make even a draw (~5 games), my ELO is ~2250
Slowchess is a strong engine, it always did well in the analysis positions from Michael Gurevich in the CSS forum, but you do not often see it as a participant in the computer tournaments that people play. And, this version is said to be considerably stronger than the previous 2.82. I'm curious which is the best version at longer time controls than Blitz, is the Blitz version stronger overall, at long timecontrols too?
Eelco
I can't read German, are the analysis positions from Michael Gurevich available for download? Can anyone provide a link?
Thanks
Dave
Hello David,
There exists not I believe not a file consisting of all the positions that Michael has posted. The positions were meant as candidates for testpositions, and they were not all shown to be 'correct' testpositions, not all positions have a single solution for instance. Michael has made several selections of these postions that are available as tests, a link to the last collection of 40 postions:
Usually the testposions Michael posted are difficult and a program may need more than twenty minutes per position to solve some of them.
But I find myself most interesting that often the positions were tested by a big variety of programs and on different hardware, so you can compare many programs in a single position with often extensive output. If you have to translate German that sometimes will be tricky I admit but the PVs at least you can follow I think. The last list by Michael Gurevich of posted testpositions in the CSS forum is from July;
The download consist of one epd file and 5 pgn files, not sure how many test positions yet, but there are obviously more than 40. I need to study the files and try to understand the numbering system. If only i could understand the German language ...
Diablo is an engine that uses a generator of moves 0x88, it seems that the only engines with generator of moves bitboard get faster at 64 bits.
Yes, I'm thinking about compiling a 64-bit TSCP (Sherwin Bitboard Version). Maybe it could win WBEC 5th div
A few months ago I made a construction of 64 bits of Olithink (not pgo), was 30% faster than the original, possibly with your construction Jim is a bit faster.
I'll see if I can squeeze a bit more nps out
How do you do train engines? Once builded your installed in the Arena or another GUI?
Ten 5-minute games in Arena then run some epd testsuites on it for an hour or so.
Jim.
I do not know nothing about sherwin tscp but
tscp has no chances to win 5th division of wbec only thanks to speed improvement.
It needs more than 300 elo improvement to do it and even making it 10 times faster is not enough
NagaSkaki 4.0 is not a champion of wbec inspite of being 300 elo better than tscp
Note that NagaSkaki has relatively better rating at long time control based on CCRL so I believe that the gap between nagaskaki and tscp is higher at long time control(I know from experience with old movei that tscp
is relatively better in blitz and movei0.07a could beat tscp with 5:1 time handicap only when the time control was long enough)
