ACT5
ACT5
Alternative Chess Test (ACT) - is series of chess tests that mostly contains tricky endgame puzzles and difficult mates.
Download:
https://www.mediafire.com/file/ir89pzfo ... 2.zip/file
It is probably the last new test from the ACT series. There will be no ACT6, instead of that I will release a refreshed ACT test with the hardest positions selected from ACT1-B, ACT2, ACT3, ACT4, ACT5 together with some new positions.
Download:
https://www.mediafire.com/file/ir89pzfo ... 2.zip/file
It is probably the last new test from the ACT series. There will be no ACT6, instead of that I will release a refreshed ACT test with the hardest positions selected from ACT1-B, ACT2, ACT3, ACT4, ACT5 together with some new positions.
Re: ACT5
Thanks a lot for the fine suite, especially grateful I'm always, if original sources with players'- (and) or composers'- (as for he studies) names are given together with the positions, there ACT includes more of that kind then most other suites to be downloaded from fora do.dorsz wrote: ↑Tue Jun 07, 2022 4:45 pmAlternative Chess Test (ACT) - is series of chess tests that mostly contains tricky endgame puzzles and difficult mates.
Download:
https://www.mediafire.com/file/ir89pzfo ... 2.zip/file
It is probably the last new test from the ACT series. There will be no ACT6, instead of that I will release a refreshed ACT test with the hardest positions selected from ACT1-B, ACT2, ACT3, ACT4, ACT5 together with some new positions.
Do you maybe have a download- link of ACT3 still too?
Dann Corbit gave one here long ago
http://talkchess.com/forum3/viewtopic.p ... 98#p820698
, but Rybka- Forum is archive only now, and search doesn't work (as well as my old username and password don't) to me there anymore.
BTW search here in OpenChess doesn't work for me neither. Tried many times over the weeks with several different terms.
Thanks again and thanks in advance,
regards
Peter.
Peter.
Re: ACT5
Everything is here from the beginning:
http://dorszcz.blogspot.com/
I will release something new this year. After that I will refresh all sets by removing too easy positions for nowadays engines and I will put them in two files: ACT2022 and TTT2022 (or 2023 if I don't make it this year).
Re: ACT5
Thank you very much again!dorsz wrote: ↑Wed Aug 17, 2022 10:32 amEverything is here from the beginning:
http://dorszcz.blogspot.com/
I will release something new this year. After that I will refresh all sets by removing too easy positions for nowadays engines and I will put them in two files: ACT2022 and TTT2022 (or 2023 if I don't make it this year).
Great news!
Best
regards
Peter.
Peter.
- Chess_Dragon
- Posts: 50
- Joined: Wed Jun 13, 2012 9:52 pm
Re: ACT5
Thank you for your testsuites! I had a lot of fun combining them:
Code: Select all
R7 2700X 14t|RTX 3080
ACT 1B-5 + TTT 1-2
60 secs/position solved/783 solved ave time
Lc0.31dagEr_t2 650 83.0% 5.58 secs *
15 secs/position
Lc0.31dagEr_LgDist 583 74.5% 1.82 secs
Stockfish-23081619 578 73.8% 2.22 secs
Lc0.31dagEr_t2 573 73.2% 1.73 secs *
Lc0.31dagEr_BT2 567 72.4% 1.78 secs
Lc0.31dagEr_t1 558 71.3% 1.79 secs
Lc0.31dagEr_T80 537 68.6% 2.05 secs
Lc0.31dagEr_T80old 533 68.1% 1.88 secs
Lc0.31dagEr_T78 521 66.5% 2.15 secs
* t2 60s additional solutions were very long
- Chess_Dragon
- Posts: 50
- Joined: Wed Jun 13, 2012 9:52 pm
Re: ACT5
Great! Testsuites do not get the attention they deserve, and it must take human skill to make them.
For example in my post virtually all the Lc0 nets are in the expected order - except it was a surprise that my expensive 3080 benefited from the smaller faster LgDist, only maybe a minor disagreement (from perhaps elsewhere, but not me) about t1 lower than BT2.
Otherwise it was as expected: newer best, oldest worse.
These are maybe a little old from the high solved rate and low ave. time ("even" Lc0 got several instantly ) - but as above they were good enough for excellent differentiation.
I liked the 783 positions at 15 secs. each better than 783 games each which would take too long for me, or would have to have too short a TC to have much meaning to me (though that is done).
If you did make more I would probably run something like my last again: it worked out very well, and there are interesting things with Lc0 going on.
For example in my post virtually all the Lc0 nets are in the expected order - except it was a surprise that my expensive 3080 benefited from the smaller faster LgDist, only maybe a minor disagreement (from perhaps elsewhere, but not me) about t1 lower than BT2.
Otherwise it was as expected: newer best, oldest worse.
These are maybe a little old from the high solved rate and low ave. time ("even" Lc0 got several instantly ) - but as above they were good enough for excellent differentiation.
I liked the 783 positions at 15 secs. each better than 783 games each which would take too long for me, or would have to have too short a TC to have much meaning to me (though that is done).
If you did make more I would probably run something like my last again: it worked out very well, and there are interesting things with Lc0 going on.
- Chess_Dragon
- Posts: 50
- Joined: Wed Jun 13, 2012 9:52 pm
Re: ACT5
To re-phase my last: the only thing about ACT & TTT was some of the Mate in .., a low number of moves might be a little easy for newer engines.
Soon I will try Swordfish & Blue Marlin
Soon I will try Swordfish & Blue Marlin