pohl4711 wrote: ↑Fri Sep 02, 2022 1:10 pm
I am working on a new version: All 324 positions combined with 2 pawn-plies (one step). Example: 1.a3 h6 or 1.c3 f6 and so on...
This allows 8*8=64 combinations, which means 64x324= 20736 different opening-lines. I built this file already and mixed it by random, using SCID. Now, I evaluate all endpositions with KomodoDragon (pgnscanner-tool) with 20 secs/position on my Ryzen 12core machine. That will take around 5 days to finish. With these many (and evaluated) lines, it should be possible to build (for example) a 500 lines file with unbalanced lines (in the eval-range of my UHO-openings). Or other files (balanced, or better for black, or whatever). I will deliver the raw-data, too, of course, with all lines containing searchdepth and eval of the endposition in the annotator-tag, like I always do. So everybody can filter lines by eval, like he wants to.
If all works as expected, the release will be in 8-10 days from now on my website.
Stay tuned.
I decided, that it makes sense to add NBC and NBSC (No Black Castling, No Shoert Black Castling) Sets of the big chess324 set, I made (see above). Because IMO, "normal" chess324 positions are very drawish (at least a lot of them).
So, this will take some more days to build, to evaluate and to test, so the release of the chess324-openings-sets will be (hopefully) in the second half of september.
AndrewGrant wrote: ↑Fri Sep 02, 2022 3:04 pm
Chess324 is ... very drawish it seems. Now an open question: compare the exit values of Chess324 to FRC to DFRC. I feel like FRC is far less drawish than Chess324 -- which makes me think either the rook/king placement is _very_ important, or that more randomization just equals more noise to produce bad openings...
FRC is more drawish than chess324.
So, it's improvement as opening set (without making book moves).
DFRC again exceeds the draw range heavily. Opening exits may vary from -3.25 to +4.70 (in evaluation scale of Stockfish development version).
At the moment seems chess324 is a best compromise between number of openings, openings distribution and rules.
AndrewGrant wrote: ↑Fri Sep 02, 2022 3:04 pm
The mid-event "update", was actually a mid-event "revert one commit", so I felt happy to do it. Not a super serious event, so no harm done really by having some buggyness.
Chess324 is ... very drawish it seems. Now an open question: compare the exit values of Chess324 to FRC to DFRC. I feel like FRC is far less drawish than Chess324 -- which makes me think either the rook/king placement is _very_ important, or that more randomization just equals more noise to produce bad openings...
So far the Chess324 event has roughly 70% draws. I don't remember the %draws in the last FRC similar event, but I suspect it was higher than that, since in the final between SF and Dragon there were nearly 99% draws. Why do you say that FRC is less drawish than 324? It sure doesn't look that way to me. It is probably more drawish than playing standard chess with the UHO22 openings, but that's to be expected, since only some of the 324 positions are that unbalanced, whereas all of the UHO positions are.
AndrewGrant wrote: ↑Fri Sep 02, 2022 3:04 pm
The mid-event "update", was actually a mid-event "revert one commit", so I felt happy to do it. Not a super serious event, so no harm done really by having some buggyness.
Chess324 is ... very drawish it seems. Now an open question: compare the exit values of Chess324 to FRC to DFRC. I feel like FRC is far less drawish than Chess324 -- which makes me think either the rook/king placement is _very_ important, or that more randomization just equals more noise to produce bad openings...
So far the Chess324 event has roughly 70% draws. I don't remember the %draws in the last FRC similar event, but I suspect it was higher than that, since in the final between SF and Dragon there were nearly 99% draws. Why do you say that FRC is less drawish than 324? It sure doesn't look that way to me. It is probably more drawish than playing standard chess with the UHO22 openings, but that's to be expected, since only some of the 324 positions are that unbalanced, whereas all of the UHO positions are.
Has anybody figured out what is the percentage of draws in Chess960, and what is the percentage in Chess324 after game 154 when a better LCO replaced the buggy one ? https://www.chess.com/computer-chess-championship
AndrewGrant wrote: ↑Fri Sep 02, 2022 3:04 pm
The mid-event "update", was actually a mid-event "revert one commit", so I felt happy to do it. Not a super serious event, so no harm done really by having some buggyness.
Chess324 is ... very drawish it seems. Now an open question: compare the exit values of Chess324 to FRC to DFRC. I feel like FRC is far less drawish than Chess324 -- which makes me think either the rook/king placement is _very_ important, or that more randomization just equals more noise to produce bad openings...
So far the Chess324 event has roughly 70% draws. I don't remember the %draws in the last FRC similar event, but I suspect it was higher than that, since in the final between SF and Dragon there were nearly 99% draws. Why do you say that FRC is less drawish than 324? It sure doesn't look that way to me. It is probably more drawish than playing standard chess with the UHO22 openings, but that's to be expected, since only some of the 324 positions are that unbalanced, whereas all of the UHO positions are.
Has anybody figured out what is the percentage of draws in Chess960, and what is the percentage in Chess324 after game 154 when a better LCO replaced the buggy one ? https://www.chess.com/computer-chess-championship
Since the switch to the supposedly better Lc0 version, I count 1 win, 3 losses, and 22 draws for Lc0, even a tad worse than in the first 154 games. So it seems that the poor results are not likely due to this issue, rather to having to play positions on which it was not trained, which hurts much more when you rely on huge nets rather than deep search. At least that seems the most plausible explanation to me.
AndrewGrant wrote: ↑Fri Sep 02, 2022 3:04 pm
The mid-event "update", was actually a mid-event "revert one commit", so I felt happy to do it. Not a super serious event, so no harm done really by having some buggyness.
Chess324 is ... very drawish it seems. Now an open question: compare the exit values of Chess324 to FRC to DFRC. I feel like FRC is far less drawish than Chess324 -- which makes me think either the rook/king placement is _very_ important, or that more randomization just equals more noise to produce bad openings...
So far the Chess324 event has roughly 70% draws. I don't remember the %draws in the last FRC similar event, but I suspect it was higher than that, since in the final between SF and Dragon there were nearly 99% draws. Why do you say that FRC is less drawish than 324? It sure doesn't look that way to me. It is probably more drawish than playing standard chess with the UHO22 openings, but that's to be expected, since only some of the 324 positions are that unbalanced, whereas all of the UHO positions are.
Has anybody figured out what is the percentage of draws in Chess960, and what is the percentage in Chess324 after game 154 when a better LCO replaced the buggy one ? https://www.chess.com/computer-chess-championship
Since the switch to the supposedly better Lc0 version, I count 1 win, 3 losses, and 22 draws for Lc0, even a tad worse than in the first 154 games. So it seems that the poor results are not likely due to this issue, rather to having to play positions on which it was not trained, which hurts much more when you rely on huge nets rather than deep search. At least that seems the most plausible explanation to me.
Therefore, LCO has to use a small Net, which allows LCO to search deeper, instead of wasting time searching a huge net that has no Chess324 training.
