Arasan 24.0 release

[jdart]

I'm maybe not following this discussion very well. To be clear, I haven't just copied solutions from any other test suite, although some of these positions are well known and may be used elsewhere. Everything in the arasan2023.epd has been run by me through a multi-pv deep depth search using Stockfish 16. It looks like I missed a couple positions that may have alternate solutions, but that should be rare.

[peter]

if you find further wrong ones, just show them too, please, so I don't have to check all of the positions on my own alone again, regards

It should be avoid move, not best move.

Code: Select all

8/1pp5/p3k1pp/8/P1p2PPP/2P2K2/1P3R2/5r2 b - - bm Rxf2+; hmvc 0; id "ERET 064 - Endspiel_?_1001, 63";
8/1pp5/p3k1pp/8/P1p2PPP/2P2K2/1P3R2/5r2 b - - am Rxf2; id "ERET 064 - Endgame";

4rrk1/Rpp3pp/6q1/2PPn3/4p3/2N5/1P2QPPP/5RK1 w - - bm Rxb7; hmvc 0; id "ERET 028 - Vergifteter Bauer_?_1001, 28";
4rrk1/Rpp3pp/6q1/2PPn3/4p3/2N5/1P2QPPP/5RK1 w - - am Rxb7; id "ERET 028 - Poisoned Pawn";

r2q1r1k/3bppbp/pp1p4/2pPn1Bp/P1P1P2P/2N2P2/1P1Q2P1/R3KB1R w KQ - bm b3; hmvc 0; id "ERET 024 - Entwicklung_?_1001, 24";
r2q1r1k/3bppbp/pp1p4/2pPn1Bp/P1P1P2P/2N2P2/1P1Q2P1/R3KB1R w KQ - am b3; id "ERET 024 - Development";

These are indeed silly syntax- errors of mine.
In this one collection

https://www.dropbox.com/s/ty11e3dm1f6tsqq/1234.epd?dl=0

, very near to the one of 1111 but in MEA- syntax and with Arasan 2021- suite instead of the new one and including some more UHO- single best move positions too, these 3 am- positions out of Eret were ok still, those nastinesses now may have occured with exchanging MEA- syntax to single best move- suite only, anyhow thanks again. Because easily to be corrected still, I'll wait with new upload till you don't find any further blunders neither anymore, regards

[peter]

I'm maybe not following this discussion very well. To be clear, I haven't just copied solutions from any other test suite, although some of these positions are well known and may be used elsewhere. Everything in the arasan2023.epd has been run by me through a multi-pv deep depth search using Stockfish 16. It looks like I missed a couple positions that may have alternate solutions, but that should be rare.

Sorry for hitchhiking your thread, Jon, thanks for the new suite anyhow once again. May I yet come back to the question of position- learning- feature of the engine?
You wrote it wouldn't work in UCI- mode, my question was, how then to make it work and if it's "only" about book- learning or about any by engine played or analysed moves as in Hiarcs' position- learning e.g., and if the learning- entries are stored in book or in any other one file?

viewtopic.php?p=951321#p951321

[jdart]

You wrote it wouldn't work in UCI- mode, my question was, how then to make it work and if it's "only" about book- learning or about any by engine played or analysed moves as in Hiarcs' position- learning e.g., and if the learning- entries are stored in book or in any other one file?

viewtopic.php?p=951321#p951321

The issue is that position learning, as it works in Arasan, operates by observing scores as they change from move to move. So if the score changes a lot between one search and the next, the later occurring score and position is remembered. However, the UCI protocol is basically stateless. All you get is a move list, and a command to go search it. So now to make position learning work, you have to basically keep game state in the engine, inferring it from the UCI commands. You can do that, but it gets a bit tricky, especially with ponder searches thrown into the mix.

[peter]

Thanks, Jon, don't understand fully but guess, the entries of learning are stored within the engine, right? Don't see any new by engine created file and if book.bin would be changed by learning, postion- learning wouldn't work independently from using the book?
BTW, I downloaded and installed the Arasan- GUI too (by Windows- installer) yesterday (engine only before) and now have a warning from Windows Defender about Wacatac Trojan:

[jdart]

Thanks, Jon, don't understand fully but guess, the entries of learning are stored within the engine, right? Don't see any new by engine created file and if book.bin would be changed by learning, postion- learning wouldn't work independently from using the book?
BTW, I downloaded and installed the Arasan- GUI too (by Windows- installer) yesterday (engine only before) and now have a warning from Windows Defender about Wacatac Trojan:

Book.bin is not changed. There is a separate learning file created, but you will only see it if using Xboard (CECP) mode. I am pretty sure you are getting a false positive from Defender. It's built with Inno setup and it may not like something in that: maybe that it may request admin privileges. However the executable was built on my system with Defender enabled and it should be ok.

[bmp1974]

Is there a way to convert arasan test suite of .epd containing all positions to a single .pgn format?

[peter]

Is there a way to convert arasan test suite of .epd containing all positions to a single .pgn format?

I always use chessbase- GUI for that. Newly created database in .pgn- format imports .epd positions of whole series with database- menue
Home-Paste-Append EPD.
Arena has a similiar way with menue EPD- Open and after loading the collection, then EPD- Save as, choosing .pgn instead of .epd as format, regards

[peter]

There is a separate learning file created, but you will only see it if using Xboard (CECP) mode.

Installed avx2-bmi2- compile in Arena as Winboard- engine, there then a checkbox Postion Learning is shown, but analysing a position that way then doesn't create any new file neither. Would Scid for PC help? (This one GUI I have installed too.) Starting binary in console with help I don't find a fitting command, sorry for being stubborn

[Dann Corbit]

Total elapsed machine time was 7-8 weeks, using about 100 cores.

--
Srdja

Maybe I'm not the only one who gets bar charts from the electric company showing how I use ludicrously more than my neighbors.