Chess Forum

massimilianogoi wrote: ↑Fri Sep 01, 2023 10:22 am I mean Raid vs these Stockfish https://github.com/official-stockfish/S ... ase%3Atrue

Orgz wrote: ↑Fri Sep 01, 2023 7:02 am

massimilianogoi wrote: ↑Fri Sep 01, 2023 1:15 am How are the tests with the latest Stockfish betas?

Orgz wrote: ↑Thu Aug 31, 2023 11:59 pm Mr MZ's suggestion:

REMOVED
code:
PRNG rng(1070372);

for (Piece pc : Pieces)
for (Square s = SQ_A1; s <= SQ_H8; ++s)
Zobrist::psq[pc] = rng.rand<Key>();

for (File f = FILE_A; f <= FILE_H; ++f)
Zobrist::enpassant[f] = rng.rand<Key>();

for (int cr = NO_CASTLING; cr <= ANY_CASTLING; ++cr)
Zobrist::castling[cr] = rng.rand<Key>();

Zobrist::side = rng.rand<Key>();

https://pixeldrain.com/u/gzRFyz2v

You mean comparing the one suggested by MZ and the one that I made?
If yes, there's not much difference

Or against other sf derivatives such as Tactical and the like?

Tests were based on the first one. It's performance is good. A few testers were involved as well. You can test it for yourself and see

If you see me releasing it means on my tests Raid came up Top.

Introducing Raid v2.74 Recon
No self learning
https://pixeldrain.com/u/B4kygSSP

Happy testing

TR

Good afternoon - What network does this engine use?

IGOR1959 wrote: ↑Sun Sep 10, 2023 6:23 am Good afternoon - What network does this engine use?

nn-c38c3d8d3920.nnue

Raid v2.75 Recon with self learning.
Thank you my beta testers, i appreciate you

you can create any exp file with a Sugar tool or Eman tool etc. and add to raid folder and rename to Raid.exp or let the engine create its own.

https://pixeldrain.com/u/4Ucem2ys

Enjoy

TR

Update Raid v2.76i_X
once again Tnx to my buddies and everyone who helped in testing Raid

https://pixeldrain.com/u/WqUnYz72

Enjoy

Update Raid v2.78i_X
https://pixeldrain.com/u/VimcqiPb

Enjoy

Orgz wrote: ↑Mon Oct 16, 2023 7:00 pm Update Raid v2.78i_X
https://pixeldrain.com/u/VimcqiPb

Enjoy

Sad, sorry!
I wanted to see what's new in Raid and I can't find anything. Only one code in the search was not taken from "Stockfish dev by Stefan Geschwentner" but the old line was kept. I don't know why I should use Raid? The author should (this is my opinion) tell people that Raid = Stockfish dev, and that the only thing implemented is the Polyglot book and the Sugar AI learning file. And this is not the author's own work, but copy & paste from SugaR AI. Compare yourself, I have attached both search.cpp files:

https://pixeldrain.com/u/x2uESANj

Raid v2.78i_X is 99,999% identical in search to Stockfish dev, Author: Stefan Geschwentner Date: Sun Oct 8. The only difference is in Step 20 (search.cpp) where RAID prefers an older Stockfish dev. entry, see at the bottom.

STOCKFISH DEV:

Author: Stefan Geschwentner
Date: Sun Oct 8 07:46:26 2023 +0200
Timestamp: 1696743986

Simplify collection of bad moves for history updates.

Search.cpp: Changes in Step 20 changed (by S. Gschwentner).

SF Gschwentner:

// If the move is worse than some previously searched move, remember it, to update its stats later
if (move != bestMove && moveCount <= 32)
{
if (capture)
capturesSearched[captureCount++] = move;

else
quietsSearched[quietCount++] = move;
}

Raid v2.78i_X:

// If the move is worse than some previously searched move, remember it, to update its stats later
if (move != bestMove)
{
if (capture && captureCount < 32)
capturesSearched[captureCount++] = move;

else if (!capture && quietCount < 64)
quietsSearched[quietCount++] = move;
}


Stockfish dev. previous versions, identical with Raid Raid v2.78i_X:

// If the move is worse than some previously searched move, remember it, to update its stats later
- if (move != bestMove)
+ if (move != bestMove && moveCount <= 32)
{
- if (capture && captureCount < 32)
if (capture)
capturesSearched[captureCount++] = move;

- else if (!capture && quietCount < 64)
+ else
quietsSearched[quietCount++] = move;

Raid v3.1 Blood Sport
https://pixeldrain.com/u/qgPRqGRu