Ltchess3

[laurietunnicliffe]

Hey folks.
You might like to have a look at ltchess3.
I have rewritten ltchess2 to use nibble board representation (just for fun) and simplifier the Eval to be
just PST blended with history scores and only double pawns.
It seems to out play ltchess2.
It can be downloaded from weebley.

[Aldus]

From your site:

To compile it use either freepascal or virtualpascal.

Virtual Pascal compilation error:

Error 3: Unknown identifier 'QWord'.

Also, why not put an executable yourself, for those who just want to test it without downloading and installing a Pascal compiler?

[mvanthoor]

Hey folks.
You might like to have a look at ltchess3.
I have rewritten ltchess2 to use nibble board representation (just for fun) and simplifier the Eval to be
just PST blended with history scores and only double pawns.
It seems to out play ltchess2.
It can be downloaded from weebley.

Hi Congratulations on the new version. However, if you want something to look at it, you'd better provide an executable for download, at the minimum. And even then, people might not trust the engine. If it's open source, it would be best to put the source code on GitHub or GitLab, in a public repository, and provide (at least) compilation instructions there.

[laurietunnicliffe]

OK, I have now added a linux exe.

[Aldus]

OK, I have now added a linux exe.

I managed to compile it for Windows with fpc 3.2.2.

Your program seems to have a bug. With 5 seconds/move, in the following position:

[d]4B3/8/1p2bN2/3pR3/8/5kP1/3n1P2/6K1 b - - 1 51
it tries to move Kg2 (f3g2), which is illegal.

[laurietunnicliffe]

Thanks,
I can reproduce that, and will fix it today.

[laurietunnicliffe]

Fixed and uploaded.

Thanks for finding that.

[Aldus]

You're welcome.

You may want to implement a perft function and test a set of standard positions, to check if there are no errors left in the move generator routines.

See here:

https://www.chessprogramming.org/Perft_Results

and here:

https://www.chessprogramming.org/Perft

Edit: I see you have a perft function, maybe you could find more positions to test. Some engines have a whole bunch of them, with published results.

[Aldus]

It seems to out play ltchess2.
It can be downloaded from weebley.

Can you provide the source code (and binary) for ltchess1 and ltchess2?

[laurietunnicliffe]

If you PM me I can send ltchess2 (0X88 board) and
ltchessbb (bitboards for pawns)

Laurie