Have you seen this Arena bug?

[Gabor Szots]

This is a partial Arena debug:

2023-07-30 18:45:29,011<--1:info multipv 1 score cp -22 wdl 101 725 174 depth 24 seldepth 34 nodes 7323474 qsNodes 3058334 nps 2206530 time 3319 tbhits 0 hashfull 153 pv e8h8 c1e3 c8e6 d1d3 b8d7 a2a4 a8c8 a4a5 d8c7 f1c1 d7c5 b3c5 d6c5 e2f3 c5c4 d3e2 e7b4 e3b6 c7c6 c1b1 h7h6 h2h3 b4c5 b6c5 c6c5 b1d1 f8d8 c3d5 e6d5
2023-07-30 18:45:29,014<--1:bestmove e8h8
2023-07-30 18:45:29,014*moveerror* Rstart piece>0 and target piece>0, can't capture own piece !...!
2023-07-30 18:45:29,014*1*---------> Arena:Illegal move!: "bestmove e8h8" ()

And this is the PV Arena prints:

9/12 00:00 8k 1 909k -0,33 Ke8xh8 Bc1-e3 Bc8-e6 h2-h3 Nb8-c6 a2-a4 h7-h6 a4-a5 Be6xb3 c2xb3 Nc6xa5 b3-b4
10/14 00:00 15k 2 083k -0,36 Ke8xh8 Bc1-e3 Bc8-e6 Nc3-d5 Nd5xf6 Be7xf6 h2-h3 Nb8-c6 c2-c3 d6-d5 e4xd5
10/11
etc.

So, apparently, the engine sends e8h8 but Arena thinks it was Ke8xh8 and does not allow the capture of an own rook.

The above was seen using Clover 5.0 but I have seen the same by Cheers 0.2 as well. It's odd that other engines do not do it (or at least I didn't notice).

Can someone cast a light on this?

[Ras]

So, apparently, the engine sends e8h8 but Arena thinks it was Ke8xh8 and does not allow the capture of an own rook.

Depending on what the position is, Arena may just be right.

[Ajedrecista]

Hello Gabor:

Assuming normal chess and not FRC, which I do not know how castling notation works, the black castle kingside should not be e8g8 instead of e8h8? The debug window on Arena 2.0.1 GUI with SF 16 after 1.- d4, Nf6; 2.- c4, e6; 3.- Nc3, Bb4; 4.- e3 shows 4.- ..., O-O as e8g8 indeed.

Here are two links that seem to suport my theory:

Description of the Universal Chess Interface (UCI)

UCI castling not recognized by some GUIs

Good luck with the bug hunting!

Regards from Spain.

Ajedrecista.

[Damir]

Why would anyone use a garbage GUI like Arena?

[Uri]

Why would anyone use a garbage GUI like Arena?

Just proves my point that some of the Chess programmers are extremely selfish and self-absorbed people.

Apparently there is a Stockfish cluster using 135,000 processor threads but it has never been made public. Just comes to show how selfish some of the program developers really are.

That's part of the reason why I quit chess.

[Gabor Szots]

Why would anyone use a garbage GUI like Arena?

There are many ignorants like me. Thanks for your help.

[Gabor Szots]

Hello Gabor:

Assuming normal chess and not FRC, which I do not know how castling notation works, the black castle kingside should not be e8g8 instead of e8h8? The debug window on Arena 2.0.1 GUI with SF 16 after 1.- d4, Nf6; 2.- c4, e6; 3.- Nc3, Bb4; 4.- e3 shows 4.- ..., O-O as e8g8 indeed.

Here are two links that seem to suport my theory:

Description of the Universal Chess Interface (UCI)

UCI castling not recognized by some GUIs

Good luck with the bug hunting!

Regards from Spain.

Ajedrecista.

Thanks for the hint. It is probably the e8h8 thing instead of e8g8. I'm going to investigate further.

[Rebel]

Why would anyone use a garbage GUI like Arena?

There are many ignorants like me. Thanks for your help.

[Gabor Szots]

It has been confirmed that indeed the notation e8h8 was the reason of the illegal moves. Clover author has already released a fix which I had not been aware of.