bizarre behavior by Leela

Discussion of chess software programming and technical issues.

Moderators: hgm, Rebel, chrisw

smcracraft
Posts: 737
Joined: Wed Mar 08, 2006 8:08 pm
Location: Orange County California
Full name: Stuart Cracraft

bizarre behavior by Leela

Post by smcracraft »

Note the illegal e7d8 instead of the correct d7h3

Not to sure what Leela is smoking but it doesn't look good.

position startpos moves e2e4 c7c5 g1f3 d7d6 d2d4 c5d4 f3d4 g8f6 f1d3 b8c6 d4c6 b7c6 e1g1 g7g6 c1g5 f8g7 b1c3 h7h6 g5e3 f6g4 d1d2 g4e3 d2e3 d8b6 e3b6 a7b6 a2a3 e7e6 a1b1 e8e7 c3e2 c8d7 c2c3 g6g5 f2f4 g5f4 e2f4 h6h5 f1f3 c6c5 a1f1 b6b5 f3h3 h5h4 f3e2 c5c4 d3c2 a8g8 f1f4 g7e5 f4h4 e5f6 h4h8 e6e5 h8g8
go movetime 100000
info depth 5 seldepth 9 time 5 nodes 503 score cp -767 hashfull 1000 nps 100600 tbhits 0 pv e7d8 g8f6 d7h3 f6d5 h3e6 d5b6 d8c7 b6d5
info depth 5 seldepth 10 time 9 nodes 1024 score cp -755 hashfull 1000 nps 113777 tbhits 0 pv e7d8 g8f6 d7h3 f6d5 h3e6 d5b6 d8c7 b6d5
info depth 5 seldepth 11 time 130 nodes 6420 score cp -788 hashfull 1000 nps 49384 tbhits 0 pv e7d8 g8f6 d7h3 f6d5 h3e6 d5b6 d8c7 b6d5 c7d7
info depth 6 seldepth 11 time 163 nodes 8715 score cp -797 hashfull 1000 nps 53466 tbhits 0 pv e7d8 g8f6 d7h3 f6d5 h3e6 d5b6 d8c7 b6d5 c7d7
info depth 6 seldepth 12 time 348 nodes 16677 score cp -817 hashfull 1000 nps 47922 tbhits 0 pv e7d8 g8f6 d7h3 f6d5 h3e6 d5b6 d8c7 b6d5 c7d7
info depth 6 seldepth 13 time 423 nodes 19758 score cp -820 hashfull 1000 nps 46709 tbhits 0 pv e7d8 g8f6 d7h3 f6d5 h3e6 d5b6 d8c7 b6d5 c7d7
info depth 6 seldepth 14 time 2070 nodes 74539 score cp -811 hashfull 1000 nps 36009 tbhits 0 pv e7d8 g8f6 d7h3 f6d5 h3e6 d5b6 d8c7 b6d5 c7d7
info depth 7 seldepth 14 time 2418 nodes 83307 score cp -808 hashfull 1000 nps 34452 tbhits 0 pv e7d8 g8f6 d7h3 f6d5 h3g4 d5f6 g4c8 f6d5 c8b7
info depth 7 seldepth 14 time 7454 nodes 144839 score cp -794 hashfull 1000 nps 19431 tbhits 0 pv e7d8 g8f6 d7h3 f6d5 h3g2 d5b6 g2e4 c2e4 b5b4
info depth 7 seldepth 14 time 12492 nodes 155147 score cp -792 hashfull 1000 nps 12419 tbhits 0 pv e7d8 g8f6 d7h3 f6d5 h3g2 d5b6 g2e4 c2e4 b5b4
info depth 7 seldepth 14 time 17493 nodes 163542 score cp -791 hashfull 1000 nps 9348 tbhits 0 pv e7d8 g8f6 d7h3 f6d5 h3g2 d5b6 g2e4 c2e4 b5b4
info depth 7 seldepth 14 time 22591 nodes 172636 score cp -789 hashfull 1000 nps 7641 tbhits 0 pv e7d8 g8f6 d7h3 f6d5 h3g2 d5b6 g2e4 c2e4 b5b4
info depth 7 seldepth 14 time 27593 nodes 181942 score cp -788 hashfull 1000 nps 6593 tbhits 0 pv e7d8 g8f6 d7h3 f6d5 h3e6 d5b6 d8c7 b6d5 c7d7 d5f6
info depth 7 seldepth 14 time 32615 nodes 191503 score cp -786 hashfull 1000 nps 5871 tbhits 0 pv e7d8 g8f6 d7h3 f6d5 h3e6 d5b6 d8c7 b6d5 c7d7 d5f6
info depth 7 seldepth 14 time 37616 nodes 200613 score cp -785 hashfull 1000 nps 5333 tbhits 0 pv e7d8 g8f6 d7h3 f6d5 h3e6 d5b6 d8c7 b6d5 c7d7 d5f6
info depth 7 seldepth 14 time 42716 nodes 210026 score cp -784 hashfull 1000 nps 4916 tbhits 0 pv e7d8 g8f6 d7h3 f6d5 h3e6 d5b6 d8c7 b6d5 c7d7 d5f6
info depth 7 seldepth 14 time 47795 nodes 219544 score cp -783 hashfull 1000 nps 4593 tbhits 0 pv e7d8 g8f6 d7h3 f6d5 h3e6 d5b6 d8c7 b6d5 c7d7 d5f6
info depth 7 seldepth 14 time 52840 nodes 228494 score cp -782 hashfull 1000 nps 4324 tbhits 0 pv e7d8 g8f6 d7h3 f6d5 h3e6 d5b6 d8c7 b6d5 c7d7 d5f6
info depth 7 seldepth 14 time 57855 nodes 237705 score cp -781 hashfull 1000 nps 4108 tbhits 0 pv e7d8 g8f6 d7h3 f6d5 h3e6 d5b6 d8c7 b6d5 c7d7 d5f6
info depth 7 seldepth 14 time 62877 nodes 246801 score cp -780 hashfull 1000 nps 3925 tbhits 0 pv e7d8 g8f6 d7h3 f6d5 h3g2 d5b6 g2e4 c2e4 b5b4 b6d5
info depth 7 seldepth 14 time 67888 nodes 255341 score cp -780 hashfull 1000 nps 3761 tbhits 0 pv e7d8 g8f6 d7h3 f6d5 h3g2 d5b6 g2e4 c2e4 b5b4 b6d5
info depth 7 seldepth 14 time 72910 nodes 264302 score cp -779 hashfull 1000 nps 3625 tbhits 0 pv e7d8 g8f6 d7h3 f6d5 h3g2 d5b6 g2e4 c2e4 b5b4 b6d5
info depth 7 seldepth 14 time 77917 nodes 272974 score cp -778 hashfull 1000 nps 3503 tbhits 0 pv e7d8 g8f6 d7h3 f6d5 h3g2 d5b6 g2e4 c2e4 b5b4 b6d5
info depth 7 seldepth 14 time 82944 nodes 281524 score cp -777 hashfull 1000 nps 3394 tbhits 0 pv e7d8 g8f6 d7h3 f6d5 h3g2 d5b6 g2e4 c2e4 b5b4 b6d5
info depth 7 seldepth 14 time 87949 nodes 290065 score cp -777 hashfull 1000 nps 3298 tbhits 0 pv e7d8 g8f6 d7h3 f6d5 h3g4 d5f6 g4c8 f6d5 c8b7 d5b6 d8c7
info depth 7 seldepth 14 time 92975 nodes 298855 score cp -776 hashfull 1000 nps 3214 tbhits 0 pv e7d8 g8f6 d7h3 f6d5 h3g4 d5f6 g4c8 f6d5 c8b7 d5b6 d8c7
info depth 7 seldepth 14 time 93490 nodes 299765 score cp -776 hashfull 1000 nps 3206 tbhits 0 pv e7d8 g8f6 d7h3 f6d5 h3g4 d5f6 g4c8 f6d5 c8b7 d5b6 d8c7
bestmove e7d8 ponder g8f6
User avatar
Guenther
Posts: 4605
Joined: Wed Oct 01, 2008 6:33 am
Location: Regensburg, Germany
Full name: Guenther Simon

Re: bizarre behavior by Leela

Post by Guenther »

smcracraft wrote: Mon May 27, 2019 8:18 am Note the illegal e7d8 instead of the correct d7h3

Not to sure what Leela is smoking but it doesn't look good.

position startpos moves e2e4 c7c5 g1f3 d7d6 d2d4 c5d4 f3d4 g8f6 f1d3 b8c6 d4c6 b7c6 e1g1 g7g6 c1g5 f8g7 b1c3 h7h6 g5e3 f6g4 d1d2 g4e3 d2e3 d8b6 e3b6 a7b6 a2a3 e7e6 a1b1 e8e7 c3e2 c8d7 c2c3 g6g5 f2f4 g5f4 e2f4 h6h5 f1f3 c6c5 a1f1 b6b5 f3h3 h5h4 f3e2 c5c4 d3c2 a8g8 f1f4 g7e5 f4h4 e5f6 h4h8 e6e5 h8g8
The feeding of your start position is illegal, so may be, we don't know what you are smoking? ;-)
https://rwbc-chess.de

trollwatch:
Chessqueen + chessica + AlexChess + Eduard + Sylwy
Joost Buijs
Posts: 1563
Joined: Thu Jul 16, 2009 10:47 am
Location: Almere, The Netherlands

Re: bizarre behavior by Leela

Post by Joost Buijs »

Guenther wrote: Mon May 27, 2019 8:40 am
smcracraft wrote: Mon May 27, 2019 8:18 am Note the illegal e7d8 instead of the correct d7h3

Not to sure what Leela is smoking but it doesn't look good.

position startpos moves e2e4 c7c5 g1f3 d7d6 d2d4 c5d4 f3d4 g8f6 f1d3 b8c6 d4c6 b7c6 e1g1 g7g6 c1g5 f8g7 b1c3 h7h6 g5e3 f6g4 d1d2 g4e3 d2e3 d8b6 e3b6 a7b6 a2a3 e7e6 a1b1 e8e7 c3e2 c8d7 c2c3 g6g5 f2f4 g5f4 e2f4 h6h5 f1f3 c6c5 a1f1 b6b5 f3h3 h5h4 f3e2 c5c4 d3c2 a8g8 f1f4 g7e5 f4h4 e5f6 h4h8 e6e5 h8g8
The feeding of your start position is illegal, so may be, we don't know what you are smoking? ;-)
This is true, but it is really stupid that Leela accepts an illegal move without noticing it and without complaining about it.
User avatar
Guenther
Posts: 4605
Joined: Wed Oct 01, 2008 6:33 am
Location: Regensburg, Germany
Full name: Guenther Simon

Re: bizarre behavior by Leela

Post by Guenther »

Joost Buijs wrote: Mon May 27, 2019 1:07 pm
Guenther wrote: Mon May 27, 2019 8:40 am
smcracraft wrote: Mon May 27, 2019 8:18 am Note the illegal e7d8 instead of the correct d7h3

Not to sure what Leela is smoking but it doesn't look good.

position startpos moves e2e4 c7c5 g1f3 d7d6 d2d4 c5d4 f3d4 g8f6 f1d3 b8c6 d4c6 b7c6 e1g1 g7g6 c1g5 f8g7 b1c3 h7h6 g5e3 f6g4 d1d2 g4e3 d2e3 d8b6 e3b6 a7b6 a2a3 e7e6 a1b1 e8e7 c3e2 c8d7 c2c3 g6g5 f2f4 g5f4 e2f4 h6h5 f1f3 c6c5 a1f1 b6b5 f3h3 h5h4 f3e2 c5c4 d3c2 a8g8 f1f4 g7e5 f4h4 e5f6 h4h8 e6e5 h8g8
The feeding of your start position is illegal, so may be, we don't know what you are smoking? ;-)
This is true, but it is really stupid that Leela accepts an illegal move without noticing it and without complaining about it.
Ofc that's not nice, but how often a user manually inputs wrong moves (manually created too!) that way, while a GUI would do a sanity check for the engine?

I did a quick test on some uci programs to record their reaction. 50% detect and report an error, the other half starts analysing
after the last legal move w/o any message. SF is in the second group.
https://rwbc-chess.de

trollwatch:
Chessqueen + chessica + AlexChess + Eduard + Sylwy
mar
Posts: 2554
Joined: Fri Nov 26, 2010 2:00 pm
Location: Czech Republic
Full name: Martin Sedlak

Re: bizarre behavior by Leela

Post by mar »

Guenther wrote: Mon May 27, 2019 2:27 pm
Joost Buijs wrote: Mon May 27, 2019 1:07 pm This is true, but it is really stupid that Leela accepts an illegal move without noticing it and without complaining about it.
Ofc that's not nice, but how often a user manually inputs wrong moves (manually created too!) that way, while a GUI would do a sanity check for the engine?

I did a quick test on some uci programs to record their reaction. 50% detect and report an error, the other half starts analysing
after the last legal move w/o any message. SF is in the second group.
Agreed, UCI is not intended for humans to input bad move sequence.
The best you can do in UCI is to output info string anyway, it would be nice to report an error at least this way.
I don't do this in Cheng, so it's not fully idiot-proof. Not sure if stupid is the right word in this case,
but I should've probably done that.

Besides - for analysis from command line, I prefer xboard mode over UCI
Martin Sedlak
mar
Posts: 2554
Joined: Fri Nov 26, 2010 2:00 pm
Location: Czech Republic
Full name: Martin Sedlak

Re: bizarre behavior by Leela

Post by mar »

mar wrote: Mon May 27, 2019 4:09 pm I don't do this in Cheng, so it's not fully idiot-proof.
Correction: Cheng does report an error (not the offending move though) - I fell for the msc stdlib bug again, where you have to manually enter twice. Sigh...
Martin Sedlak
Joost Buijs
Posts: 1563
Joined: Thu Jul 16, 2009 10:47 am
Location: Almere, The Netherlands

Re: bizarre behavior by Leela

Post by Joost Buijs »

mar wrote: Mon May 27, 2019 5:44 pm
mar wrote: Mon May 27, 2019 4:09 pm I don't do this in Cheng, so it's not fully idiot-proof.
Correction: Cheng does report an error (not the offending move though) - I fell for the msc stdlib bug again, where you have to manually enter twice. Sigh...
Isn't this bug already solved a long time ago?
mar
Posts: 2554
Joined: Fri Nov 26, 2010 2:00 pm
Location: Czech Republic
Full name: Martin Sedlak

Re: bizarre behavior by Leela

Post by mar »

Joost Buijs wrote: Mon May 27, 2019 6:54 pm Isn't this bug already solved a long time ago?
You're probably right, it only seems to happen when I compile against older Win SDK (8.1 or earlier).
It also happens when I try to compile for WinXP, no idea if it's related.
Martin Sedlak
smcracraft
Posts: 737
Joined: Wed Mar 08, 2006 8:08 pm
Location: Orange County California
Full name: Stuart Cracraft

Re: bizarre behavior by Leela

Post by smcracraft »

mar wrote: Mon May 27, 2019 7:23 pm
Joost Buijs wrote: Mon May 27, 2019 6:54 pm Isn't this bug already solved a long time ago?
You're probably right, it only seems to happen when I compile against older Win SDK (8.1 or earlier).
It also happens when I try to compile for WinXP, no idea if it's related.
This is running on an Apple Mac.

Does illegal moves.

Not good.
User avatar
hgm
Posts: 27790
Joined: Fri Mar 10, 2006 10:06 am
Location: Amsterdam
Full name: H G Muller

Re: bizarre behavior by Leela

Post by hgm »

Demons can fly out of your nose, and such...