There is a problem in 'cmd.c' . I see you made some changes here, but it causes the windows build to crash when run >
OK, I just rewrote that function because the old one was from some old BSD code, and I wasn't sure if I could GPL it. Simple fix, just put *string instead of string on the first line where it checks if its NULL.
1. Occasionally engine choses to play opponents move instead of it's own.
2. Some time control problems. Over 40 moves engine started playing moves instantly.
1. is strange, maybe another input bug. Is ponder on? Do you have a log of any examples?
2. I think I don't have all of the time control parsing implemented in "go", so I imagine it thinks it has 0 time. I'll investigate.
2. Forced ZCT to play in uci mode.
This is very strange if it is necessary to make it work. You are setting the protocol to UCI, but you don't call initialize_cmds(), so it will continue parsing input against the default protocol. Maybe try sending me a logfile with and without it?
There is a problem in 'cmd.c' . I see you made some changes here, but it causes the windows build to crash when run >
OK, I just rewrote that function because the old one was from some old BSD code, and I wasn't sure if I could GPL it. Simple fix, just put *string instead of string on the first line where it checks if its NULL.
1. Occasionally engine choses to play opponents move instead of it's own.
2. Some time control problems. Over 40 moves engine started playing moves instantly.
1. is strange, maybe another input bug. Is ponder on? Do you have a log of any examples?
2. I think I don't have all of the time control parsing implemented in "go", so I imagine it thinks it has 0 time. I'll investigate.
2. Forced ZCT to play in uci mode.
This is very strange if it is necessary to make it work. You are setting the protocol to UCI, but you don't call initialize_cmds(), so it will continue parsing input against the default protocol. Maybe try sending me a logfile with and without it?
Here's the end of the log where ZCT plays illegal move. This is ponder off - uci mode. This is first game I just played to get a log. It looks like a ponder bug, but game is played ponder off.
There is a problem in 'cmd.c' . I see you made some changes here, but it causes the windows build to crash when run >
OK, I just rewrote that function because the old one was from some old BSD code, and I wasn't sure if I could GPL it. Simple fix, just put *string instead of string on the first line where it checks if its NULL.
1. Occasionally engine choses to play opponents move instead of it's own.
2. Some time control problems. Over 40 moves engine started playing moves instantly.
1. is strange, maybe another input bug. Is ponder on? Do you have a log of any examples?
2. I think I don't have all of the time control parsing implemented in "go", so I imagine it thinks it has 0 time. I'll investigate.
2. Forced ZCT to play in uci mode.
This is very strange if it is necessary to make it work. You are setting the protocol to UCI, but you don't call initialize_cmds(), so it will continue parsing input against the default protocol. Maybe try sending me a logfile with and without it?
Here's the end of the log where ZCT plays illegal move. This is ponder off - uci mode. This is first game I just played to get a log. It looks like a ponder bug, but game is played ponder off.
It's a buffer overflow. I suppose BUFSIZ is only 512 bytes on Windows. This could be a simple fix for just increasing the buffer size, but I'd like to elegantly handle any size buffer as well as any amount of arguments. I'll see what I can do later today.
There is a problem in 'cmd.c' . I see you made some changes here, but it causes the windows build to crash when run >
OK, I just rewrote that function because the old one was from some old BSD code, and I wasn't sure if I could GPL it. Simple fix, just put *string instead of string on the first line where it checks if its NULL.
1. Occasionally engine choses to play opponents move instead of it's own.
2. Some time control problems. Over 40 moves engine started playing moves instantly.
1. is strange, maybe another input bug. Is ponder on? Do you have a log of any examples?
2. I think I don't have all of the time control parsing implemented in "go", so I imagine it thinks it has 0 time. I'll investigate.
2. Forced ZCT to play in uci mode.
This is very strange if it is necessary to make it work. You are setting the protocol to UCI, but you don't call initialize_cmds(), so it will continue parsing input against the default protocol. Maybe try sending me a logfile with and without it?
Right, here are the logs for loading ZCT into ChessBase.
First the fixed one with forced uci mode and fixed id which works:
ZCT Version 0.3.2443 JA
EgbbProbe failed to load! egbbdll.dll
Trying opening book book.zbk... successful.
trans hash size set to 32M
qsearch hash size set to 1M
pawn hash size set to 1M
eval hash size set to 2M
(zct)1. uci
id name ZCT0.3.2443 JA
author Zach Wegner
uciok
isready
readyok
position startpos
setoption name MultiPV value 1
go infinite
stop
No move.
bestmove b1c3 ponder b8c6
position startpos moves f2f4
d7d5: games=5 winp=70.0%
bestmove d7d5 ponder b8c6
setoption name MultiPV value 1
d7d5: games=5 winp=70.0%
bestmove d7d5 ponder b8c6
go infinite
stop
bestmove d7d5 ponder b8c6
d7d5: games=5 winp=70.0%
bestmove d7d5 ponder b8c6
quit
Using just id fix alone without forced uci doesn't work. They both need to be in.
Hmm... interesting. I think I might know what's going on, but I'm not positive. I'll get it fixed later today. I'll just send you an email about it when I do, not necessary to keep flooding CCC.
I also see that you were doing some tuning. Modifying the source is not necessary. Just put this in ZCT.ini:
12 Kf2 looked a bit weird,but it managed to win!
Sometimes it plays a book move,but none in this game.
Seems i'll have to play around with that a bit more.
Wow, great game, thanks. Kf2 looks like a pretty bad move, my king safety/castling evaluation is kind of messed up. At that point, black looks to have a strong attack building up. 13...g5 is a nice looking pawn sac to move the knight into action, but I think it was the fatal blunder. I love how after 20.gxh6 ZCT's eval shoots up, and it takes Bright 12 moves to fully realize the danger. This probably is an example of both ZCT's deeper searching making up for crappy eval, and likely a too aggressive passed pawn eval. The game was very entertaining though, with three promotions.
My little engine looks like it has some potential, once I get the eval tuned up.
And about the book, I think Jim made the book with the "win percent" turned up too high, so only moves which are strongly winning are stored in the book. These are scattered all around the tree. I'd recommend setting win_percent either to 0 or to a low value around 40 when making books. Too many good book moves can get abandoned, it's not a very good metric... I'll try to get some PGNs soon so I can make a nice solid book.
Zach Wegner wrote:Wow, great game, thanks. Kf2 looks like a pretty bad move, my king safety/castling evaluation is kind of messed up. At that point, black looks to have a strong attack building up. 13...g5 is a nice looking pawn sac to move the knight into action, but I think it was the fatal blunder. I love how after 20.gxh6 ZCT's eval shoots up, and it takes Bright 12 moves to fully realize the danger. This probably is an example of both ZCT's deeper searching making up for crappy eval, and likely a too aggressive passed pawn eval. The game was very entertaining though, with three promotions.
My little engine looks like it has some potential, once I get the eval tuned up.
And about the book, I think Jim made the book with the "win percent" turned up too high, so only moves which are strongly winning are stored in the book. These are scattered all around the tree. I'd recommend setting win_percent either to 0 or to a low value around 40 when making books. Too many good book moves can get abandoned, it's not a very good metric... I'll try to get some PGNs soon so I can make a nice solid book.
About the book: I cannot blame Jim(probably only myself ) as I generated one from Olivier Deville's pgn collection.
I used
Zach Wegner wrote:Wow, great game, thanks. Kf2 looks like a pretty bad move, my king safety/castling evaluation is kind of messed up. At that point, black looks to have a strong attack building up. 13...g5 is a nice looking pawn sac to move the knight into action, but I think it was the fatal blunder. I love how after 20.gxh6 ZCT's eval shoots up, and it takes Bright 12 moves to fully realize the danger. This probably is an example of both ZCT's deeper searching making up for crappy eval, and likely a too aggressive passed pawn eval. The game was very entertaining though, with three promotions.
My little engine looks like it has some potential, once I get the eval tuned up.
And about the book, I think Jim made the book with the "win percent" turned up too high, so only moves which are strongly winning are stored in the book. These are scattered all around the tree. I'd recommend setting win_percent either to 0 or to a low value around 40 when making books. Too many good book moves can get abandoned, it's not a very good metric... I'll try to get some PGNs soon so I can make a nice solid book.
Hi Zach,
Whenever there is a good release version,buzz me and leave the opening book issue for me....just send me the instructions what exactly to do
_No one can hit as hard as life.But it ain’t about how hard you can hit.It’s about how hard you can get hit and keep moving forward.How much you can take and keep moving forward….