Yace

[Anonymous]

Is there a way where you can make Yace .87 NOT resign at all, just play to the end?

[Daniel Mehrmann]

Is there a way where you can make Yace .87 NOT resign at all, just play to the end?

Please try in yace.ini:

resign off

my yace.ini with some balls:

Code: Select all

log_path log\
; Jedes Spiel eine neue Datei
auto_log
; Alles in eine Datei
; log log\yace.log
; Append logging
; loga log\yace.log
hash 128M
egtb_cache 128M
tbldir e:\chess\tb
sgames yace_icc.pgn
book_learn 7
; noise 500k
; show_eco 2
pos_learn 1
; Yace-Author(C) tells you: 5+3 ist 300 + 60*3
min_learn_tc 350
bbpath bittables\
; book_random 1
autokib 0
say kibitz
max_PV_tellics 0
nmoves_i 1
whisper whisper
announce_mate whisper
resign on
rscore -6
rmoves 3
version experimental - Book by Daniel Mehrmann
; ignore_system 1
overwrite_vgm 1
max_PV_WB 256
book_holes 8
margin_book_learn_white -45
margin_book_learn_black -45
; tb_usage 3
null_high 3.5
; null_low 3.0
; null_endgame 2.0
; null_split 5.5


[Anonymous]

What commands in the yace.ini makes it to where it learns from the games it plays?

Because I have seen Yace play the same exact line, and it loses. book_learn 7 is supposed to make it learn correct? But it doesn't seem like it is doing so.

Dblearn.bin is supposed to be its learn file to where it reads from that file and is supposed to have its eval saying that it was a bad move in this position, so it shouldn't do it, but still does... Maybe thats just the best move for that position?

[Anonymous]

What commands in the yace.ini makes it to where it learns from the games it plays?

Because I have seen Yace play the same exact line, and it loses. book_learn 7 is supposed to make it learn correct? But it doesn't seem like it is doing so.

Dblearn.bin is supposed to be its learn file to where it reads from that file and is supposed to have its eval saying that it was a bad move in this position, so it shouldn't do it, but still does... Maybe thats just the best move for that position?

Depending on the version, book learning might be enabled by default. book_learn 7 should do book learning as expected, as will do book_learn 6. Most people won't need some additional information, that is written into files by 7. Position learning is enabled by default since many versions.

Which time control did you play. For fast Blitz, book learning is disabled by default. You can use

min_learn_tc 180

to enable it for 3 minute and longer games. You can use a smaller number.

Can you post the two games? Did you look at dblearn2.pgn? You should be able to identify any lost game there.

Regards,
Dieter