James, the documentation is a bit sprinkled over the source code, so it will involve some searching. Most of these options can be found if you grep for 'mandated_' inside of the source code. For example, some of the implications of depth are documented within the function set_depth_values inside of engine/utils.c. If you set one of these toggles from the command line when launching the client, then it will override one of the mandated_ variables (so if you set depth, it will override mandated_depth which would have otherwise been set to -1 and inherited from the level set itself).
>From what I've seen of some of the other ones like KataGo, they might be able to better mimic more realistic lower ranks, so depending on your ultimate goal it might indeed be an alternative to explore. Having said that, I'm chatting with folks from GNU to see if we can get development of GNU Go a bit more active again, and hope to have good news there soon. Best regards, On Tue, Aug 1, 2023 at 10:20 AM James Dempsey <james.e.demp...@gmail.com> wrote: > Luis, do you know where the minimum values of the "options affecting > strength and speed" you linked me to are documented? I couldn't determine > from the source code what values would put GNU Go 3.8 at its weakest rank. > > GNU Go 2.6 does not support GTP, so I've scratched that idea. > > Were you implying that I'd have better luck with a different engine (e.g., > KataGo) if I wanted to control rank? > > On Sat, Jul 22, 2023 at 10:38 AM James Dempsey <james.e.demp...@gmail.com> > wrote: > >> Ah, thanks Luis, somehow I missed that level of granularity given by >> those options. >> >> I would wonder, though, if `--level` already presets those, and at least >> from running GNU Go bots on KGS where I've observed `--level 0` and >> `--level 10` both settle at 4k, would they really make much difference in >> strength/rank? >> >> There's a GNU Go 2.0 bot on KGS that plays at 11k, which is more what I'm >> looking for, FYI. >> >> Thank you. >> >> On Fri, Jul 21, 2023 at 5:59 PM Luis Felipe Strano Moraes < >> luis.str...@gmail.com> wrote: >> >>> James, >>> >>> being quite honest, it will be difficult for you to toggle difficulty by >>> going to older versions. Likely your best bet if you really want to stay >>> with the GNU Go engine, would be to toy around with some of configurations >>> that are also exposed (which level in some cases just defines a preset for) >>> here: https://www.gnu.org/software/gnugo/gnugo_3.html#SEC31 (see >>> section Other options affecting strength and speed) >>> >>> Best regards, >>> >>> >>> On Fri, Jul 21, 2023 at 2:55 PM James Dempsey <james.e.demp...@gmail.com> >>> wrote: >>> >>>> Thanks Luis, I replaced `-lang-c89` with `-std=c90` and that let me >>>> compile successfully. >>>> >>>> I wanted to try compiling a weaker version :). Even with setting >>>> `--level` for 3.8 I see no real difference in strength/rank. >>>> >>>> On Fri, Jul 21, 2023 at 4:47 PM Luis Felipe Strano Moraes < >>>> luis.str...@gmail.com> wrote: >>>> >>>>> James, >>>>> >>>>> is there any reason why you are trying to build even older versions? >>>>> >>>>> The error there is that that specific compiler flag is not valid for >>>>> this version of GCC. I'd guess you would have to look at how to switch it >>>>> for something else to see if it would still build, this seems to be >>>>> implying it should be using the C89 standard, which you can do in >>>>> different >>>>> ways now: >>>>> https://gcc.gnu.org/onlinedocs/gcc/Standards.html >>>>> "-ansi, -std=c90 or -std=iso9899:1990" >>>>> >>>>> I'm trying to chat with folks from GNU to see if we can sort out >>>>> making a new release of GNU Go to at the very least solve some compilation >>>>> issues and make it build again, hopefully have something happen soon on >>>>> it. >>>>> >>>>> Best regards, >>>>> >>>>> >>>>> On Fri, Jul 21, 2023 at 6:08 AM James Dempsey < >>>>> james.e.demp...@gmail.com> wrote: >>>>> >>>>>> Hello, >>>>>> >>>>>> I'm experiencing another compilation issue, this time for GNU Go 2.6 >>>>>> on Debian GNU/Linux 11 (bullseye). >>>>>> >>>>>> gcc is >>>>>> >>>>>> gcc (Debian 10.2.1-6) 10.2.1 20210110 >>>>>> >>>>>> >>>>>> Error is: >>>>>> >>>>>> gcc -DHAVE_CONFIG_H -I. -I. -I.. -I../engine -I../interface -g -O2 >>>>>>> -Wall -W -Wpointer-arith -Wbad-function-cast -Wcast-qual -Wcast-align >>>>>>> -Wwrite-strings -Wstrict-prototypes -Wmissing-prototypes >>>>>>> -Wmissing-declarations -Wp,-lang-c89 -c sgf.c >>>>>>> cc1: error: command-line option ‘-lang-c89’ is valid for the driver >>>>>>> but not for C >>>>>>> make[2]: *** [Makefile:155: sgf.o] Error 1 >>>>>>> make[1]: *** [Makefile:158: all-recursive] Error 1 >>>>>>> make: *** [Makefile:306: all-recursive-am] Error 2 >>>>>> >>>>>> >>>>>> One again, not a C programmer and couldn't exactly make sense of the >>>>>> answers I had searched for. Any help would be appreciated. >>>>>> >>>>>> Thank you. >>>>>> _______________________________________________ >>>>>> gnugo-devel mailing list >>>>>> gnugo-devel@gnu.org >>>>>> https://lists.gnu.org/mailman/listinfo/gnugo-devel >>>>>> >>>>> >>>>> >>>>> -- >>>>> Luís Felipe Strano Moraes >>>>> _______________________________________________ >>>>> gnugo-devel mailing list >>>>> gnugo-devel@gnu.org >>>>> https://lists.gnu.org/mailman/listinfo/gnugo-devel >>>>> >>>> _______________________________________________ >>>> gnugo-devel mailing list >>>> gnugo-devel@gnu.org >>>> https://lists.gnu.org/mailman/listinfo/gnugo-devel >>>> >>> >>> >>> -- >>> Luís Felipe Strano Moraes >>> _______________________________________________ >>> gnugo-devel mailing list >>> gnugo-devel@gnu.org >>> https://lists.gnu.org/mailman/listinfo/gnugo-devel >>> >> _______________________________________________ > gnugo-devel mailing list > gnugo-devel@gnu.org > https://lists.gnu.org/mailman/listinfo/gnugo-devel > -- Luís Felipe Strano Moraes
_______________________________________________ gnugo-devel mailing list gnugo-devel@gnu.org https://lists.gnu.org/mailman/listinfo/gnugo-devel
