• Questions about running bot vs bot experiments

    From MK@21:1/5 to All on Fri Oct 29 15:21:28 2021
    I have never tried such experiments but the idea crossed
    my mind lately. Could anyone help with telling us all how
    it's done and share the tools used to do it?

    I became curious about what dice is used in games between
    two different bots, between a bot and its mutant or even the
    same bot against itself.

    Is it manual dice fed to the bots through simulated keyboard
    input? (Which I suppose would be the preferable way.)

    I wonder if different bots would do better using their own dice
    and if anyone has tested for this?

    Also, assuming for the sake of the argument that bots cheat,
    with the good and noble intentions of expedited teaching of
    course ;) but if for some other reason let's say only if they
    have to cheat when they begin to fall behind by too much, etc.
    I wonder if many short sessions vs a single very long session
    would make a difference?

    I will appreciate any help and suggestions...

    MK

    BTW: anyone wanting to claim to be a good analys would test
    for all these kinds of things assumed to be unlikely. For example,
    Nasti's tests with an external ddl's for XG that rolled always the
    same number or never rolled a double were good but amateur
    attempts if they only tested for a few dozen rolls, etc. But since
    we're not going to try landing a bg bot on Mars, such efforts may
    not go beyond time passing brain exercises...

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Axel Reichert@21:1/5 to murat@compuplus.net on Sat Oct 30 22:35:08 2021
    MK <murat@compuplus.net> writes:

    I have never tried such experiments but the idea crossed
    my mind lately. Could anyone help with telling us all how
    it's done and share the tools used to do it?

    In short: Have GNU Backgammon play in text mode, parse the output, feed
    it to a bot your wrote yourself and send your move/cube action back to
    GNU Backgammon. My experiments with the mutant were trivial, since it
    did simplistic cube action after asking GNU Backgammon for a hint and
    blatantly copied its checker play. No, I will not share my code ("exotic programming language, and I am not eager for support nightmares).

    what dice is used in games between two different bots

    GNU Backgammon uses Mersenne Twister by default, which is fine. GNU
    Backgammon is open source, just read the code. I am no dice paranoid and
    will not support any dice paranoids. My bot gets the dice from GNU
    Backgammon and tries to make the most of it.

    Is it manual dice fed to the bots through simulated keyboard
    input? (Which I suppose would be the preferable way.)

    Hell, no, 3000 games! I am neither paranoid nor crazy.

    I wonder if different bots would do better using their own dice
    and if anyone has tested for this?

    Probably not. The people here who would be up to the task are no dice paranoids, and the dice paranoids for some reason never deliver.

    they have to cheat when they begin to fall behind by too much

    "Read the source, Luke", to find out the facts. This is not cloud cuckoo
    land.

    Axel

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From MK@21:1/5 to Axel Reichert on Sat Oct 30 15:24:25 2021
    On October 30, 2021 at 2:35:10 PM UTC-6, Axel Reichert wrote:

    MK <mu...@compuplus.net> writes:

    I have never tried such experiments but the idea crossed
    my mind lately. Could anyone help with telling us all how
    it's done and share the tools used to do it?

    In short: Have GNU Backgammon play in text mode, .....

    I was thinking about not just gnubg against itself but any
    bot against any other bot.

    No, I will not share my code ("exotic programming language,
    and I am not eager for support nightmares).

    There was once a tool called "Dueller" which was free and
    worked with Jellyfish, Snowie and Gnubg. Perhaps someone
    may have a similar tool that would work with current bots
    also that he may be willing to share.

    what dice is used in games between two different bots

    GNU Backgammon uses Mersenne Twister by default, which
    is fine. GNU Backgammon is open source, just read the code.
    I am no dice paranoid and will not support any dice paranoids.

    Jumping to conclusions from mere questions, lecturing from
    high and misdiagnosing must be your own dogmatism. I am
    a catmatist myself and I don't particularly like mutts sniffing
    each orher's butt in circularity. To each his own...

    Is it manual dice fed to the bots through simulated keyboard
    input? (Which I suppose would be the preferable way.)

    Hell, no, 3000 games! I am neither paranoid nor crazy.

    Sorry, this was my terribly wording it. I meant each bot is set
    to accept manual dice and an external process will pass dice
    rolls to them, through keyboard events, when they are waiting
    for dice input.

    One Michael here had created such a tool to feed encrypted
    external dice rolls to XG, as part of a test for me to prove to
    him that I could really beat XG, but it was abandoned because
    of too many problems with timing and/or other issues. Maybe
    he can modify his code to feed dice to two processes taking
    turns. Is he still around in RGB or other forums?

    I wonder if different bots would do better using their own dice
    and if anyone has tested for this?

    Probably not. The people here who would be up to the task are
    no dice paranoids,

    Speak for yourself and try to overcome your own dogmatism.

    Some people were open minded enough to acknowledge that
    bots' decisions based on their own equity calculations could
    be biased on what dice algorythm they were trained with.

    and the dice paranoids for some reason never deliver.

    Of course they do. Some people will remember Jellyfish not
    only rolling but playing 7's, Snowie's rejecting perfectly valid
    manual dice numbers, how modifying a bot's dice roller also
    changed how it played, etc.

    Some of such may even be "honest" ;) bugs that people didn't
    know existed in the code they stole from others, no? :)

    they have to cheat when they begin to fall behind by too much

    "Read the source, Luke", to find out the facts. This is not cloud
    cuckoo land.

    This is a small planet of inbred mentally ill gamblers who cater
    to one another. No outsider observer would be a naive believer
    that people capable of killing their wives, selling their daughters,
    stealing boards and equipment at tournaments, cheating with
    rigged dice, knowingly shuffling checkers, erasing and modifying
    match scores, etc. would not be tempted and capable of developing
    cheating bots.

    You come down from your high horse on the clouds, put your feet
    on solid ground and start walking...

    MK

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)