1. Forum
  2. Usenet
  3. REC.GAMES.BACKGAMMON

  • Alpha-Experiment suggestions for bot developers, especially Gnubg team.

    From MK@21:1/5 to Philippe Michel on Wed Oct 13 21:57:46 2021
    On October 13, 2021 at 3:20:38 PM UTC-6, Philippe Michel wrote:

    ..... someone claiming he can consistently beat entities even stronger
    and more equanimous than Mochy with some mysterious style of play
    everyone else is too dumb to grasp.

    Why don't some of you with math and programming knowledges stop
    being lazy ass deniers byt try to make a little effort to test for your own sakes, whether your "cube skill" based on Jackoffski formulas and your
    "checker skills" based on equestrian stables are accurate or not.

    With Gnubg being already capable of playing 1, 2, 3 checker hypergammon,
    people like Philippe Michel are best positioned to run such experiments.

    The playing code is already there. So, you are already half way there.
    Just train the bot differently and see what happens.

    My hypestgammon is perfect as a first test, better than Gnubg's 1checker hypergammon which uses 2 dice and thus still has room for some checker
    skill in playing the conbination of numbers rolled in alternative ways. In my hypestgammon there is zero checker skill. It has purely cube skill. So, what better way to test your "doubling window", etc. fantasies...?

    Of course, even though experimenting with an alpha-hypestgammon will
    give you an initial clue, I realize it's too far from the regular backgammon.

    I have a suggestion fo that also. Since you guys have mastery of Gnubg
    code, it should be trivial to create a "tweaked" version of it that always doubles if its opponent opens with 63 and plays 24/18 13/11, with not
    other changes at all. Then let it play a large number of long matches
    against Gnubg and see what happens...?

    I'm perfectly willing to live with the results of such experiments but I the pain for you guys might be unbearable. And so, I understand why you all
    resist even giving it a try...

    MK

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Timothy Chow@21:1/5 to All on Thu Oct 14 21:03:21 2021
    On 10/14/2021 12:57 AM, MK wrote:
    My hypestgammon is perfect as a first test, better than Gnubg's 1checker hypergammon which uses 2 dice and thus still has room for some checker
    skill in playing the conbination of numbers rolled in alternative ways. In my hypestgammon there is zero checker skill. It has purely cube skill. So, what better way to test your "doubling window", etc. fantasies...?

    You're the lazy one. I already posted the details of the solved
    hypestgammon bot. So go ahead. Knock yourself out. Invent your
    own bot and play it against the solved bot and see if you can get
    an advantage.

    ---
    Tim Chow

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From MK@21:1/5 to Tim Chow on Sat Oct 16 02:11:16 2021
    On October 14, 2021 at 7:03:25 PM UTC-6, Tim Chow wrote:

    You're the lazy one. I already posted the details of the solved
    hypestgammon bot. So go ahead. Knock yourself out. Invent your
    own bot and play it against the solved bot and see if you can get
    an advantage.

    It's not because of my being lazy. Since 1996, hard to belive for 25
    years, not only did I post here in RGB but I played literally tens of
    thousands of matches against Jellyfish, Snowie, Gnubg and XG,
    "tested" them, discovered numerous serious bugs with them that
    had eluded everyone else, ran hundreds of experiments, etc.

    Of those experiments, many ended up proving nothing but dozens
    of them have been revealing. I made a web site, spending quite a
    lot of time to make it look nice, easy to follow and consistent and
    posted dozens of my experiments against various bots at:

    https://www.montanaonline.net/backgammon/

    Each experiment is described in a text file, as to why and how it
    was conducted. I shared the results in raw game records, tabulated
    spreadsheets and charts. I also posted after-the-fact comments on
    most of my experiments here in RGB.

    In responding to suggestions that those were fabricated and/or
    cherry-pickes sessions, I had dared any of you to do the same just
    to prove that they could fabricate similar long sessions in less than
    years. None of you came forward.

    Then, I went to the time consuming effort of playing long sessions
    against bots while screen-capturing in real time and posted the
    video files on youtube at my channel here:

    https://www.youtube.com/user/benboru1/videos

    None of you have ever dared to suggest that the videos had been
    manipulated in any way. But none of you had the decency to
    acknowledge the results.

    If I had done what you suggest, you wouldn't have accepted the
    results of my alpha-bot experiments either.

    That's why I suckered you (and "your colleague") into running your
    own experiments, to prove it to yourselves, in baby steps, starting
    my hypestgammon.

    You pretentious rocket-scientist couldn't even carry that little
    stupid experiment to completion.

    And we all know why...

    You can run but you can't hide. And it's all a metter of time...!

    MK

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Timothy Chow@21:1/5 to All on Sat Oct 16 20:58:01 2021
    On 10/16/2021 5:11 AM, MK wrote:
    On October 14, 2021 at 7:03:25 PM UTC-6, Tim Chow wrote:

    You're the lazy one. I already posted the details of the solved
    hypestgammon bot. So go ahead. Knock yourself out. Invent your
    own bot and play it against the solved bot and see if you can get
    an advantage.

    It's not because of my being lazy.

    [lots of stuff deleted]

    In short, you're lazy.

    ---
    Tim Chow

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From MK@21:1/5 to Tim Chow on Mon Oct 18 01:05:57 2021
    On October 16, 2021 at 6:58:04 PM UTC-6, Tim Chow wrote:

    On 10/16/2021 5:11 AM, MK wrote:

    It's not because of my being lazy.

    [lots of stuff deleted]
    In short, you're lazy.

    Even on inconsequential stuff, you must feel a need to be
    right. Poor you. Sad... :(

    MK

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Timothy Chow@21:1/5 to All on Mon Oct 18 22:09:58 2021
    On 10/18/2021 4:05 AM, MK wrote:
    On October 16, 2021 at 6:58:04 PM UTC-6, Tim Chow wrote:

    On 10/16/2021 5:11 AM, MK wrote:

    It's not because of my being lazy.

    [lots of stuff deleted]
    In short, you're lazy.

    Even on inconsequential stuff, you must feel a need to be
    right. Poor you. Sad... :(

    Wow, so R.B. Sahi's true identity is Donald Trump! Whodathunk.

    ---
    Tim Chow

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