• Some thought on the fascinating and elusive "luck".

    From MK@21:1/5 to Axel Reichert on Mon Aug 15 22:47:56 2022
    On August 13, 2022 at 7:06:39 AM UTC-6, Axel Reichert wrote:

    b...@panix.com (Bradley K. Sherman) writes:

    | This is the paradox of luck and skill: the more skilled the
    | competitors are, the more the outcome is determined by
    | luck.

    Add-on: If two clueless backgammon players compete,
    the outcome is also determined mostly by luck.

    If "clueless play" means "random play" (or near-random
    play), according to this definition of luck, if two clueless
    chess players compete, the outcome is also determined
    entirely (or mostly) by luck as well.

    Yet, that wouldn't make chess a game of luck because if
    the game is analysed without knowing that it was played
    by clueless players and tracing backwards from the final
    position to the opening move, one may find that there
    were many skillful moves made and that the player who
    made more of those skillful moves did indeed win.

    My point is that how a game is played in certain ways
    doesn't change its nature (i.e. a game of skill or luck).

    Therefore, because backgammon is played with dice, it's a
    game of luck by nature and no particular ways of playing
    it will ever turn it into a game of skill.

    I have always disagreed with how "luck" was defined and
    calculated by gamblegammon players and/or bots. Many
    times, I tried to start discussions using more creative and
    comprehensive approaches to defining and quantifying
    "luck" with very limited results or no success at all. Based
    on the above comments, I want to give it one more try...

    Regarding backgammon, Bradley's quote would apply to
    a bot playing against itself "perfectly". Although I totally
    reject bots being perfect, let's assume it to be so for the
    sake of the argument here and accept it as the upper limit
    of skill, i.e. 100%.

    Similarly, what Axel said would establish the lower limit
    of skill, i.e. (*near*) 0%.

    Thus, it's clear that in terms of absolutes, backgammon is
    a game of luck with some relative skill occuring in between.

    Skill "happens" in backgammon (or in any other game of
    luck) just as luck "happens" in chess (or in any other game
    of skill).

    Above, I said *near* because even in a randomly played
    backgammon game, some amount of skill will "happen".
    In other words, even a "clueless" player will win more
    than 0% against a "perfect" player.

    Thus, the "real" skill begins at some higher level than 0%,
    i.e. the "background skill" or "ambient skill" but we have
    no idea where it begins. For years and years I suggested
    that we make a "random bot" play against a "perfect bot"
    to find out what that number is but nobody has shown
    any interest. (Why are you all so fearful of knowledge??)

    Now, let's go back to Axel's comment and talk a little
    about "luck" itself. What if his two players are clueless
    about luck also? If a tree falls in a forest and no one is
    around to hear, does it still make a noise?

    Once I had fun pretending that I was trying to enact a new
    law making luck a punishable crime... :)

    Surely I would have to take very seriously the definition of
    luck and punishments proportionate to its severity. I would
    have to consider as many aspects of it as possible, such as:
    ignorance, intent, sanity, attempt, failing, repeating, petty,
    misdemeanor, felony, premeditated, aggravated, etc.

    For example, if a player clueless about luck gets a lucky roll,
    the falling dice wouldn't make noise unless observed by an
    opponent or a spectator and he wouldn't be guilty. Even if
    there were witnesses, he would be unfit for punishment.

    A player who is "clued" about bogus "temperature maps" of
    bots and who gets a lucky roll without wishing for a specific
    roll would only be guilty of petty luck.

    A player who gets two doubles in a row is a repeat offender.
    Jokers are misdemeanors. Super jokers are felonies.

    A player who wishes for a specific roll and gets it, would be
    guilty of premediteted luck.

    A player who wishes for but doesn't get a specific roll, would
    only be guilty of attempted luck.

    If a player like Murat, in defying the bot's temperature map,
    wishes for a roll that he considers lucky and gets it, he may
    be accused of voodoo luck but would "get away with luck".

    All this can be summarized as: Backgammon is by nature a
    game of luck with undefinable and unmeasurable amounts
    of luck and skill occurring in it.

    "Luck + skill = 1" is plain bullshit, pretentious false science.

    I congratulate you for having read my entire article. I won't
    try to dissuade you from framing and hanging it on your wall. ;)

    MK

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Axel Reichert@21:1/5 to murat@compuplus.net on Tue Aug 16 09:06:04 2022
    MK <murat@compuplus.net> writes:

    In other words, even a "clueless" player will win more than 0% against
    a "perfect" player.

    Sure.

    For years and years I suggested that we make a "random bot" play
    against a "perfect bot" to find out what that number is but nobody has
    shown any interest.

    I have done so about two years back in a different context. No need to
    program a single line, just have GNU Backgammon play itself, say,
    "Expert" against "Beginner". And since "Beginner" is still far better
    than "Random", set the "noise" ("Advanced Options") to 1, the maximum
    value. Have fun watching!

    A 21-point match just now during breakfast resulted in "Random" 2 versus "Expert" 21, a rather typical result, here are my older experiments:

    Casual Player-Random
    64:1
    64:1

    Advanced-Casual Player
    64:5
    64:10

    World Class-Advanced
    64:21
    64:55

    And since you are obsessed with denying the existence of cube skill, I
    have now done another one, "Expert with Random Cube" against "Expert".
    The random cuber won 21:12. But it seems (also for "random" checker
    play) that a noise value of 1 is not enough and does not result in
    completely random play, be it cube or checkers.

    So maybe I need to reactivate my mutant bot.

    Best regards

    Axel

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From MK@21:1/5 to Axel Reichert on Tue Aug 16 00:36:39 2022
    On August 16, 2022 at 1:06:07 AM UTC-6, Axel Reichert wrote:

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

    For years and years I suggested that we make a
    "random bot" play against a "perfect bot" to find
    out what that number is but nobody has shown
    any interest.

    I have done so about two years back in a different
    context. No need to program a single line, just have
    GNU Backgammon play itself, say, "Expert" against
    "Beginner". And since "Beginner" is still far better
    than "Random", set the "noise" ("Advanced Options")
    to 1, the maximum value. Have fun watching!

    Not quite the same but similar idea/experiment.
    I'm glad to hear you have done those but can you
    tell us what was your "context"/purpose in doing
    your experiments? I would like to know. It may be
    interesting to me.

    Advanced-Casual Player
    64:5
    64:10
    World Class-Advanced
    64:21
    64:55
    .....

    When you were doing those, did you pay attention
    to see if the results met the expextations, i.e. the
    expected win/lose ratio based on the error rates
    of the two sides?

    As you know, based on my own experiment results,
    one of my arguments is that the error rates are bs.

    And since you are obsessed with denying the
    existence of cube skill,

    When genericly speaking I refer to cube skill as one
    but I have clarified many times that I actually mean
    it's much less than what it's hyped up to be. More
    specificly, I only accept the existance of cube skill
    towards the very ends of games.

    But it seems (also for "random" checker play) that
    a noise value of 1 is not enough and does not result
    in completely random play, be it cube or checkers.

    So maybe I need to reactivate my mutant bot.

    Yay! I always thought that someday you would.

    BTW: I'm not trying to insult you but trying tu push
    your buttons to "activate you". ;)

    I think you are the person most open minded about
    running experiments here. I value that.

    MK

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Axel Reichert@21:1/5 to murat@compuplus.net on Sun Aug 28 14:40:59 2022
    MK <murat@compuplus.net> writes:

    On August 16, 2022 at 1:06:07 AM UTC-6, Axel Reichert wrote:

    But it seems (also for "random" checker play) that a noise value of 1
    is not enough and does not result in completely random play, be it
    cube or checkers.

    So maybe I need to reactivate my mutant bot.

    Yay! I always thought that someday you would.

    No need to, as I found some old "mutant" session results with expert
    checker play mimicking GNU Backgammon and random cube play (double,
    hold, beaver, take, pass). After 1000 games the mutant was trailing with
    17394 points versus gnubg's 114822. You can suspect from the points per
    game that this is Petersburg Paradox terrain, which I confirmed with my stability analysis techniques.

    It seamed too trivial to warrant a post.

    Best regards

    Axel

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From MK@21:1/5 to Axel Reichert on Sun Aug 28 21:38:08 2022
    On August 28, 2022 at 6:41:02 AM UTC-6, Axel Reichert wrote:

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

    On August 16, 2022 at 1:06:07 AM UTC-6, Axel Reichert wrote:

    So maybe I need to reactivate my mutant bot.

    Yay! I always thought that someday you would.

    No need to, as I found some old "mutant" session results

    I have to admit I'm rather surprised that you had already
    done these kinds of "mutant" experiments in the past.

    with expert checker play mimicking GNU Backgammon

    What's "mimicking" mean? Is your mutant bot a slightly
    modified version of Gnubg or a variant of a different bot?

    and random cube play (double, hold, beaver, take, pass).
    After 1000 games the mutant was trailing with 17394
    points versus gnubg's 114822.

    That's almost exactly 15%. I don't' know how to refer to it
    in plain language. I'm tempted to use words like "tare"(??)

    I think setting the checker skill to the highest possible of
    the bot would better isolate "cube skill" and 1,000 ganes
    are hardly enough to be significant but your numbers are
    all we have and thus surely better than nothing at all.

    I should also clarify that since I'm not trying to predict the
    "cube skill value of random cube decisions", I'm willing to
    accept whatever properly done experiments will reveal.

    Are anyone willing to comment on your 15% result and/or
    what would they have guesses/predicted it would be?

    Also one very important point that I keep repeating and
    you all keep avoiding is the predicted results based on the
    error rate in comparison to the actual results.

    In your above experiment, what was the cube error rate
    of the mutant that made random cube decisions?!?!

    Based on that error rate, was the mutnat expected to win
    more or less than 15%?!?!

    Knowing these is very important in proving my argument
    that equities, error rates, etc. as calculated by the current
    bots are "bullshit"! If you disagree, provide the numbers
    from your experiments. You can run but you can't hide...

    You can suspect from the points per game that this is
    Petersburg Paradox terrain, which I confirmed with my
    stability analysis techniques.

    I won't even question your "techniques" if you can try to
    explain what your "Petersburg Paradox terrain" mean in
    relation to your "cube skill theory"??

    MK

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From Axel Reichert@21:1/5 to murat@compuplus.net on Tue Aug 30 08:45:48 2022
    MK <murat@compuplus.net> writes:

    On August 28, 2022 at 6:41:02 AM UTC-6, Axel Reichert wrote:

    with expert checker play mimicking GNU Backgammon

    What's "mimicking" mean? Is your mutant bot a slightly
    modified version of Gnubg or a variant of a different bot?

    I programmed it to ask for GNU Backgammon's move and then send it
    back. So checker play will be identical.

    I think setting the checker skill to the highest possible of the bot
    would better isolate "cube skill" and 1,000 ganes are hardly enough to
    be significant

    I will not spend time on this. Perhaps some day on "never double, always
    take" and "never double, never take", two limiting cases similar to some beginners' cube "handling". But we do know the results already, don't
    we? These "strategies" will be beaten terribly, even with perfect
    checker play. Maybe I can save the experimental effort and derive
    something analytically from

    https://www.bkgm.com/rgb/rgb.cgi?view+513

    Also one very important point that I keep repeating and
    you all keep avoiding is the predicted results based on the
    error rate in comparison to the actual results.

    I will not spend time on this.

    I won't even question your "techniques" if you can try to
    explain what your "Petersburg Paradox terrain" mean in
    relation to your "cube skill theory"??

    We had this before. I will not spend time on this.

    Axel

    --- SoupGate-Win32 v1.05
    * Origin: fsxNet Usenet Gateway (21:1/5)
  • From MK@21:1/5 to Axel Reichert on Tue Aug 30 17:52:57 2022
    On August 30, 2022 at 12:45:51 AM UTC-6, Axel Reichert wrote:

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

    On August 28, 2022 at 6:41:02 AM UTC-6, Axel Reichert wrote:

    I think setting the checker skill to the highest possible
    of the bot would better isolate "cube skill" and 1,000
    ganes are hardly enough to be significant

    I will not spend time on this.

    I wasn't asking you either. The rest of my sentence that
    you cut off said "but your numbers are all we have and
    thus surely better than nothing at all." meaning that we
    can make do with what we have for the moment being.

    If you are going to ignore the results of your own tests
    and are not willing to discuss their implications, why do
    waste your time doing such experiment to begin with??

    Perhaps some day on "never double, always take" and
    "never double, never take", two limiting cases similar to
    some beginners' cube "handling". But we do know the
    results already, don't we? These "strategies" will be
    beaten terribly, even with perfect checker play.

    The "ramdom cube" experiment was about "no strategy"
    and you left is unfinished as with your other experiments.

    Of course, more experiments about various "strategies"
    can be revealing but you have demonstrated that you are
    incapable of doing an experiment properly. So, I would
    advice against wasting more of your time and then ours.

    Maybe I can save the experimental effort and derive
    something analytically from
    https://www.bkgm.com/rgb/rgb.cgi?view+513

    Since you say you already know the results, why would
    you even bother with more "smokes and maths"?

    Especially if you ignore and deny the "relative" results??

    You make fart-ass claims like "will be beaten terribly"
    with no effort to quantify "how terribly" and, of course,
    "in relation/proportion to what other other quantity".

    I can even accept that you skip this "prediction" step to
    be based on your "cube skill theory" (bullshit!) but can't
    let you off the hook without retroactively comparing the
    actual results to what would have been expected by your
    "cube skill theory bullshit"!

    Also one very important point that I keep repeating and
    you all keep avoiding is the predicted results based on
    the error rate in comparison to the actual results.

    I will not spend time on this.

    Wise decision, because if you did, then you would have to
    execute "anal harakiri" using your cube skill theory bullshit!

    I won't even question your "techniques" if you can try to
    explain what your "Petersburg Paradox terrain" mean in
    relation to your "cube skill theory"??

    We had this before. I will not spend time on this.

    You never did. That was another half-brained, half-baked
    experiment of yours. I drafted a digest of it and set it aside
    in favor of more current discussions but it's there for me to
    go back to it at an oppurtune moment.

    MK

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