On October 15, 2021 at 3:33:33 PM UTC-6, Philippe Michel wrote:
On 2021-10-14, MK <[email protected]> wrote:
And lastly, there is nothing "mysterious" about the "style" or "strategy" I >> used because I explained it openly in detail, such as my doubling right
after XG++'s opening with a 63 and its playing the way it does.
My reasoning is simple: the first version of TD-Gammon was an "alpha"
bot even if limited to playing 1-pointers. According to it, opening 63 is a >> bad roll.
This is an example of why you are mocked here.
You mean Mocky is mocking at me..? :)
You make extraordinary claims but support them with reasons ranging
from weak to patently wrong.
This is indeed precious for me, that you acknowledge that not all of my
claims are patently wrong but some are supported by at least a weak
reasoning. Would you be brave enough to specify which ones are which?
Here, you simply neglect the fact that, even if opening with 63 with the
cube centered were sligthly below average, opening with 63 *and holding
the cube* is very different, and a lot better.
No. Not in a rollout of 4 billion tries. And that what you and your ilk don't understand. The cube is not part of the backgammon game but it's just
a gambling device to manipulate the stakes. It can at best affect checker
play considering the n-away scores in a match play but not even that in
money play.
What are you gonna do with holding the cube after 63? Hatch chicks?
In 4 billion tries, some times you will hold it until the end of the game, sometimes you will redouble and play to the end of the game with your
opponent holding the cube, etc. etc.
4 billion is a pretty big number. It will cover all possibilities after the opening 63 sufficently enough to trivialize the role of "cube ownership",
etc. and thus take the cube out of the backgammon in a way, by forcing
the games to be played out to the end.
If one side is played by an alpha-bg bot, the cube will go sky high ( as in Chow's hypesygammon experiment already) and the result will be decided
by checker play skill...
To that, I added my argument that the player who first gains an
equity adge in a game will never lose statistically based on a significant >> number of trials like 4 billion games.
Unhelpful argument again. This implicitely assumes that the players
are equal.
Well, duh! Of course. Isn't that what's assumed in all bot rollouts and two-handed jackoffs?
In fact, just as "equal" but "indentical"! Two wrestlers "equal in weight" can't be compared if one does "free-style" and the other "greco-roman".
In every game the player who gets the opening roll "first gains
an equity edge" (on average, there are a few bad rolls,
Similar to white's getting the first move in chess.
although it is not clear 63 belongs to them).
I based my 63 experiments on TD-Gammon opening roll equities and
my humble logical argument.
Feel free to offer you own specific arguments as to why not?
If we play a variation where one of the players always rolls first
(but doublets are rerolled for the first move), he won't always win
very long sessions if he is not, at worst, only slightly weaker than
his opponent.
There is no way to theorize between unequal opponents, at least
not as of yet with the existing bots that are worshipped as being
better then any human. And so, between two "identical" opponents,
i.e. the bot masturbating doing rolllouts, with everything else being
equal, the player who gets the opening roll will win more than 50%.
If you can't agree to that, it must be because either you don't
understand statistics or that you are in denial of statisctics...
MK
--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)