While searching for old RGB threads relevant to
subjects that we have been discussing recently,
I found this: "Weird JF moves".
https://groups.google.com/g/rec.games.backgammon/c/_ZsA5s7Nowc/m/EeTLqfYxaS0J
It's a long, tedious one that I won't recommend
reading but I'll paraphrase some selected lines
from it to give you the gist of it.
I was accused of "jacking up the cube", even to
the extent of trying to "decide a 25-point match
in 1 game". Why wouldn't I do it if I thought that
it was to my advantage against the bots..? Yet,
I had also made, (and I kept making since), the
point that "it takes two to tango".
It's interesting to see that my recent comments
that gamblegammon matches of any length can
be over in a single game" have roots so far back.
I was arguing that the "cube could be trivialized,
"eliminated" from the game by keep taking and
doubling back", (i.e. turning it into a practically
cubeless game with more checker decisions, by
forcing it to be played out to the end), creating
an advantage for the stronger checker player.
I was already talking about using "my own criteria"
for cube decisions based on "stages of the game",
(i.e. "how much play is left in the game"), which I
have been using against bots ever since.
I worked out the details for such a mutant cube
skill experiment and hopefully I will find the time
to post it here eventually.
Some people were willing to extend some credit
to my ideas by making statements such as: "Given
extreme differences in skill between two players,
however, jacking up the cube may be justified" but
I never accepted them because I didn't see myself
as weaker than the bots.
David desJardins, (who later became a math PHD),
had said: "It seems to me that Murat has a single
valid point: Jellyfish doesn't realize that it's a much
stronger player than Murat is. So it doesn't take skill
differences into account when doubling. If Murat
does, then he can achieve a higher win percentage
(especially in long matches) than his relative strength
would otherwise predict".
He must be given the due credit for having said this
back then, (even though I don't agree that Jellyfish
was "much stronger" than me). In my experiments
playing *long* matches and money sessions against
Snowie, Gnubg and XG, I have consistently achieved
much higher win percentages than predicted by the
bots but I believe that the discrepancy is due to the
bots being wrong in estimating my strength based
on inaccurate, fancyful ER/PR calculations.
What's really interesting is what David desJardins
said in this post which I recommend you to read:
https://groups.google.com/g/rec.games.backgammon/c/_ZsA5s7Nowc/m/-wf-LR8MCvAJ
I'll comment on some of his selective paragraphs:
"Your opponent doesn't have to have any weaknesses
"in order for you to improve your results against it by
"adopting a "theoretically suboptimal" strategy.
Music to my ears, from a wax cylinder record... :)
"Taking doubles that a player just as strong as Jellyfish
"would drop, is a good example. Since his personal
"match equity table against JF is different than the
"theoretical match equity table between equal players,
"his correct double and take points will be different.
Did you guys catch the words "personal match equity
table"? This concept is what led me to post this thread.
What a novel idea!
Unfortunately, that was the first and only usage of the
expression. It was only last year that I started talking
about running experiments with "mutant match equity
tables" and suggested that bots offer the functionality.
I thought about it independently of what he had said
almost 25 years ago. What a huge loss that nobody
else developed on the idea during all those years... :(
I wonder if he had realised himself the potential value
of his own novel/unique idea. I think this is something
that needs to be further explored.
He also said:
"I think it's possible to make the practical observation
"that taking more and doubling more than theoretical
"calculations would predict will improve your match
"winning chances, and then to implement that
"observation, without having to have a detailed
"understanding of the theory.
I think this is as close a gamblegammon math PHD
candidate could come to saying that the so-called
"doubling theory" is mostly bullshit. ;) For this to be
"possible", the match winning chances predicted by
theoretical calculations, (and thus what is wrongly
called a "theory"), must be erroneous!
What is more likely is that some people, (like me?),
may be able to make this observation instead and
"knowingly" take advantage of their opponents who
implement the inaccurate "cube skill theory".
If everyone in a congregation implements a same
ridiculous "cube theory", like "if it's Monday and it's
raining, double early", "if it's full moon and the third
week of the month, drop late", etc. it will eventually
self-validate because everyone will mistakenly end
up concluding that they are winning or losing based
on how well they implement that so-called theory.
I was already arguing this back then using "doubling
after every 10 moves" as an example in my post.
Any wonder why in the year 2023, gamblegammon
mathematicians still can't calculate probabilities of
match winnings based on rating differences, etc... :(
MK
PS: A bit of information that may be useful later is
that the ten 25-point matches I had played against
Jellyfish had lasted 9 games on the average.
--- SoupGate-Win32 v1.05
* Origin: fsxNet Usenet Gateway (21:1/5)