I see that all kinds of discussions have been going
on in Bgonline and Facebook BG group, about the
fate of XG and the future of AI gamblegammon bots.

Let's start by first going back in time for a minute.

Until 2001, the last Jellyfish was v3.2 and users were
asking if there would ever be a v4.0 until this post:

https://groups.google.com/g/rec.games.backgammon/c/Wa5yFNkgMqI/m/TIlXI5pyaPgJ

After that and until 2008, the last Snowie was v4.71
and users were asking if there would ever be a v5.0
until this post:

https://groups.google.com/g/rec.games.backgammon/c/NE1HhK27yAA/m/udzwG8exJt0J

Then until 2014, the last XG was v2.1 and users were
wondering if there would ever be a v3.0 while rumors
were circulating such as in this thread about Wachtel's
book (save your time and skip reading the long thread):

https://groups.google.com/g/rec.games.backgammon/c/0oGf7_Dmhpk/m/OjEMeJPWryoJ

In one post, Tim Chow mentions that the book included
rollouts done using XG v3.0 beta. But by 2015, there was
only a minor update to v2.19 and with that, in 2016 I had
predicted that there would never be an XG v3.0 and in
2017 I had reiterated my prediction that Xavier would go
down the way of Fredrik Dahl and Olivier Egger. See:

https://groups.google.com/g/rec.games.backgammon/c/nYr_oLBYH0g/m/8jgOFsvUBgAJ

In 2018, I was mockingly soliciting bets on XG v3.0 in:

https://groups.google.com/g/rec.games.backgammon/c/K9vYdtgUI9M/m/Qz5yTxZwAwAJ

With the comment that "Cows have been milked and
milking bulls isn't lucrative??" ;)

Lo and behold, another 5 years later, it's said that there
won't be an XG v3.0 and that Xavier is fancying selling
it for $1,000,000++ (pun intended;)

People offer all kinds of comments, arguments, advice,
etc. about what to do now/next, with most of the posts
showing what a pathetic group of mentally ill gambling
addicts they are. :(

There nothing I can do to help them and even though I
don't equate the gamblegammon variant with the real
backgammon that I love, I still would like to see better
AlphaZero type BG bots developed.

I can't create an incentive for a better bot but I think I
can create a "need" for it by demonstrating that the
existing bots are no good.

Creating and training AlphaZero type BG bots may not
be easy but demolishing Ex-Gee/Noo-BG/BG-Bzzt/etc.
by doing "AlphaZero rollouts" will only take a couple of
hours to add to the existing bots and then rollouts will
only take a couple of hours each to do.

What I'm talking about is what I had previously called
"clinical rollouts", i.e. millions of trials with purely
random checker and purely random cube decisions.

Since an AlphaZero BG bot would be trained the same
way and since you all are using AlphaZero in referring
to them, I thought I may get my idea through your thick
heads using the same word to call it "AlphaZero rollout".

Once the current bots are debunked to be beyond being
worthy of improvement, then the only remaining choice
will be to develop new and better bots from scratch...

MK

inhale the "good stuff"

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

I assume <doing "AlphaZero rollouts" will only take a couple of hours >
refers how fast Alphazero trained a chess AI better than Stockfish? Then do yourself a favour and check which hardware they used. IIRC in the post from Tim was a link to ELF and there were some numbers or in a talk of Aja Huang (or similar) from 2016.
IIRC a fast PC with let's say 16 cores needs more than a year, so that is not that quick (and the structure of BG is not well suited to GPUs) Unless you have someone with very deep pockets you'll stick with month.

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

On July 24, 2023 at 9:50:29 AM UTC-6, Frank Berger wrote:

I assume <doing "AlphaZero rollouts" will only
take a couple of hours refers how fast Alphazero
trained a chess AI better than Stockfish?

No. You assumed wrong. Didn't you read my posts
about "clinical rollouts", i.e. through random checker
and random cube decisions?

I had done some rollouts as close as possible to a
random, i.e. AlphaZero, rollout using Noo-BG's "noise"
settings, which to its credit is the only bot that offers
such a feature, (as opposed to bots like your piece of
shit BG-Bzzt), that had taken only about two hours on
my almost obselete by now Intel i7-8700 CPU to do a
1 million 296 thousand trials! See:

https://groups.google.com/g/rec.games.backgammon/c/F5xoWKrOe8I/m/50xP3DhhAQAJ

Then do yourself a favour and check which hardware
they used. ..... IIRC a fast PC with let's say 16 cores
needs more than a year, so that is not that quick .....

I knew very well that it would take a ton of dynamite to
blow the wall in one shot and that's why I proposed my
"AlphaZero rollouts" approach to take down the wall
one brick at a time using a sledge hammer...

Do you understand? I don't think you can. If I were the
developer of a piece of shit gamblegammon bot like
your BG-Bzzt, I would defend that it's godd enough and
that no significantly better AI bots can be developed. I
do understand you and feel your pain... :(

(and the structure of BG is not well suited to GPUs)

I'll just ignore this remark for now in order to not stray
from our main subject on hand.

Unless you have someone with very deep pockets
you'll stick with month.

I don't. But IBM does. As far as I know Tseauro is still
working for IBM. I have tried and will try my part to put
Tseauro/IBM to shame for the damage they have done
to backgammon/gamblegammon AI bot development
by sending all their idiot followers on the wrong path,
starting with injecting human bias into TD-Gammon v2.

IBM has the machine and human power to clean up
Tseauro's shit and redeem itself but public pressure
must be put on it. Unfortunately none of you mentally
ill gambling scums of humanity will help me with that.

So, my "AlphaZero rollouts" is a compromise for now.

I have posted here how it could be added to Noo-BG
by adding/modifying only 5 lines but the coq mangeurs
won't implement it for the obvious reasons.

If your piece of shit BG-Bzzt bot isn't programmed in a
worse "spaghetti code" than Noo-BG, you should be
able to add the AlphaZero, i.e random, rollouts feature
to it just as easily and I'm daring you to do it! I bet you
won't!!! For the reason you know better than all those
mentally ill gamblers that you try to sell your piece of
shit bot for $19.95...! ;)

MK

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

I'm very confident that IBM will be very open to your suggestions, especially if your wording is as usually.

And yes, it would be dead easy to add that to BGBlitz. It's just that I spend time for a target group of a single user only if I regard it as interesting.
And even the charming and throughout positve way you communicate wont change that.

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

On July 26, 2023 at 3:40:57 PM UTC-6, Frank Berger wrote:

I'm very confident that IBM will be very open to your
suggestions, especially if your wording is as usually.

I don't think IBM would be open doing anything based
on suggestions, no matter how worded and by whom.

Do you remember "Deep Blue"? Why do you think IBM
spent all that money for it, including the custom CPU?
For the love of science and chess? Guess again.

Not so much later, a team of maverics came along to
piss on IBM's Deep Blue, more for the glory and fame
of it than for financial profits.

Another team of mavericks could similarly piss on IBM
and Tseauro in backgammon or in gamblegammon but
there is no competition for backgammon AI bots and
gamblegammon is so much infested and monopolized
by a handful mentally ill scum of humanity that it is no
surprize that people like AlphaZero team don't want to
get involved in that.

I certainly don't blame them. I wouldn't shake hands of
those scumbags, wearing 4-ply gloves (pun intended:).

And yes, it would be dead easy to add that to BGBlitz.
It's just that I spend time for a target group of a single
user only if I regard it as interesting.

This can be a perfect "David and Goliath" story but you
are no David material. You are just a regular pathetic
sick gambler with programming skills trying to peddle
a piece of shit so-called AI bot for $19.95... :(

You can't be interested in anything beyond that for any
reason whatsoever! Your limits are very apparent.

And even the charming and throughout positve way
you communicate wont change that.

Consider this: some people otherwise very polite and
nice in general towards most people, eventully start to
enjoy swearing mouthfuls to people like you because
thay realize that it wouldn't/doesn't make a difference!

I have never accomplished anything better by being
nice to you or your coq mangeurs ilks of you. So why
not enjoy myself by pissing on you scums instead...? ;)

MK

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