In 2010, I had started a short thread titled:

"Skill within strategy - ideas for better bg bots"

https://groups.google.com/g/rec.games.backgammon/c/X3t-7sPFbaw/m/nMcBfUPx-WEJ

Do yourselves a favor and read (or reread even if
you had read in the past but forgotten) at least my
initial post, her reply to me and my response to her.

It contains paragraphs like this from me:

"I believe that bots will eventually become smart
"enough to use dynamic strategy against their
"opponents, even if it may not happen soon.

Even if not fully interlocking with mine, Patti Beadles'
ideas are very similar and contributive to the subject
of future AI bots.

Although she can't completely break free of the notion
of "optimal play" and doesn't think that bots can reliably
exploit human weaknesses because it will be impossible
to get large enough data sets on particular individuals,
she talks about "exploitive play", "categorizing positions",
"identifying patterns of games", etc.

I found her web site (http://pattib.org/) but it was last
updated in 2013. I wonder if she still operates FIBS and
is interested in BG. It would be interesting to hear what
she would have to say on the same subject today..?

MK

On 11/8/2022 5:58 PM, MK wrote:
> It contains paragraphs like this from me:
>
> "I believe that bots will eventually become smart
> "enough to use dynamic strategy against their
> "opponents, even if it may not happen soon.
>
> Even if not fully interlocking with mine, Patti Beadles'
> ideas are very similar and contributive to the subject
> of future AI bots.

I've probably mentioned this before here on r.g.b., but already
in 1999 there had been held a computer roshambo (rock-paper-scissors)
tournament that had demonstrated the concept of exploiting the
opponent's weaknesses.

https://web.archive.org/web/20120114173409/http://webdocs.cs.ualberta.ca/~darse/rsbpc.html

Of course, it was impossible to gain an advantage in head-to-head play
against the "optimal" strategy of purely random choices, but the top
programs scored massively better than the "optimal" strategy did in the
round-robin tournament.

---
Tim Chow

On November 10, 2022 at 5:56:35 AM UTC-7, Tim Chow wrote:

> On 11/8/2022 5:58 PM, MK wrote:

>> It contains paragraphs like this from me:

>> "I believe that bots will eventually become smart
>> "enough to use dynamic strategy against their
>> "opponents, even if it may not happen soon.

> I've probably mentioned this before here on r.g.b.,
> but already in 1999 there had been held a computer
> roshambo (rock-paper-scissors) tournament that
> had demonstrated the concept of exploiting the
> opponent's weaknesses.

I vaguely remember you mentioning it briefly in a
similar context, as well as other things like coin
toss football, chickens crossing the road, both gg
players with 10 pieces each on their 1-point, etc.
in relation to skills and strategies in bg/gg.

> Of course, it was impossible to gain an advantage
> in head-to-head play against the "optimal" strategy
> of purely random choices, but the top programs
> scored massively better than the "optimal" strategy
> did in the round-robin tournament.

As with your other examples that were either irrelevant
to bg/gg, or failed to prove "cube skill", or maintained
the argument for "optimum strategy", I don't understand
what's your point again this time?

My argument is that there is no such thing as a single
"optimum strategy", that currently humans are capable
of flexible/alternative strategies but bots aren't but that
bots will also become capable of this in the future.

You may call a "consistent player's" (i.e. a bot's) strategy
"optimum", (regardless of its skill level), only while it's
playing against itself.

When it plays against another bot or human opponent,
that opponent's strategy can be just as "optimum" and
either side can exploit the other, whether consistently
one-sidedly or both sides continuously taking turns at
exploiting each other.

Once a bot (or human) deviates from the strategy that
it considers "optimum" in order to exploit the opponent,
then you can't talk about "optimum strategy" anymore.

You can't have your optimum and eat it too! (© Murat K:)

If you think you understand my argument, can you make
a clear statement on whether you agree or disagree with
me? In other words, are you still arguing that there is and
always will be an "optimal strategy" in bg/gg such that no
other deviating strategy will be able to gain an advantage
against it?

MK

On 11/11/2022 6:45 AM, MK wrote:
> If you think you understand my argument, can you make
> a clear statement on whether you agree or disagree with
> me? In other words, are you still arguing that there is and
> always will be an "optimal strategy" in bg/gg such that no
> other deviating strategy will be able to gain an advantage
> against it?

As long as there is a fixed limit on how high the cube can get,
yes, no deviating strategy will be able to gain a consistent
advantage over the "optimal strategy" *in the long run*. The
long run could be very long if the cap on the cube value is high.

In the short run, even in pure games of chance such as roulette,
it may be possible (for example) to obtain a >50% chance of coming
out ahead (at the risk of losing big if you do lose).

---
Tim Chow