I don't know much about them but apparently there are quite a number of collections/databases of games/matches that are kept for various reasons
like rating/ranking players, analysing positions, etc.

What I'm wondering is whether there are any stats about the frequency of backgammon positions?

I know the the number of all possible legal positions is very large but the number of positions that actually occur during real playing is probably not very large at all. I think this would be a very interesting and useful thing to know for all kinds of practical applications.

I would propose that the most frequent positions in human-v-human,
human-v-bot and bot-v-bot games will be distinctively different due to the characteristic styles/strategies of different types of players, and perhaps
be also different based on the strength level of the players in each of the above pairing combinations.

Most stored games are probably already saved as sequences of unique
position ID's or can be batch converted to be so. And then a script can
read through all available games and tabulate the frequency of all the
unique position ID's encountered.

Any thoughts on this subject?

Any ideas about how difficult would this be to accomplish?

Could the effort be justified for some beneficial uses like "frequent
positions books" to speed up bots or just to satisfy human curiosity?

Personally, it would be worth for me to contribute some of my time and
effort towards this if I can be useful in some way.

MK

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

MK <murat@compuplus.net> writes:

I know the the number of all possible legal positions is very large
but the number of positions that actually occur during real playing is probably not very large at all. I think this would be a very
interesting and useful thing to know for all kinds of practical
applications.

I agree that this is an interesting question, but have some doubts about
the usefulness. Some thoughts:

Imaging a vast area of land. A backgammon position is, well, a position somewhere in this area. A backgammon move is a path from one position to another one. Some of these paths are well-trodden (kind of a six-lane autobahn), e.g., the one from the starting position to the one after 31:
8/5 6/5, others are rarely used and have become overgrown, e.g., the one
from the starting position to the one after 31: 8/7 8/5. Other paths
from the starting position might have an intermediate state, such as the
3 most common moves with 43 ("Down", "Split", "Zplit"), whereas the "mediterranean" "Up" has become a smaller one over the decades. An
interesting detail is that this "map of backgammon country" is
mathematically a directed multi-graph (if I got the terminology right)
with all its consequences regarding application of graph theory. It
could be also seen a Markov chain, with probabilities assigned to status changes, e.g., after 31 in the starting position, we will end up almost certainly with the move 8/5 6/5, whereas for 43 we might have something
like 35 % for Down, 35 % for Split, 30 % for Zplit and 5 % for Up, see
(the perhaps outdated) https://www.bkgm.com/openings.html.

Continuing with the opening stage, we will have a "combinatorial
explosion" because of the much higher branching factor of backgammon
compared to chess (which is one of the reasons why it took roughly 10
years longer for the computers to become better than world class
players).

Coming to the end of the game, in a pure race there will a "squeezing"
zone and both the number of positions and the number of paths will
shrink. Especially in bear-off I expect a large discrepancy between
positions theoretically possible versus positions encountered during
live play. After I published my Isight count, I got a comment from
someone stating roughly "looks like a bad method, since it failed my
basic test". When I asked about the "basic test", it turned out that he
had used this:

GNU Backgammon Position ID: Pp8PAACfzwcAAA
Match ID : cAkAAAAAAAAA
+24-23-22-21-20-19------18-17-16-15-14-13-+ O: gnubg
| O O O | | | 0 points
| O O O | | |
| O O O | | |
| O O O | | |
| O O O | | |
| |BAR| |v (Cube: 1)
| X X X | | |
| X X X | | |
| X X X | | |
| X X X | | | On roll
| X X X | | | 0 points
+-1--2--3--4--5--6-------7--8--9-10-11-12-+ X: axel
Pip counts: O 60, X 45

This is "Double, take", but my Isight count said "Double, pass", which
sealed the deal for him. I then asked whether he had ever seen this
position over the board ("No") and asked whether it makes more sense to
devise a method such that it works well for positions encountered in
real life ("I never thought along these lines, but maybe yes"). This is
why I used Tom Keith's database of endgame positions gathered from real
matches played on FIBS.

I would propose that the most frequent positions in human-v-human, human-v-bot and bot-v-bot games will be distinctively different due to
the characteristic styles/strategies of different types of players,
and perhaps be also different based on the strength level of the
players in each of the above pairing combinations.

Yes, sure: Splitters versus slotters, backgamers versus blitzers,
novices running when behind, not daring to hit openly but preferring
stacking plays, ...

Most stored games are probably already saved as sequences of unique
position ID's or can be batch converted to be so. And then a script
can read through all available games and tabulate the frequency of all
the unique position ID's encountered.

Yes, sure, technically this should be simple. But what would you do with
this? The combinatorical explosion is really, really bad, so I am quite
sure that since backgammon started being played we (which is all human
players ever in existence) have so far covered only a fraction of
"practically relevant" positions, let alone "theoretically possible"
positions.

Could the effort be justified for some beneficial uses like "frequent positions books" to speed up bots or just to satisfy human curiosity?

My guess is that such a look-up table would have to be so huge that it
would not save time anymore. Also, despite being huge, it would be too
small to reasonably cover "backgammon country". Which is why humans use
rules of thumb ("if in doubt, hit"). Even bots do not work by memorizing positions (again, in contrast to chess programs, at least in the opening stages), but by assigning weights to "features" of positions. This
amounts to a huge compression of information. In fact, it is quite
embarrassing that a neural network with a ridiculously low number of
"brain cells" beats almost all human players in the long run.

So to summarize: I do not think that such a brute-force approach to
backgammon will work. But it might serve well in extracting information
for a "compressed" heuristics (which is what both humans and bots do and
did).

Best regards

Axel

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

On 10/04/2022 1:21 pm, MK wrote:

I don't know much about them but apparently there are quite a number of collections/databases of games/matches that are kept for various reasons
like rating/ranking players, analysing positions, etc.

Indeed.

What I'm wondering is whether there are any stats about the frequency of backgammon positions?

Probably not, or you'd have found them surely?

I know the the number of all possible legal positions is very large but the number of positions that actually occur during real playing is probably not very large at all. I think this would be a very interesting and useful thing to
know for all kinds of practical applications.

Not sure why, but ok.

I would propose that the most frequent positions in human-v-human, human-v-bot and bot-v-bot games will be distinctively different due to the characteristic styles/strategies of different types of players, and perhaps be also different based on the strength level of the players in each of the above pairing combinations.

Most stored games are probably already saved as sequences of unique
position ID's or can be batch converted to be so.

They are more likely saved as match logs - plain text files of moves
and/or .xg files.

And then a script can
read through all available games and tabulate the frequency of all the
unique position ID's encountered.

Yes, if it knew how to build position ids from move lists.

Any thoughts on this subject?

A few.

Any ideas about how difficult would this be to accomplish?

Technically easy, even I could do it. Practically, it depends on the accessibility of the data and the format it's saved in.

Could the effort be justified for some beneficial uses like "frequent positions books" to speed up bots or just to satisfy human curiosity?

For the latter, yes, if one were willing to devote the time and effort required.

Personally, it would be worth for me to contribute some of my time and
effort towards this if I can be useful in some way.

As you're a self-confessed programmer, you can do it all yourself. All
the tools you need are publicly available.

Let us know how you go.

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

MK schrieb am Sonntag, 10. April 2022 um 05:21:24 UTC+2:

Most stored games are probably already saved as sequences of unique
position ID's or can be batch converted to be so. And then a script can
read through all available games and tabulate the frequency of all the unique position ID's encountered.

Any thoughts on this subject?

Any ideas about how difficult would this be to accomplish?

Wasn't MK the one who named bgblitz rubbish because I said I fixed about 50 bugs and claimed that he is quite well in the programming area even providing a link to his website?
And now you ask this question for an pretty trivial task? (Why I'm not really astonished)

There are about 10^20 possible positions and if only 1% of the them occur in real games and you need about 20 bytes to describe a position you can't do that counting even with the current largest supercomputer.
So if you only use recorded matches (I have a colection with about 4 million human matches, let's say if all BG servers collaborate we might get 100 million matches) it might be doable with a decent PC (probably within a 4 digit budget) But counting the
individual positions is pointless unless you categoryze them.

BTW the effort is moderate it took me about 1 or 2 evenings to code that stuff, and I used it for categorizing.
Using that for a lookup is not a relevant idea as Axel already pointed out.

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

On April 10, 2022 at 3:36:33 AM UTC-6, Axel Reichert wrote:

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

I agree that this is an interesting question, but have
some doubts about the usefulness. Some thoughts:

Well, it's already great that you and a couple others are
taking interest in the subject. Let's just delve into it with
open minds and see what we will duscover. Often times
some use is found for ideas that initially appear totally
useless.

Imaging a vast area of land. A backgammon position is,
well, a position somewhere in this area. A backgammon
move is a path from one position to another one. Some
of these paths are well-trodden ..... others are rarely used
.....
Coming to the end of the game, in a pure race there will a
"squeezing" zone and both the number of positions and
the number of paths will shrink.

This is a good start but I will try to paint it a little differently
and then tie it to some of my previous ideas.

I won't call "a backgammon move a path" but "an element
of scenery", along "a path of dice rolls sequence" going
from "a starting location to an end destination" within your
"a vast area of backgammon-land".

Let's say we are going from Seattle to New York. At the
start of our travel, we will see the same familiar elements
of scenery again and again, like the Space Needle. Towards
the end of our travel, we will see the same familiar elements
of scenery again and again also, like the Statue of Liberty.

This is similar to your above ideas about beginning and end
positions but using different analogies.

Continuing with the opening stage, we will have a "combinatorial
explosion" because of the much higher branching factor of
backgammon compared to chess ....

Yes, and what I want to explore is kind of what you are calling
"branching factor", not between different games but between
different types of backgammon players like bots and humans
of varying skills and styles.

GNU Backgammon Position ID: Pp8PAACfzwcAAA
Match ID : cAkAAAAAAAAA
This is "Double, take", but my Isight count said "Double, pass",
which sealed the deal for him. I then asked whether he had
ever seen this position over the board ("No").....

This in my analogy would be an element of scenery, a very odd
shaped cactus somewhere in the desert that nobody will never
travel by and never see.

When I say "backgammon" I mean the classic, cubeless variety
and thus I was really only talking about "positions" without any
regard to match scores, how to play a given rolls, etc. let alone
cube ownership, cube decisions, etc.

But by mentioning the cube, you made me realize that there is
value in discussing how "shortened cubeful travels" will effect
the frequency of positions themselves. I will get into this later
on in this post.

BTW: I insist that referring to cubeful-backgammon simply as
backgammnon is wrong. It should be given a different name
like any other backgammon variant which it clearly is. I call it "gamble-gammon" and hope that it or another suggested name
will find acceptance and common usage.

I would propose that the most frequent positions in
human-v-human, human-v-bot and bot-v-bot games will
be distinctively different due to the characteristic
styles/strategies of different types of players, and perhaps
be also different based on the strength level of the
players in each of the above pairing combinations.

Yes, sure: Splitters versus slotters, backgamers versus
blitzers, novices running when behind, not daring to hit
openly but preferring stacking plays, ...

Yes, it's great that you acknowledged and even elaborated some.
So, let's keep going.

Many times in the past, I likened bots to "trains on tracks" and
called their single-strategy (i.e. "best play", "perfect play" or to
make Tim happy "optimum play", etc.) style of play words like
"sterile", "metallic", "robotic" :), etc. and complained that playing
against bots eventually becomes rather boring.

I also likened other strategies and styles of play to riding cars, atv's/horses. So, let me develop by combining those analogies
with my new one here about going from Seattle to New York.

If you take the train, it will follow a single track all the way, each
and every time. You will always see the same limited scenery
that you will eventually memorize and maybe even get bored of.

This is bot-vs-bot play. I predict that this style/strategy will lead
to the lowest number of possible positions. Especially so in
gamble-gammon because of dropped cubes, which can happen
after just a couple of rolls and cut the trip short before even
getting out of city limits.

A strong human player would have the capability of steering his
car but would stick to taking the major freeways, often running
along railways in parallel. Through occasional "PR-sacrificing
moves", he will get to see some more possible positions.

At weaker and weaker human (or bot) players may end up taking
some back-roads, dirt-roads or go off-road on their atv's/horses,
weaving and meandering around, thus encountering increasingly
larger and larger numbers of possible positions, maybe even get
totally lost at times to even see the odd shaped cactus that you
gave in your example position above.

Just to clarify, I'm not including here odd human players like me
who may claim that they can beat the train on horsback, even
though purposefully making "PR-sacrificing moves" and better
yet "PR-defying moves" will result in a larger number of possible
positions as well but I think that should be a different discussion.

An interesting detail is that this "map of backgammon country"
is mathematically a directed multi-graph (if I got the terminology
right) with all its consequences regarding application of graph
theory. It could be also seen a Markov chain, with probabilities
assigned to status changes, e.g., after 31 in the starting position,
we will end up almost certainly with the move 8/5 6/5, whereas
for 43 we might have something like 35 % for Down, 35 % for .....

I'm not sure if I really understand what you are talking about but
I would like to see you elaborate and explain further.

The various stats I'm talking about can, of course, be presented
as graphs. Trying to understand what you wrote, I thought of
color spectrums of elements. I wonder if the graphs of bot-v-bot, human-v-human, human-v-bot positions graphs can be similarly
recognizable enough characteristic? What about even subsets
like strong-bot-v-weak-human, weak-human-v-weak-human,
weak-bot-v-strong-human, etc...?? Can we look at a "spectrum
of positions" and be able to say that those come from games
played by weak-bots and strong-humans?

Also, back to the destructive effect of the cube: here I'm talking
about positions independent of what dice rolls the may come
after, etc. (as in your above mentioning of what will come after
31 in the starting position).

But cube decisions effect positions relevantly to the subject
here, because after a double/drop, the game gets truncated
and comes to an unnatural end as far as real backgammon
is concerned and not gamble-gammon. When that happens,
positions that could come up after each specific double/drop
position do no longer come up. The spectrum of positions in
gamble-gammon may be visibly different than in cubelessly
played "real" backgammon.

In fact, with this said, while tabulating frequency of positions,
each time a certain position is the final position of a game,
it should be marked and counted as such also.

Yes, sure, technically this should be simple. But what would
you do with this? The combinatorical explosion is really, really
bad, so I am quite sure that since backgammon started being
played we (which is all human players ever in existence) have
so far covered only a fraction of "practically relevant" positions,
let alone "theoretically possible" positions.

Am I using the "possible positions" wrongly? I'm only referring
to that "fraction" of positions that has have already come up
during the recorded games. I just want to count their frequencies
and tabulate/rank/plot them.

Among my reasons is just to look at them to see if something
alien will jump at my face... :)

Could the effort be justified for some beneficial uses like
"frequent positions books" to speed up bots or just to satisfy
human curiosity?

My guess is that such a look-up table would have to be so
huge that it would not save time anymore.

We can control the size to a practical number of most frequent
positions.

Also, despite being huge, it would be too small to reasonably
cover "backgammon country". Which is why humans use
rules of thumb ("if in doubt, hit"). Even bots do not work by
memorizing positions (again, in contrast to chess programs,
at least in the opening stages), but by assigning weights to
"features" of positions.

Is this like you guys givin creative/cute names to positions
when discussing them here? Also, I thought that at least some
bots were using opening books and even large positions files.

This amounts to a huge compression of information. In fact,
it is quite embarrassing that a neural network with a ridiculously
low number of "brain cells" beats almost all human players in
the long run.

How about using frequency stats during the training of neural
networks?

Since no entity with the necessary means is showing an interest
in working towards Alpha-Zero-BG-Bots, I keep trying to think of
things that can be subtituted towards better bots in the meantime.

Especially now that we know that the existing bots are not even
that good nor on the right track to becoming considerably better,
living happy enough with them is not possible. We must dump all
the old bullshit and try to create bots that are "differently better".

MK

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

On April 10, 2022 at 3:42:03 AM UTC-6, Simon Woodhead wrote:

On 10/04/2022 1:21 pm, MK wrote:

What I'm wondering is whether there are any stats
about the frequency of backgammon positions?

Probably not, or you'd have found them surely?

I looked. Not found. Never even heard any mention of
anything along those lines. I guess it takes an unusal
mind capable of free thinking to come up with unusual
questions.

I think this would be a very interesting and useful
thing to know for all kinds of practical applications.

Not sure why, but ok.

If there is ongoing interest in the subject, I have a few
ideas that I may offer for conversation.

Any ideas about how difficult would this be to accomplish?

Technically easy, even I could do it.

Your name sounded familiar and I checked. Sure enough
you are the creator of www.bglog.org/rgbnews.html which
I use every so rarely. So, this should be your specialty.

Could the effort be justified for some beneficial uses
like "frequent positions books" to speed up bots or just
to satisfy human curiosity?

For the latter, yes, if one were willing to devote the time
and effort required.

You said the thechnical part was easy. Could you provide
the scripts? Then we can see can do the legwork of data
gathering. After that, I would bet more than a few people
would be interested to analyse and interpret the stats.

Personally, it would be worth for me to contribute some
of my time and effort towards this if I can be useful in
some way.

As you're a self-confessed programmer, you can do it all
yourself. All the tools you need are publicly available.

It may be a little too much for one person to do it quickly
enough. I haven't looked to see what tools are available.
I have a few other major projects of deep interest on my
list that I only have time and energy to peck at as I can.

As far as backgammon, I am also pecking at the design
of a "Lego-Bg-Bot" and any amount of time and effort I
could spare for backgammon would first go towards that.

Even if I get no help, I may be able to hack it from Gnubg
but it may take me too long to start on it and then too long
to complete it.

Let us know how you go.

Don't worry, I will keep it a secret... ;)

MK

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

On April 10, 2022 at 8:57:50 AM UTC-6, bgbl...@googlemail.com wrote:

MK schrieb am Sonntag, 10. April 2022 um 05:21:24 UTC+2:

Any ideas about how difficult would this be to accomplish?

Wasn't MK the one who named bgblitz rubbish because I
said I fixed about 50 bugs and claimed that he is quite well
in the programming area even providing a link to his website?

So? BTW, I called your bot garbage for other more important
reasons than the 50 bugs you fixed. If you want to prove me
"right" for a change, try to give it away for free. Let's see how
many people will switch to using your shitty bot instead of XG
or Gnubg... :)

And now you ask this question for an pretty trivial task?

Okay, I have been humiliated for having to ask to learn that it
is a trivial task. It's very encouraging and promising to know.

(Why I'm not really astonished)

I don't know that either. :( Please tell.

There are about 10^20 possible positions and if only 1% of
the them occur in real games

How do you know this? Are you that sure it is not .5% or 2%?

and you need about 20 bytes to describe a position

Actually, strictly speaking about positions, it only takes 10
bytes but this is not a major issue to debate. For what I am
talking about, there is no need to know the match ID which
is the other 10 bytes.

you can't do that counting even with the current largest
supercomputer.

Can you give specific numbers instead of a generic claim?

So if you only use recorded matches (I have a colection with
about 4 million human matches, let's say if all BG servers
collaborate we might get 100 million matches)

These numbers are beyond what my wildest guesses would be.
Wow! Great. That means we can make somewhat statistically
meaningful observations.

it might be doable with a decent PC (probably within a 4
digit budget) But counting the individual positions is
pointless unless you categoryze them.

Surely we can find a few thousand dollars worth of volunteer
work, no? Encouraging still.

I would only mark the final moves in order to investigate the
doubling cube's destructive effect on backgammon, by its
unnaturally ending games and depriving the players of some
possibly very exciting, enjoyable positions to play if the game
wasn't abruptly truncated for the thrill of gambling.

But hey, I'm glad that you could see some use in categoryzing
positions in perhaps many other ways. While at it, who would
keep you from doing that too.

I faintly remember some comments by bott-kissing PR-nuts
that positions should be ranked based on their complexities
and be used in skill and luck calculations accordingly. There
is an opportunity to improve your shitty bot... ;)

BTW the effort is moderate it took me about 1 or 2 evenings
to code that stuff, and I used it for categorizing.

Oh, you already did it. I'm impressed. BTW, I hope I never said
anything bad about your programming abilities and assumed
that you could do better than your existing bot.

That's why I gave you some suggestions but obviously my
opinion/advice wasn't valuable for you. I have no problem
with that. If you can only be a follower and not an innovator,
there is no value nor fairness in making you wrong for that.

Using that for a lookup is not a relevant idea as Axel already
pointed out.

That's not what he said at all. He said it would require too much
resources for the usefulness it could provide. And that's just one
man's opinion which I acknowledge and value as such. But then
he also said that it could be useful for other puprposes. Go back
and read it again...

MK

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

MK <murat@compuplus.net> writes:

On April 10, 2022 at 3:36:33 AM UTC-6, Axel Reichert wrote:

dropped cubes, which can happen after just a couple of rolls and cut
the trip short before even getting out of city limits.

Yes.

If you take the train, it will follow a single track all the way, each
and every time. You will always see the same limited scenery that you
will eventually memorize and maybe even get bored of.

A resounding "No", see below.

players may end up taking some back-roads, dirt-roads or go off-road
on their atv's/horses, weaving and meandering around, thus
encountering increasingly larger and larger numbers of possible
positions, maybe even get totally lost at times to even see the odd
shaped cactus that you gave in your example position above.

Yes. Being an adventure cyclist, I know what you mean.

An interesting detail is that this "map of backgammon country" is
mathematically a directed multi-graph (if I got the terminology
right) with all its consequences regarding application of graph
theory. It could be also seen a Markov chain, with probabilities
assigned to status changes, e.g., after 31 in the starting position,
we will end up almost certainly with the move 8/5 6/5, whereas for 43
we might have something like 35 % for Down, 35 % for .....

I'm not sure if I really understand what you are talking about but I
would like to see you elaborate and explain further.

You understood "graph" as something like a contour plot or this:

https://en.wikipedia.org/wiki/Graph_of_a_function#/media/File:Three-dimensional_graph.png

I meant this:

https://en.wikipedia.org/wiki/Graph_(discrete_mathematics)

The various stats I'm talking about can, of course, be presented as
graphs. Trying to understand what you wrote, I thought of color
spectrums of elements. I wonder if the graphs of bot-v-bot,
human-v-human, human-v-bot positions graphs can be similarly
recognizable enough characteristic?

Sounds like

https://www.strava.com/heatmap

Different activities (running versus cycling) yield different results.

In fact, with this said, while tabulating frequency of positions, each
time a certain position is the final position of a game, it should be
marked and counted as such also.

Like I said, easy, but for which purpose?

Among my reasons is just to look at them to see if something alien
will jump at my face... :)

Without classification, you will simply be drowned in data.

My guess is that such a look-up table would have to be so
huge that it would not save time anymore.

We can control the size to a practical number of most frequent
positions.

No. The branching factor is so high, that even in "boring" bot-vs-bot
play you will rarely encounter exactly the same position after you have
left the city limits and before you have entered New York. Try it
yourself, you have played some longer sessions against GNU
Backgammon. Extract the Match and Position ID, sort it and check for
multiple occurrences.

I did so just now, with one file containing 1096762 different
positions. Of these, 770042 occured exactly once and 321517 occured
exactly twice. The fraction of positions occuring more than twice was
thus less than 5 permille, and the highest number of occurences at all
was 6. That much to your statement from above: "If you take the train,
it will follow a single track all the way, each and every time." It is
simply not true.

I thought that at least some bots were using opening books and even
large positions files.

GNU Backgammon has a bear-off database. So this covers Manhattan. I do
not think it covers Seattle.

How about using frequency stats during the training of neural
networks?

No. The branching factor is too high. The outcome will be fed back
immediately, AFAIK.

Since no entity with the necessary means is showing an interest
in working towards Alpha-Zero-BG-Bots, I keep trying to think of
things that can be subtituted towards better bots in the meantime.

A table lookup-up will not do, for the reasons given above.

Best regards

Axel

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

MK <murat@compuplus.net> writes:

On April 10, 2022 at 3:36:33 AM UTC-6, Axel Reichert wrote:

dropped cubes, which can happen after just a couple of rolls and cut
the trip short before even getting out of city limits.

Yes.

If you take the train, it will follow a single track all the way, each
and every time. You will always see the same limited scenery that you
will eventually memorize and maybe even get bored of.

A resounding "No", see below.

players may end up taking some back-roads, dirt-roads or go off-road
on their atv's/horses, weaving and meandering around, thus
encountering increasingly larger and larger numbers of possible
positions, maybe even get totally lost at times to even see the odd
shaped cactus that you gave in your example position above.

Yes. Being an adventure cyclist, I know what you mean.

An interesting detail is that this "map of backgammon country" is
mathematically a directed multi-graph (if I got the terminology
right) with all its consequences regarding application of graph
theory. It could be also seen a Markov chain, with probabilities
assigned to status changes, e.g., after 31 in the starting position,
we will end up almost certainly with the move 8/5 6/5, whereas for 43
we might have something like 35 % for Down, 35 % for .....

I'm not sure if I really understand what you are talking about but I
would like to see you elaborate and explain further.

You understood "graph" as something like a contour plot or this:

https://en.wikipedia.org/wiki/Graph_of_a_function#/media/File:Three-dimensional_graph.png

I meant this:

https://en.wikipedia.org/wiki/Graph_(discrete_mathematics)

The various stats I'm talking about can, of course, be presented as
graphs. Trying to understand what you wrote, I thought of color
spectrums of elements. I wonder if the graphs of bot-v-bot,
human-v-human, human-v-bot positions graphs can be similarly
recognizable enough characteristic?

Sounds like

https://www.strava.com/heatmap

Different activities (running versus cycling) yield different results.

In fact, with this said, while tabulating frequency of positions, each
time a certain position is the final position of a game, it should be
marked and counted as such also.

Like I said, easy, but for which purpose?

Among my reasons is just to look at them to see if something alien
will jump at my face... :)

Without classification, you will simply be drowned in data.

My guess is that such a look-up table would have to be so
huge that it would not save time anymore.

We can control the size to a practical number of most frequent
positions.

No. The branching factor is so high, that even in "boring" bot-vs-bot
play you will rarely encounter exactly the same position after you have
left the city limits and before you have entered New York. Try it
yourself, you have played some longer sessions against GNU
Backgammon. Extract the Match and Position ID, sort it and check for
multiple occurrences.

I did so just now, with one file containing 10000 games with 491623
positions. Of these, 485584 occured exactly once, which is almost 99 %
of all positions. One position came up 10000 times (guess which), and 17 positions came up some hundred times (the ones after the 15 different
opening ROLLS, played somewhat differently by my program and GNU
Backgammon, hence 17 POSITIONS after the play, not 15).

That much to your statement from above: "If you take the train, it will
follow a single track all the way, each and every time." It is simply
not true and backgammon is a more scenic ride than you thought. Which is
why we love this game. (-:

I thought that at least some bots were using opening books and even
large positions files.

GNU Backgammon has a bear-off database. So this covers Manhattan. I do
not think it covers Seattle.

How about using frequency stats during the training of neural
networks?

No. The branching factor is too high. The outcome will be fed back
immediately, AFAIK.

Since no entity with the necessary means is showing an interest
in working towards Alpha-Zero-BG-Bots, I keep trying to think of
things that can be subtituted towards better bots in the meantime.

A table lookup-up will not do, for the reasons given above.

Best regards

Axel

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

On 12/04/2022 5:08 pm, MK wrote:

You said the thechnical part was easy. Could you provide
the scripts? Then we can see can do the legwork of data
gathering. After that, I would bet more than a few people
would be interested to analyse and interpret the stats.

As Frank pointed out, developing this part of what you want to
do is easy enough, a few hours work. To generate ids from move
lists is a little extra. No I cannot provide the scripts, the
parts you want are embedded deep in bglog and would be meaningless
outside of that context. Besides, if you were to do this work,
you'd gain credibility in a community that you spend half your
time swearing at.

It may be a little too much for one person to do it quickly
enough. I haven't looked to see what tools are available.
I have a few other major projects of deep interest on my
list that I only have time and energy to peck at as I can.

Me too, probably Frank also. So why would we spend time doing your
bidding when we have more interesting projects to do? And again,
as Frank pointed out this is not a hard thing to do so if you are
serious about this, show us by developing the code to generate id
lists from move lists. That would be a good start and even with a
learning curve should only take you a week or two.

As far as backgammon, I am also pecking at the design
of a "Lego-Bg-Bot" and any amount of time and effort I
could spare for backgammon would first go towards that.

Even if I get no help, I may be able to hack it from Gnubg
but it may take me too long to start on it and then too long
to complete it.

No-one helped me write bglog. I'm sure no-one helped Frank write
bgblitz. These are labours of love, and tens of thousands of hours
of work over literally decades of development. Anyone can start a
project. Finishing it to the point where it's useful to others is
another matter. You have yet to perform, so go and make something
great for the backgammon community. You'll be showered with gifts
and admiration. Ok, you won't, but you'll get the self-satisfaction
of having achieved something useful. Assuming it's useful ;-)

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

Simon Woodhead <simon@bglog.org> writes:

As Frank pointed out, developing this part of what you want to
do is easy enough, a few hours work.

Easy? It is trivial! The statistics I have given in a previous reply
were generated by 1 line of code. After reading "Classic Shell
Scripting" and "Unix Power Tools" any willing learner should be ready to
go. The problem is probably the "willing" ...

No I cannot provide the scripts, the parts you want are embedded deep
in bglog and would be meaningless outside of that context.

Same here. Code is living language, not cryo-conserved quotes.

Besides, if you were to do this work, you'd gain credibility in a
community that you spend half your time swearing at.

This pretty much sums it up.

Me too, probably Frank also. So why would we spend time doing your
bidding when we have more interesting projects to do?

Me too. Although I have to admit that the non-swearing half of
contributions (deliberately not using quotation marks here) is sometimes original and stimulating. At least I was thus enticed to learn and
program ("hark, hark") something with Markov chains, which proved useful
for my investigations about the beaver-caused "Petersburg paradox".

But the "value proposition" of this current project is very poor indeed,
at least so far.

Best regards

Axel

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

On 13/04/2022 5:05 pm, Axel Reichert wrote:

Simon Woodhead <simon@bglog.org> writes:

As Frank pointed out, developing this part of what you want to
do is easy enough, a few hours work.

Easy? It is trivial! The statistics I have given in a previous reply
were generated by 1 line of code. After reading "Classic Shell
Scripting" and "Unix Power Tools" any willing learner should be ready to
go. The problem is probably the "willing" ...

If you have long lists of just ids, sure. It's a bit more effort to
generate ids from move lists. But yeah, "willing" is the key.

"Code is living language, not cryo-conserved quotes."

Love this :-)

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

On 4/10/2022 5:36 AM, Axel Reichert wrote:

Continuing with the opening stage, we will have a "combinatorial
explosion" because of the much higher branching factor of backgammon
compared to chess (which is one of the reasons why it took roughly 10
years longer for the computers to become better than world class
players).

Let me see if I can translate the core of MK's question into more
commonly understandable language.

Create two large piles of backgammon positions. Pile 1 consists of
positions taken from bot-versus-bot games. Pile 2 consists of
positions taken from human-versus-human games.

Can we train a classifier to distinguish between a randomly chosen
position from Pile 1 and a randomly chosen position from Pile 2?
Of course it won't be perfect because some positions will appear in
both piles. But how high an accuracy/recall is possible?

---
Tim Chow

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

On April 12, 2022 at 1:53:06 AM UTC-6, Axel Reichert wrote:

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

On April 10, 2022 at 3:36:33 AM UTC-6, Axel Reichert wrote:

dropped cubes, which can happen after just a couple of rolls
and cut the trip short before even getting out of city limits.

Yes.

It's good that you acknowledge this but will you also accept
that it has consequences?

If we take all the recorded gamble-gammon games we have
and look at the frequence of very early positions, we will see
that the positions that could come up immediately following
the dropped cube will not come up as frequently as in regular
backgammon games.

From that, by looking at a "spectrum" of positions from a data
set, we can tell if those came from gamble-gammon games
or backgammon games.

What can be the practical use of that? I don't know yet but we
may find some as the discussion progresses. This is similar to
staring at posters (like the one in this link) with hidden images
until you get cross eyed and see them.

https://i.pinimg.com/originals/fe/43/d8/fe43d88059ec9a06c4377410739e7004.jpg

I can do this on my own but sharing thought here is like looking
at clouds together to see shapes. When I say I think I'm seeing
a cat up there, I don't do it asking to be negated but hoping that
someone may say it looks more like shaggy little dog than a cat.

If you take the train, it will follow a single track all the way,
each and every time. You will always see the same limited
scenery that you will eventually memorize and maybe even
get bored of.

A resounding "No", see below.

I will illustrate and explain that is indeed a resounding "Yes", but
in a separate post. Addressing specific issues separately while
still in the same thread may be easier and more productive.

My guess is that such a look-up table would have to be
so huge that it would not save time anymore.

We can control the size to a practical number of most
frequent positions.

No. The branching factor is so high, that even in "boring"
bot-vs-bot play you will rarely encounter exactly the same
position after you have left the city limits.....

I'm not talking about "play" but "positions". We will be looking
at positions in already played games. Whatever the braching
factor was in those games, it has already happened and it's
nearly not as high as theoretically possible.

I a given set of games, most positions will occur very rarely
and perhaps mostly only once. But some beginning and end
positions will occur frequently and that may/will be enough
to see some patterns, distinctive markers.

I did so just now, with one file containing 1096762 different
positions. Of these, 770042 occured exactly once and....

You're on the wrong track here. I'll explain in the other post.

I thought that at least some bots were using opening books
and even large positions files.

GNU Backgammon has a bear-off database. So this covers
Manhattan. I do not think it covers Seattle.

If it makes sense to have bots with doable opening and closing
(bear-off) books, why not a doable mid-game book? We don't
need to map every tree and telephone pole all the way. We can
create a city size map of only major elemets of scenery, like
recognizable landmarks, and stretch it across the states.

How about using frequency stats during the training of neural
networks?

No. The branching factor is too high. The outcome will be fed
back immediately, AFAIK.

Branching factor has nothing to do with what I'm proposing. I
defend this idea much more avidly than the look-up table idea
as I will explain it in much detail in another post.

MK

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

On Thursday, April 14, 2022 at 12:10:22 AM UTC+1, peps...@gmail.com wrote:

On Wednesday, April 13, 2022 at 1:36:45 PM UTC+1, Tim Chow wrote:

On 4/10/2022 5:36 AM, Axel Reichert wrote:

Continuing with the opening stage, we will have a "combinatorial explosion" because of the much higher branching factor of backgammon compared to chess (which is one of the reasons why it took roughly 10 years longer for the computers to become better than world class players).

Let me see if I can translate the core of MK's question into more
commonly understandable language.

Create two large piles of backgammon positions. Pile 1 consists of positions taken from bot-versus-bot games. Pile 2 consists of
positions taken from human-versus-human games.

Can we train a classifier to distinguish between a randomly chosen
position from Pile 1 and a randomly chosen position from Pile 2?
Of course it won't be perfect because some positions will appear in
both piles. But how high an accuracy/recall is possible?

If we take the entirety of human backgammon games, no matter how weak the players, then the positions are completely different.
Probably over 99% of humans are weak beginners or worse.
I think I could distinguish between a random human position and a random strong bot vs strong bot position with over 90% confidence.
For example, beginners are terrified of getting hit so you get massive candlesticks
around the 8 and 6 points. A 5 point slotted is a strong suggestion of a non-beginner
and therefore a non-human.
A better question is whether we can readily distinguish between positions encountered
when strong bots play each other, and positions encountered when world-class humans
play each other. Almost certainly, no, we can't.

Paul

More fundamentally, beginners (over 95% of humans) are completely ignorant of priming
and closing out. So if all humans are in the sample then, if one of the players has a strong inner board
or a prime, then it's a bot position.

Paul

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

On Wednesday, April 13, 2022 at 1:36:45 PM UTC+1, Tim Chow wrote:

On 4/10/2022 5:36 AM, Axel Reichert wrote:

Continuing with the opening stage, we will have a "combinatorial
explosion" because of the much higher branching factor of backgammon compared to chess (which is one of the reasons why it took roughly 10
years longer for the computers to become better than world class
players).

Let me see if I can translate the core of MK's question into more
commonly understandable language.

Create two large piles of backgammon positions. Pile 1 consists of
positions taken from bot-versus-bot games. Pile 2 consists of
positions taken from human-versus-human games.

Can we train a classifier to distinguish between a randomly chosen
position from Pile 1 and a randomly chosen position from Pile 2?
Of course it won't be perfect because some positions will appear in
both piles. But how high an accuracy/recall is possible?

If we take the entirety of human backgammon games, no matter how weak the players, then the positions are completely different.
Probably over 99% of humans are weak beginners or worse.
I think I could distinguish between a random human position and a random
strong bot vs strong bot position with over 90% confidence.
For example, beginners are terrified of getting hit so you get massive candlesticks
around the 8 and 6 points. A 5 point slotted is a strong suggestion of a non-beginner
and therefore a non-human.
A better question is whether we can readily distinguish between positions encountered
when strong bots play each other, and positions encountered when world-class humans
play each other. Almost certainly, no, we can't.

Paul

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

On April 12, 2022 at 1:53:06 AM UTC-6, Axel Reichert wrote:

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

An interesting detail is that this "map of backgammon
country" is mathematically a directed multi-graph (if I
got the terminology right) with all its consequences
regarding application of graph theory. It could be also
seen a Markov chain, with probabilities assigned to
status changes, e.g., after 31 in the starting position,
we will end up almost certainly with the move 8/5 6/5,
whereas for 43 we might have something like 35 % for
Down, 35 % for Split, 30 % for Zplit and 5 % for Up, see
(the perhaps outdated) https://www.bkgm.com/openings.html.

I'm not sure if I really understand what you are talking about
but I would like to see you elaborate and explain further.

You understood "graph" as something like a contour plot or this:

https://en.wikipedia.org/wiki/Graph_of_a_function#/media/File:Three-dimensional_graph.png

Yes, I knew you were speaking a different language but your
use of it led me to make use of it in its common meaning.

The graph in your link is much beyond what would be enough
for my purpose.

The various stats I'm talking about can, of course, be presented
as graphs. Trying to understand what you wrote, I thought of
color spectrums of elements. I wonder if the graphs of bot-v-bot,
human-v-human, human-v-bot positions graphs can be similarly
recognizable enough characteristic?

Visualizing helps understand thing better and I was trying use
the idea of graphs to help you guys understand me better but
I was thinking more about graphs like these:

https://www.astro.rug.nl/~ndouglas/teaching/IMAGES/achilles.elements.gif

https://smallpond.ca/jim/misc/cfl/spectra.jpg

Maybe "chart" would be a better word here. Replace element
names in these with various combinations bot, human, strong,
weak, etc. or light bulb names with names of various bots.

Can we substitute frequencies of backgammon positions for
wavelengths of colors and create spectrum charts of positions
taken from certain sets of games, and then looking at them
observe characteristics of certain types of gamble-gammon,
backgammon, players, styles, strayegies, etc...??

Am I flyings too high above you guys? Look up! It's a bird, it's a
plane, it's Murat...!

I meant this:
https://en.wikipedia.org/wiki/Graph_(discrete_mathematics)

Your heatmap of different activities (running versus cycling)
doesn't apply at all here but let's go back to your other ideas
and let me requote for clarity:

An interesting detail is that this "map of backgammon
country" is mathematically a directed multi-graph (if I
got the terminology right) with all its consequences
regarding application of graph theory.

I did read the wiki and my understanding is that it would be
a "mixed multigraph". Trying to speak your lingo, if positions
are "vertices" and dice rolls are "edges", dice rolls played in
one or more possible ways would be "directed edges", each
to a different position. If it ended here, it would be a "directed multi-graph". But if we look at those resulting positions and
the same dice rolls, if it's the opposite side's (i.e. "white's")
turn to play, one of the (necessarily legal) plays would be the
exact reverse of the first side's (i.e. "black's) play. Thus those
"edges" would also be "undirected edges" and thus the graph
would be a "mixed multigraph" not a "directed multi-graph".
(Warning! I'm not a math PHD. :)

And now it's my turn to ask you what is the practical use (if
any) of this? Can you expand on what you mean by "with all
its consequences regarding application of graph theory"??

seen a Markov chain, with probabilities assigned to
status changes, e.g., after 31 in the starting position,
we will end up almost certainly with the move 8/5 6/5,
whereas for 43 we might have something like 35 % for
Down, 35 % for Split, 30 % for Zplit and 5 % for Up, see
(the perhaps outdated) https://www.bkgm.com/openings.html.

Looking at that page, no two opening moves have the same
equity.

Humans may chose to make an inferior, "PR-sacrificing" move
but bots will always make the move they conside the best.

In your example of 43, the bot will always play 13/10 13/9.
In bot play you will never see "35 % for Down, 35 % for Split,
30 % for Zplit and 5 % for Up" regardless of which bot and
regardless of what strength level of that bot.

They may occur in human play but in measuring, consistency
is essential and we can only use bots for measuring beacuse
they play 100% consistently, regardless of which bot and
regardless of what strength level of that bot.

So, it's again my turn again to ask you what's the practical use
(if any) of this nonsense of yours?

Your trying to bring in the "Markov chain" again fails badly.
Are you trying to impress people by bulshitting big with high
math stuff?

And, BTW, what was the result of your experiment with "Murat
mutant" against Gnubg? Is the hury still out? Will you ever
share your finddings with us, hopefully in a common language
that we all can understand?

MK

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

On 4/13/2022 7:10 PM, peps...@gmail.com wrote:

I think I could distinguish between a random human position and a random strong bot vs strong bot position with over 90% confidence.
For example, beginners are terrified of getting hit so you get massive candlesticks
around the 8 and 6 points. A 5 point slotted is a strong suggestion of a non-beginner
and therefore a non-human.
A better question is whether we can readily distinguish between positions encountered
when strong bots play each other, and positions encountered when world-class humans
play each other. Almost certainly, no, we can't.

Even if what you say is true, there is still a rather large zone
of players between beginners and top humans. I'd guess that the
available recorded games contain a fairly large sample of games from
this "middle zone." True beginners generally don't have their games
recorded, and only a handful of players are at the very top.

---
Tim Chow

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

On April 12, 2022 at 12:18:03 PM UTC-6, Axel Reichert wrote:

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

opening ROLLS, played somewhat differently by my program and
GNU Backgammon, hence 17 POSITIONS after the play, not 15).

That much to your statement from above: "If you take the train, it
will follow a single track all the way, each and every time." It is
simply not true and backgammon is a more scenic ride than you
thought. Which is why we love this game. (-:

I'll use this post to try explain in detail at least one possible use
analysing frequency of positions.

I've never said that backgammon-land wasn't scenic. I said that gamble-gammon-land was less scenic. I tried to say that either
one is more scenic on horseback or on a bike than on train.

Anologies are difficult. You may have misunderstood my train
analogy. By train ride, I didn't mean you or your bot (perhaps
the Murat mutant bot?) or I playing against a bot. Right after
my sentence that you quoted above, I said "This is bot-vs-bot
play".

In fact, initially I thought of it as strongest-bot-vs-strongest-bot
but later I expanded it to include any boy of any strength level
as long as it plays against itself (which is necessary to achieve
100% consistent play).

Now, let's start by using an example from your first post.

Some of these paths are well-trodden (kind of a six-lane
autobahn), e.g., the one from the starting position to the
one after 31: 8/5 6/5, others are rarely used and have
become overgrown, e.g., the one from the starting position
to the one after 31: 8/7 8/5.

If we look at positions from games played by, let's say, Gnubg
4-Ply against itself, we will never see the position resulting
from playin an opening 31 as 8/7 8/5 because it will alwas
play it as 8/5 6/5.

The same is true for all other 8 preset levels of Gnubg, from
Grandmaster to Beginner.

This means not only that we wont see 31: 8/7 8/5 when those
levels play against themselves but also that we won't see it
even when 4-Ply plays against Beginner. Never. Not even once.

Thus, the absence of 31: 8/7 8/5 is a uniquely distinctive marker
of Gnubg-vs-Gnubg games.

Now if we look at positions from a set of randomly picked
games and see 31: 8/7 8/5 even once, we can conclude that
these positions didn't come from Gnubg-vs-Gnubg games.

This in itself may not be useful but since we know that even
Gnubg Beginner leven dosen't play 31: 8/7 8/5, we conclude
also that 31: 8/7 8/5 can only be played by a bot or human
weaker than Gnubg Beginner. (Except, of course, while they
were playing against Gnubg, if Mockey or Mickey had a few
too many sakes. :)

Having a marker like this can be very useful. Let me explain.

We know that something the current bots lack is to be able
to adjust their games to the skill level of their opponents, like
some human players can claimedly do, in order to exploit a
weaker (or stronger?) opponent. Ways of improving bots to
do this have been often discussed in the past.

One suggestion I had made was to train bots making them
play not just against themselves ("masturbation") but also,
and more so, against opponents ("intercourse") of different
skill levels.

Assuming that this (or similar) idea has merit, after being
so trained, bots would still need to assess their opponents
during actual play in order to adjust their games accordingly.

One way of achieving this may be for the bot to constantly
monitor the opponents error rate keep updating a calculated
assessment of its opponent.

Another way may be for the bot to look for unique markers
such as the example above, which may be more efficient.

Markers not as strong as the above example can be used in
various combinations to make them stronger and more precise.

I realize that all can do and all I'm doing here is sowing a seed.
It may grow a tree, a bush or a bonsai. Who kows...?

MK

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

On 4/13/2022 9:28 PM, MK wrote:

Another way may be for the bot to look for unique markers
such as the example above, which may be more efficient.

Markers not as strong as the above example can be used in
various combinations to make them stronger and more precise.

What you're calling a "marker" is usually called a "feature"
in the machine learning community.

The frequency with which a position occurs might be called a
"static" feature, whereas tracking behavior over the course of
a game might be called a "dynamic" feature.

In the old days, having human programmers propose specific
features was more or less the only way to get off the ground
with machine learning. Nowadays, though, the increase in
computing power means that it's often better to just let the
programs discover their own features rather than hand-craft
features using human expertise. Static features are simpler
but dynamic features tend to be more powerful, and again
because of the increase in computing power, dynamic features
tend to be the way to go if they are available.

---
Tim Chow

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

On Thursday, April 14, 2022 at 2:02:48 AM UTC+1, Tim Chow wrote:

On 4/13/2022 7:10 PM, peps...@gmail.com wrote:

I think I could distinguish between a random human position and a random strong bot vs strong bot position with over 90% confidence.
For example, beginners are terrified of getting hit so you get massive candlesticks
around the 8 and 6 points. A 5 point slotted is a strong suggestion of a non-beginner
and therefore a non-human.
A better question is whether we can readily distinguish between positions encountered
when strong bots play each other, and positions encountered when world-class humans
play each other. Almost certainly, no, we can't.

Even if what you say is true, there is still a rather large zone
of players between beginners and top humans. I'd guess that the
available recorded games contain a fairly large sample of games from
this "middle zone." True beginners generally don't have their games
recorded, and only a handful of players are at the very top.

My first guess is that distinguishing positions from the best bots and from competent humans
is extremely difficult. I'd expect to be able to do this (or teach a computer to do this) with somewhere
around 55 to 60% accuracy.
The key difference is that, in contact positions, bots are far more likely to create a mega-blotty home board
in order to make points later or preserve timing. The bots can see that the trillions of blots are not a problem
because no sequence exists to take advantage of these blots. We are unable to get ourselves hit to attack them
and we are unable to take advantage of these blots by loose plays that wouldn't work otherwise.
Since many thousands of sequences need to be assessed to make this assessment, humans can't generally
do these mega-blotty plays successfully.

Paul

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

MK <murat@compuplus.net> writes:

If it makes sense to have bots with doable opening and closing
(bear-off) books, why not a doable mid-game book?

Because in contrast to bear-off positions and especially opening
positions you will almost never come across exactly the same
"landmark". Hence the cartographer's work would be futile. No one else
will visit the same spot in forseeable time. This is caused by the
branching factor of the dice, even if an algorithm plays the same
position always in the same way.

Best regards

Axel

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

MK <murat@compuplus.net> writes:

On April 12, 2022 at 1:53:06 AM UTC-6, Axel Reichert wrote:

https://en.wikipedia.org/wiki/Graph_(discrete_mathematics)

[...]

And now it's my turn to ask you what is the practical use (if any) of
this? Can you expand on what you mean by "with all its consequences
regarding application of graph theory"??

I thought we were still in brainstorming mode? Also, it is YOUR burden
to tell what you are after. I am just helping out with ideas.

In your example of 43, the bot will always play 13/10 13/9. In bot
play you will never see "35 % for Down, 35 % for Split, 30 % for Zplit
and 5 % for Up" regardless of which bot and regardless of what
strength level of that bot.

They may occur in human play but in measuring, consistency is
essential and we can only use bots for measuring beacuse they play
100% consistently, regardless of which bot and regardless of what
strength level of that bot.

So, it's again my turn again to ask you what's the practical use (if
any) of this nonsense of yours?

As before, I am just helping out with ideas. No need to flip-flop back
into insulting mode. Some paragraphs before, you were hypothesising
about distinguishing bots from human, now humans are out. And since you
mention "measuring", but do not say what you want to measure, I am out
as well. Good luck with your project.

[Further insults deleted]

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

On 15/04/2022 12:40 pm, MK wrote:

[blah blah blah]

If you don't want to be around me while I'm trying so,
you can stuff your gifts and admirations up your asses
and go fuck yourselves away from me... ;)

Yeah, no worries Murat, you get what you deserve. You could have had
help here but, as Avon once said "If you didn't want the answer, why
did you ask the question?"

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

On April 13, 2022 at 1:06:05 AM UTC-6, Axel Reichert wrote:

Simon Woodhead <si...@bglog.org> writes:

As Frank pointed out, developing this part of what
you want to do is easy enough, a few hours work.

Easy? It is trivial! The statistics I have given in a
previous reply were generated by 1 line of code.
After reading "Classic Shell Scripting" and "Unix
Power Tools" any willing learner should be ready
to go. The problem is probably the "willing" ...

Sad to see you joing the club of dickheads. :(

Although I have to admit that the non-swearing half
of contributions (deliberately not using quotation
marks here) is sometimes original and stimulating.
At least I was thus enticed to learn and program
("hark, hark") something with Markov chains, which
proved useful for my investigations about the
beaver-caused "Petersburg paradox".

In the computer world, programming is the easiest
thing to do. That's why people advance from that to
system analysis and design, to testing, to innovating.

I'm glad someone expressedly benefited from my
contribution here, in this case was my proposing
some easy ways that people can use to prove to
themselves that the "so-called cube theroy" is just
a elaborate, fanciful pile of bullshit.

Unfortunately, even after his experiments proved
it to him, Axel is unable overcome his denial, or
order to accept it and contribute back by openly
sharing his findings. :(

But the "value proposition" of this current project is
very poor indeed, at least so far.

Yes, not only "so far" is the keyword here but also
everyone may not get the same mileage from the
same tank of gas... Especially not if you take off
making bullshit assertions by flooring the gas pedal
and coming to a backflip stop... ;)

MK

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

On April 12, 2022 at 4:20:33 PM UTC-6, Simon Woodhead wrote:

On 12/04/2022 5:08 pm, MK wrote:

As Frank pointed out, developing this part of what
you want to do is easy enough, a few hours work.
To generate ids from move lists is a little extra.

Apparently Frank has already done what I propose
partially, for a different purpose. So for him it may
take not even a few hours but only a few minutes
to adapt his scripts.

You guys are stingy of donating a few hours to a
project for the sake of the game but you don't mind
wasting as many hours at being dickheads to me. :(

No I cannot provide the scripts, the parts you
want are embedded deep in bglog and would
be meaningless outside of that context.

Fine, just say so and leave it at that.

Besides, if you were to do this work, you'd gain
credibility in a community that you spend half
your time swearing at.

If I were concerned about credibility, respectability,
likability, etc. in a dog pack community of mentally
ill gamblers who created for themselves a closed,
comfortable little universe in denial of realities, I
would't swear at you all in the first place, would I?

Me too, probably Frank also. So why would we
spend time doing your bidding when we have
more interesting projects to do?

I couldn't have known this without asking probing
questions to the entire RGB. If you don't want to
contribute that's fine. Simply not responding at
all would have been good enough for me and save
you your precious time. But you guys had to be a
dickheads, didn't you?

And again, as Frank pointed out this is not a hard
thing to do so if you are serious about this, show
us by developing the code to generate id lists...

I didn't ask for anything from you, Frank or anybody
else directly. Neither can you ask anything from me.

Even if I get no help, I may be able to hack it from
Gnubg but it may take me too long to start on it
and then too long to complete it.

No-one helped me write bglog. I'm sure no-one
helped Frank write bgblitz. These are labours of
love, and tens of thousands of hours of work over
literally decades of development.

Which one of you devised the position encoding?
Or maybe one of you even invented the wheel??

You couldn't have produced what you have without
the helps of "the labours of love and tens of thousands
of hours of work over literally decades of development"
of other people before you!

Knowledge is accumulative and eventually belongs to
all, as we all somehow benefit from the innovations
and efforts of other. So, it's okay for you two to have
done the same and hopefully at least morally/legally,
as opposed to some petty crook scumbags who steal
in broad daylight, make money from it and still menage
to get more respect from this community that I could
never get. I guess I lack what it takes to belong among
you; fortunately... :)

You have yet to perform, so go and make something
great for the backgammon community.

As a lone ranger outsider, I have contributed greatly
and uniquely to this community, for over 25 years,
constantly having to suppress my disgust for the sake
of the game that I love (and surely love differently and
for different reasons than the majority here).

You'll be showered with gifts and admiration. Ok,
you won't, but you'll get the self-satisfaction of
having achieved something useful. Assuming it's
useful ;-)

I do my thing and I am self-content with my existance
here regardless of how much I achieve as perceived
by you all. It's not easy to crack some people's thick
skulls to open their minds but I'll keep trying.

If you don't want to be around me while I'm trying so,
you can stuff your gifts and admirations up your asses
and go fuck yourselves away from me... ;)

MK

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

On April 13, 2022 at 6:36:45 AM UTC-6, Tim Chow wrote:

On 4/10/2022 5:36 AM, Axel Reichert wrote:

Continuing with the opening stage, we will have a
"combinatorial explosion" ....

Let me see if I can translate the core of MK's question
into more commonly understandable language.

I thought I was the common man speaking the
commonly understandable language here. Do
you think it may help better if you translate to a
laguage that mathematicians can understand?

Create two large piles of backgammon positions.
Pile 1 consists of positions taken from bot-versus-bot
games. Pile 2 consists of positions taken from
human-versus-human games.

Can we train a classifier to distinguish between a
randomly chosen position from Pile 1 and a randomly
chosen position from Pile 2? Of course it won't be
perfect because some positions will appear in both
piles. But how high an accuracy/recall is possible?

That's not what I tried to say. In fact, I believe that
distinguishing between randomly chosen positions
would be not just "imperfect" but "impossible".

What I proposed was to create a chart from all the
positions in your Pile 1 and another chart from all
the positions in your Pile 2. Then take a chart made
from all the positions in a ramdom pile and compare
it to previous two charts to see if it looks like one of
them.

Depending on whether it looks "somewhat", "mostly"
or "exactly" like one of them, we may be able to say
that it must consists of positions taken from bot-
versus-bot games or from a human-versus-human
games with a proportionate confidence.

Thanks for trying to pitch in. I would like to see you
and others delve deeper into the discussion.

MK

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

On April 13, 2022 at 8:37:39 PM UTC-6, Tim Chow wrote:

On 4/13/2022 9:28 PM, MK wrote:

In the old days, having human programmers propose
specific features was more or less the only way to get
off the ground with machine learning.

You must be, in your way, referring to 3rd paragrah of
4th chapter at this link, for example:

https://www.cs.cornell.edu/boom/2001sp/Tsinteris/gammon.htm

I quote for your convenience:

"Tesauro also released subsequent versions of TD-Gammon
"that built upon the initial success. Instead of just using a
"raw" board encoding (number of white and black pieces at
"each location) as input, a set of handcrafted features deemed
"important by experts was also added. This included pre-
"computed information such as the probability of being hit,
"or the strength of a blockade.

We have discussed this many times, in different contexts,
and I did/do object to it as the injected human bias in all
current bots that are all offsprings of TD-G v2. which has
become much worse when bots started to be trained to
play matches, cubeful money games, etc....

Nowadays, though, the increase in computing power means
that it's often better to just let the programs discover their
own features rather than hand-craft features using human
expertise.

Yes, of course. As you know I have long advocated that new
alpha-zero type bots should be trained through only random
checker play for backgammon and through random cube
play in addition to that for gamble-gammon.

Static features are simpler but dynamic features tend to
be more powerful, and again because of the increase in
computing power, dynamic features tend to be the way to
go if they are available.

Again, I agree in essence. However, what you call "static
features" may be used as a compromise to improve the
existing bots while waiting for alspha-zero type bots.

Often updated "static features" can also be a reasonable
compromise for "dynamic features" in the meantime.

Are you also talking about bots running on user computers
continuing to train themselves during actual play?

Would they then collect all the ongoing trainings at a hub
and redistribute to users?

What would be the likelyhood of "rooster fights" between
privately trained bots by world-class-rooster-training giants
instead of them competing themselves against one another? :)

MK

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

On April 14, 2022 at 1:51:58 PM UTC-6, Axel Reichert wrote:

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

If it makes sense to have bots with doable opening and
closing (bear-off) books, why not a doable mid-game book?

Because in contrast to bear-off positions and especially
opening positions you will almost never come across
exactly the same "landmark". Hence the cartographer's
work would be futile. No one else will visit the same spot
in forseeable time. This is caused by the branching factor
of the dice, even if an algorithm plays the same position
always in the same way.

After a break, I'm back. Sorry folk, I can't stop posting
tonight. I'm still under the influence of a covid booster.

Instead of talking about them as Manhattan or Seattle
size, I wanted to use the actual numbers of possible
opening and closing positions. I searched and found
that there are 54,263 possible bear-off positions but
couldn't find the number for opening positions. Does
anyone know it offhand?

About mid-game positions, I understand the branching
factor and all that but I personally think that they won't
branch out as much as what would be expected.

I believe the "theoretical" number of possible positions
is calculated based on backgammon and if so, I would
propose that the branching factor for gamble-gammon
will be smaller.

For either game, as the skill level of the opponents go
up, the "actual" branching factor will go down because
the moves will become more consistent and thus more
repetitive, causing the resulting positions to branch out
less and occur more frequently.

I'm as happy to offer my estimates and opinions on it
as anyone else of you but I would be much happier to
have some actual numbers. Even if they may be based
on a small sample of only 100,000 games and thus not
very accurate, anything will be better than nothing.

(Please guys, spare me from dickhead comments like
I should count them myself and such. Thanks.)

MK

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

On April 14, 2022 at 2:04:40 PM UTC-6, Axel Reichert wrote:

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

On April 12, 2022 at 1:53:06 AM UTC-6, Axel Reichert wrote:

https://en.wikipedia.org/wiki/Graph_(discrete_mathematics)

And now it's my turn to ask you what is the practical use (if
any) of this? Can you expand on what you mean by "with all
its consequences regarding application of graph theory"??

I thought we were still in brainstorming mode? Also, it is YOUR
burden to tell what you are after. I am just helping out with ideas.

Come on Axel, I can tell when you are trying to helpt out with
ideas and when you are "mathshitting" ;)

Even so, at that point I was only asking you elaborate on your
own words which didn't make any sense to me.

Since yesterday, I probably thought about discrete math graphs
more than you did. For example, after my last comments on it,
I realized the some positions (vertices) would have unplayable
rolls (loop edges). In some special positions like closed boards
with the opponent on the bar, none of the rolls will be playable.
Thus they all will be "loops" of that vertex, with none connecting
it to nother vertices. It will look like da daisy flower floating out
there in the space by itself. In fact, I understand that in that case
it becomes a separate graph by itself with a single vertex. And
do you know what it's called then? "A bouquet". :))

So, it's again my turn again to ask you what's the practical use
(if any) of this nonsense of yours?

As before, I am just helping out with ideas. No need to flip-flop
back into insulting mode.

Wow. I didn't know that you were so sensitive to feel insulted by
by my calling some of your comments "nonsense".

I was going to make pun of you by calling you "nonsensitive" but
I decide not to, fearing that you may think I made fun of you and
run off crying... ;)

Some paragraphs before, you were hypothesising about
distinguishing bots from human, now humans are out.

No, they are not out. I only clarified that by "train ride" I meant
bot-vs-bot play. We can still talk about humans riding cars,
horses, bikes... And the idea is to differentiate between styles,
strategies, etc. more than simple between bots and humans.

And since you mention "measuring", but do not say what you
want to measure, I am out as well. Good luck with your project.

This is a "nonsensitive" excuse. If the kitchen got too hot for
you, fine, you can go.

If you want to come back later to accept by your own will that
some of your comments were wrong, incomplete, useless or
whatever, I'll allow you to be polite to yourself and use your
choice of nice words... ;) Okay?? :)

MK

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

On 4/14/2022 4:04 PM, Axel Reichert wrote:

As before, I am just helping out with ideas. No need to flip-flop back
into insulting mode.

Fun fact: there's an early Persian version of the fable in which it
stings a turtle rather than a frog.

---
Tim Chow

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

On 4/14/2022 10:40 PM, MK wrote:

You guys are stingy of donating a few hours to a
project for the sake of the game but you don't mind
wasting as many hours at being dickheads to me. :(

Someone should invent a saying about the color of common kitchen items.

---
Tim Chow

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

On Friday, April 15, 2022 at 5:30:20 AM UTC-3, MK wrote:

[...]
(Please guys, spare me from dickhead comments like
I should count them myself and such. Thanks.)
[...]
Wow. I didn't know that you were so sensitive to feel insulted by
by my calling some of your comments "nonsense".

I was going to make pun of you by calling you "nonsensitive" but
I decide not to, fearing that you may think I made fun of you and
run off crying... ;)
[...]
This is a "nonsensitive" excuse. If the kitchen got too hot for
you, fine, you can go.

If you want to come back later to accept by your own will that
some of your comments were wrong, incomplete, useless or
whatever, I'll allow you to be polite to yourself and use your
choice of nice words... ;) Okay?? :)

MK

And once again, Murat is left alone, king in his island, talking to the winds. His lack of manners and zeppelin-sized ego strike again, leading him to dilapidate his interlocutors' willingness to contribute to a discussion he himself proposed.

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

Grunty <gruntingdwarf@yahoo.com> writes:

And once again, Murat is left alone, king in his island, talking to
the winds. His lack of manners and zeppelin-sized ego strike again,
leading him to dilapidate his interlocutors' willingness to contribute
to a discussion he himself proposed.

(-:

Or, since we were having it from trivial programming tasks:

1. Come up with a vague idea that is somewhere between weird and
stimulating.

2. Get criticism from a diverse range of others, who point out
theoretical issues.

3. Fail to know/learn the theory behind the others' arguments.

4. Switch to ranting/insulting mode as next attempt to get the others do
the work for you.

5. Get further criticism (theoretical, satirical) or get ignored.

6. Switch to whining mode.

7. Stay silent for some weeks.

8. Repeat.

Best regards

Axel

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

On 4/16/2022 2:41 AM, Axel Reichert wrote:

4. Switch to ranting/insulting mode as next attempt to get the others do
the work for you.

I think it would be fun to see MK try to coerce Nasti Chestikov
into doing work for him.

---
Tim Chow

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

Timothy Chow <tchow12000@yahoo.com> writes:

I think it would be fun to see MK try to coerce Nasti Chestikov
into doing work for him.

Yes! I was thinking along the same lines. At least this would have a
famous precedent: If I remember correctly, David Hilbert (though upon re-searching it, it could also have been Edmund Landau) was for a time responsible to deal with claimants of the Wolfskehl prize for proving
Fermat's last conjecture. Since it was a highly tiresome work, at some
point he answered to a claimant along these lines: "Unfortunately the mathematics you are using is too difficult for me, but I will refer you
to another more knowledgeable expert in this field, namely ..." (he
inserted the name of another claimant here). This turned out to reliably
dry up the stream of claimant correspondence. (-:

Best regards

Axel

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

On April 15, 2022 at 9:17:35 PM UTC-6, Tim Chow wrote:

On 4/14/2022 10:40 PM, MK wrote:

You guys are stingy of donating a few hours to a
project for the sake of the game but you don't mind
wasting as many hours at being dickheads to me. :(

Someone should invent a saying about the color of
common kitchen items.

I'll try if you tell me what color you want me to call you..?

MK

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

On April 15, 2022 at 11:50:31 PM UTC-6, Grunty wrote:

On Friday, April 15, 2022 at 5:30:20 AM UTC-3, MK wrote:

And once again, Murat is left alone,

I was not "left" alone by them. I chased them off myself.

king in his island, talking to the winds.

This neither new nor the first time. During my 26 years in
RGB, I've been mostly talking to the winds...

His lack of manners and zeppelin-sized ego strike again,

Being a zeppeling among hot air balloons wasn't by my
choices. Manners and all are part the role I got cast here.
It's a dirty, nay, disgusting job but if somebody has to do it
but nobody else here has what it takes to do it, and if the
reward is worth it, then I don't mind being the one doing it.

leading him to dilapidate his interlocutors' willingness to
contribute to a discussion he himself proposed.

What you may have not noticed is that static noises were
already degrading the sound quality of the discussion...
In radio terminology static noise is called "parasite". Think
of an elephant dirtying himseld by taking a mud bath in
order to get rid of the parasites...

MK

PS: Have you guys ever heard the King Crimson song
called "I talk to the wind"? Here are the partial lyrics
and a link to listen to it:

https://www.youtube.com/watch?v=UlKrH07au6E

.....
I'm on the outside looking inside
What do I see?
Much confusion, disillusion
All around me

I talk to the wind
My words are all carried away
I talk to the wind
The wind does not hear, the wind cannot hear

You don't possess me, don't impress me
Just upset my mind
Can't instruct me or conduct me
Just use up my time

I talk to the wind
My words are all carried away
I talk to the wind
The wind does not hear, the wind cannot hear
.......

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

On April 15, 2022 at 9:22:02 PM UTC-6, Tim Chow wrote:

On 4/14/2022 4:04 PM, Axel Reichert wrote:

As before, I am just helping out with ideas. No need
to flip-flop back into insulting mode.

Fun fact: there's an early Persian version of the
fable in which it stings a turtle rather than a frog.

What is fun about that..?? BTW, not all stingers are
bad. Think of the stinging nettle with its numerous
beneficial and medicinal uses. When you run out of
toilet paper during a discussion, just don't try to wipe
yourself with it.

How about a real fun math fact gift from me to you?

When a sick pack of mother loving, rooster sucking,
indiscreet mathematicians form a Markovian daisy
chain, in discrete mathematics it's called a "directed
cycle graph"... ;)

https://en.wikipedia.org/wiki/Cycle_graph

MK

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

On April 16, 2022 at 12:41:41 AM UTC-6, Axel Reichert wrote:

1. Come up with a vague idea that is somewhere between
weird and stimulating.
2. Get criticism from a diverse range of others, who point
out theoretical issues.
3. Fail to know/learn the theory behind the others' arguments.
4. Switch to ranting/insulting mode as next attempt to get
the others do the work for you.
5. Get further criticism (theoretical, satirical) or get ignored.
6. Switch to whining mode.
7. Stay silent for some weeks.
8. Repeat.

Here is my version:

1- Murat comes up with an interesting idea regarding BG.
2- After mulling over it for a while, he introduces to RGB.
3- Some people find it "stimulating" and join the discussion.
4- They fail to add much to it but they can't let Murat own it.
5- They start bulshitting big, mostly mathshitting since there
is quite a crowd of sick gambler mathematicians in RGB,
about even things not directly related to the subject, in
order to rise above Murat.
6- Murat slaps senseless one of the senseless sick puppies.
7- Little puppy feels offended and starts wailing and yipping.
8- Hearing him, all the other hyena puppies of RGB come out
of their holes, break their silence and all start yipping and
nipping at Murat's toes.
9- At first he is tickled and amused. Then he shoos them away.
10- The chorus of yelping, yowling, howling pack scatters off.
11- They go back to their daily addictive, compulsive routine
of discussing positions, pip counting, etc. ad nauseam.
12- After a hiatus, Murat comes up with a new idea about BG.
13- The routine repeats... :)

MK

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

On April 18, 2022 at 2:04:13 AM UTC-6, MK wrote:

PS: Have you guys ever heard the King Crimson song
called "I talk to the wind"? Here are the partial lyrics
and a link to listen to it:

https://www.youtube.com/watch?v=UlKrH07au6E

Shit, I'm on my second glass of wine at 4:00am and
following up to my own post, to myself.

This is all Grunty's fault. Please folks, stop reading.

After hearing the above song, I couldn't keep myself
from also listening to "Epitaph". But I managed to stop
there.

When that album came out, I was in 12th grade in a
boarding school. The best we could have then were
small transistor radios, with sound quality of police
walkie-talkies, that we could only listen to pressed
against our ears, with our head under our pillows.
Going to sleep listening to one of those songs, despite
the "parasite", was the high point of our days.

But what I want to talk about now is something else.
On that same youtube page, there were links to many
ELP songs. One of which took me weirdly from 1970's
to 2010's...

After the movie "Burlesque", all of a sudden an aged,
once wanna-be ballerina, then dance instructor to
children, had gathered enough bored wives in a small
rural town to put on a burlesque show. Frankly, they
did well beyond expectations, in their sexy fishnets.

After rehearsals, they would join everyone else in our
regular pubs where everyone almost knew everyone else.

One night, one of the youngish players complained to
me that she was scolded and forbidden from chewing
gum on the stage because that would interfere with her
acting.

I told her to go tell the older (near my age) chreographer
that Greg Palmer could sing while chewing gum.

It didn't register with her even enough to give me a blank
stare. It just flew over her head. But what if she had told
her boss? Would she know what I was referring to?

Here: https://www.youtube.com/watch?v=7nyt57LxWy8

Buth then again how many people on this entire planet
would have ever noticed it and then remembered it 40+
years later and then again 10+ years later talking about
the frequency of BG positions.

I guess this is what happens when different Markov
chains and discrete math graphs collide like galaxies
swirling into one another, combinatorial supernovae
exploding, etc... This is all Grunty's fault... :(

MK

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

On 4/18/2022 3:21 AM, MK wrote:

What is fun about that..??

It's fun because the comment elicits more entertaining nonsense
from you.

---
Tim Chow

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

On Saturday, 16 April 2022 at 14:07:16 UTC+1, Tim Chow wrote:

On 4/16/2022 2:41 AM, Axel Reichert wrote:

4. Switch to ranting/insulting mode as next attempt to get the others do the work for you.

I think it would be fun to see MK try to coerce Nasti Chestikov
into doing work for him.

---
Tim Chow

Or Timothy Chow coming out as Alex Choi and admitting that he hawks fast cars around YouTube.

So transparent.

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

On 4/26/2022 1:30 PM, Nasti Chestikov wrote:

On Saturday, 16 April 2022 at 14:07:16 UTC+1, Tim Chow wrote:

On 4/16/2022 2:41 AM, Axel Reichert wrote:

4. Switch to ranting/insulting mode as next attempt to get the others do >>> the work for you.

I think it would be fun to see MK try to coerce Nasti Chestikov
into doing work for him.

---
Tim Chow

Or Timothy Chow coming out as Alex Choi and admitting that he hawks fast cars around YouTube.

So transparent.

I don't hawk fast cars much any more because I'm so busy
developing Eraser.

---
Tim Chow

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

On April 27, 2022 at 6:46:49 AM UTC-6, Tim Chow wrote:

I don't hawk fast cars much any more because I'm so busy
developing Eraser.

This is your fifth or sixth asinine one-liner just in this thread diluting/degrading the discussion along with several others
from your ilks... :(

Why can't you pathetic little sick cretins stay out of it if you
have nothing meaningful to contribute?

Why can't you mother loving, rooster sucking scumbags do
it in your own moronic discussions in threads that you can
start yourselves instead of spoiling mine?

MK

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

On 4/27/2022 5:46 PM, MK wrote:

Why can't you pathetic little sick cretins stay out of it if you
have nothing meaningful to contribute?

Why can't you mother loving, rooster sucking scumbags do
it in your own moronic discussions in threads that you can
start yourselves instead of spoiling mine?

If your sock puppet talks, it's only common courtesy to respond!

---
Tim Chow

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