Backgammon is a game of skill and luck, we all know that.

I am going to write a DLL to use for the dice rolls in eXtreme Gammon.

It will eliminate any of those "lucky" 6-6s in the bear-off that turn me into being 97% favourite to the bot being 97% favourite (AI Factory, I'm looking at you here).

The DLL will roll two dice - let's say the bot gets first dibs, it makes it's roll. The user then gets the same two dice, he/she makes his/her roll.

Then a new set of dice are generated, the bot makes it's roll, the user gets the same two dice and moves accordingly.

Rinse and repeat.

Any thoughts, caveats? This will ensure that luck is completely eliminated and the better player wins, yes?

On 12/1/2022 12:31 PM, Nasti Chestikov wrote:
> Backgammon is a game of skill and luck, we all know that.
>
> I am going to write a DLL to use for the dice rolls in eXtreme Gammon.
>
> It will eliminate any of those "lucky" 6-6s in the bear-off that turn me into being 97% favourite to the bot being 97% favourite (AI Factory, I'm looking at you here).
>
> The DLL will roll two dice - let's say the bot gets first dibs, it makes it's roll. The user then gets the same two dice, he/she makes his/her roll.
>
> Then a new set of dice are generated, the bot makes it's roll, the user gets the same two dice and moves accordingly.
>
> Rinse and repeat.
>
> Any thoughts, caveats? This will ensure that luck is completely eliminated and the better player wins, yes?

Well, it won't eliminate luck, since a great roll for one player is a
terrible roll for the opponent.

For example, an early 66 is great for one player, but stinks for the
second player.

Also, in no-contact positions the result would be a foregone conclusion
most of the time, eliminating the skill factor of using the cube.

I don't think this idea will accomplish what you think it does.

On 12/1/2022 12:31 PM, Nasti Chestikov wrote:
> Any thoughts, caveats? This will ensure that luck is completely eliminated and the better player wins, yes?

As long as there are dice, there is going to be luck. Who
gets to play first? That's still determined by luck, right?

I seem to recall that there have been some attempts to play
"duplicate backgammon" where you have teams of two people.
Say Austin and Alyssa are on one team, and Brian and Brooke
are on another team. Austin plays Brian and Alyssa plays
Brooke. The referees ensure that whatever rolls Austin gets
are identical to whatever rolls Brooke gets, and whatever
rolls Brian gets are identical to whatever rolls Alyssa gets.
The players don't get to see what's happening on the other
board until the games are over. This might sound like a good
way to eliminate luck, just as duplicate bridge eliminates
the luck of the deal. The trouble is, as soon as someone
makes a play on one board that diverges from the play on the
other board, the fact that the dice are the same on both
boards becomes almost irrelevant. For example, suppose the
first roll of the game is 52, which Austin plays 24/22 13/8
and Brooke plays 13/11 13/8. Then suppose the next roll is
55. The positions on the board are now massively different;
Austin now has two checkers on the bar and snake eyes are
disastrous for him, while snake eyes are a joker for Brooke.
Brooke's huge advantage compared to Austin at this point
cannot be reasonably attributed to any difference in skill.

---
Tim Chow

On Friday, December 2, 2022 at 1:25:29 PM UTC, Tim Chow wrote:
> On 12/1/2022 12:31 PM, Nasti Chestikov wrote:
> > Any thoughts, caveats? This will ensure that luck is completely eliminated and the better player wins, yes?
> As long as there are dice, there is going to be luck. Who
> gets to play first? That's still determined by luck, right?
>
> I seem to recall that there have been some attempts to play
> "duplicate backgammon" where you have teams of two people.
> Say Austin and Alyssa are on one team, and Brian and Brooke
> are on another team. Austin plays Brian and Alyssa plays
> Brooke. The referees ensure that whatever rolls Austin gets
> are identical to whatever rolls Brooke gets, and whatever
> rolls Brian gets are identical to whatever rolls Alyssa gets.
> The players don't get to see what's happening on the other
> board until the games are over. This might sound like a good
> way to eliminate luck, just as duplicate bridge eliminates
> the luck of the deal. The trouble is, as soon as someone
> makes a play on one board that diverges from the play on the
> other board, the fact that the dice are the same on both
> boards becomes almost irrelevant. For example, suppose the
> first roll of the game is 52, which Austin plays 24/22 13/8
> and Brooke plays 13/11 13/8. Then suppose the next roll is
> 55. The positions on the board are now massively different;
> Austin now has two checkers on the bar and snake eyes are
> disastrous for him, while snake eyes are a joker for Brooke.
> Brooke's huge advantage compared to Austin at this point
> cannot be reasonably attributed to any difference in skill.
>
> ---
> Tim Chow

I've heard this argument before which does have some validity.
I've never played this version, but I do like the idea a lot and would
like to see it tried more.
Two counter-arguments. There are a few rolls that have a very high
probability of being lucky -- for example an early 31 and early doubles.
So, it's good to see the luck of these events approximately cancelled out.
Yes, somebody could roll these doubles and thereby dance on the bar,
but the likely effect is for these lucky events to be made less lucky or for
the luck to cancel out completely.
I also think there's a reasonable likelihood that the racing phases of the two
games will overlap.
For all pure races, the luck of the roll has an extremely strong positive correlation
with the pip count of the roll. So if the races overlap, we'll get a lot of this
luck-cancelling or reduced luck.

Paul

On December 1, 2022 at 10:31:14 AM UTC-7, Nasti Chestikov wrote:

> I am going to write a DLL to use for the dice rolls
> in eXtreme Gammon.
> The DLL will roll two dice - let's say the bot gets
> first dibs, it makes it's roll. The user then gets the
> same two dice, he/she makes his/her roll.
> Any thoughts, caveats?

Personally, I think it's impossible to measure luck
in BG. Trying to do so is just a vain, pretentious
attempt to make science out BG. Since you can't
measure it, you can't compare the luck of players
either but I will nevertheless suggest an idea based
on measuring (done by the bot and assumed to be
accurate) and comparing luck.

After the 1st player rolls and plays, and before the
2nd player rolls, look at the "dice map" ("temperature
map in Gnubg) of equities of all possible rolls and
roll the numbers with the closest equity (i.e. "luck")
to what the 1st player had gained.

Keep track of the -/+ equity difference so that when
the 2nd player is on roll again, you can adjust his
nearest "equity to be" by it (and keep repeating so).

The players may not necessarily make the best use
of their "luck" but will be at least given equal "luck",
(with, of course, all these statements being based
on the bot's calculations of "equity" and "luck"...)

I don't think this can be done by writing and external
DLL but it would be easy to add it to Gnubg, since it
already has a temperature map feature and the inane
"dice manipulation" feature which can be modified for
this purpose and turned into something a little more
useful. However, I don't think the Gnubg team cares
to do any real improvements that would broaden the
minds (or "horizons" as in the cliche) of BG players. :(

MK

On Friday, 2 December 2022 at 20:17:15 UTC, MK wrote:
>
> However, I don't think the Gnubg team cares
> to do any real improvements that would broaden the
> minds (or "horizons" as in the cliche) of BG players. :(
>
> MK

Yes I was somewhat disappointed that the latest release the other week (the first update in four years) contained no groundbreaking advances.

I'd love it if they brought back the "read dice from user DLL" feature that I understand was present many moons ago (and yet, strangely dropped from subsequent releases).

On Friday, 2 December 2022 at 15:41:57 UTC, peps...@gmail.com wrote:

> I've heard this argument before which does have some validity.
> I've never played this version, but I do like the idea a lot and would
> like to see it tried more.
>
> Paul

I was a little disappointed that there was so little appetite for your idea here

https://groups.google.com/g/rec.games.backgammon/c/IQjNOzKU0rk

and Murat's subsequent idea that everyone should use the same sequence of dice rolls and settings......it would have been a truly level playing field.

On 12/2/2022 10:41 AM, peps...@gmail.com wrote:
> Two counter-arguments. There are a few rolls that have a very high
> probability of being lucky -- for example an early 31 and early doubles.
> So, it's good to see the luck of these events approximately cancelled out.

This is a good point.
> I also think there's a reasonable likelihood that the racing phases of the two
> games will overlap.

I'm less sure about this point. I can see that the luck will be
significantly reduced if the racing phase starts at *exactly* the
same time and the pip counts are approximately the same on both
boards. But all it takes to disturb the balance is for there to be
one big roll and for the race to start one roll later on one board.

I think the main problem is that the concept of a "team" would be
a hard sell in backgammon. It's caught on in bridge, but I have
a hard time seeing it catch on in backgammon. It's also harder to
implement in backgammon as a matter of logistics. In practice, I'm
sure tournament organizers would resort to electronic dice, and
players might not trust electronic dice.

---
Tim Chow

On Sunday, December 4, 2022 at 2:28:01 PM UTC, Tim Chow wrote:
> On 12/2/2022 10:41 AM, peps...@gmail.com wrote:
> > Two counter-arguments. There are a few rolls that have a very high
> > probability of being lucky -- for example an early 31 and early doubles.
> > So, it's good to see the luck of these events approximately cancelled out.
> This is a good point.
> > I also think there's a reasonable likelihood that the racing phases of the two
> > games will overlap.
> I'm less sure about this point. I can see that the luck will be
> significantly reduced if the racing phase starts at *exactly* the
> same time and the pip counts are approximately the same on both
> boards. But all it takes to disturb the balance is for there to be
> one big roll and for the race to start one roll later on one board.
>
> I think the main problem is that the concept of a "team" would be
> a hard sell in backgammon. It's caught on in bridge, but I have
> a hard time seeing it catch on in backgammon. It's also harder to
> implement in backgammon as a matter of logistics. In practice, I'm
> sure tournament organizers would resort to electronic dice, and
> players might not trust electronic dice.
>
>
The extent to which this form reduces the luck would seem readily measurable.
Simply look at the correlation between the results and the PR differentials.

Doubles seems reasonably common in backgammon, and those are teams.
I like the team arrangement where one player is much stronger than the other, and
the weaker player does sanity checking for clear blunders. If there's not a clear leader,
you just get squabbling. Another approach might be to agree in advance with your doubles
partner that, whenever you have an unresolved disagreement over a play (meaning the captain's
play happened but the non-captain still hated it), then there will always be a bot prop on the issue
for a reasonable stake agreed in advance. That way, no disagreement need be too frustrating.
If you don't get your play, then either your play is not clearly right (and you shouldn't complain) or
you will win money in the prop (and therefore also shouldn't complain).

Paul

On 2022-12-02, MK <murat@compuplus.net> wrote:

> After the 1st player rolls and plays, and before the
> 2nd player rolls, look at the "dice map" ("temperature
> map in Gnubg) of equities of all possible rolls and
> roll the numbers with the closest equity (i.e. "luck")
> to what the 1st player had gained.

This won't work very well since the rolls (with the exception of the
first one) will not be random but determined by the previous rolls and
plays.

For instance, after 3-1, the second will will always be the same. After
2-1 it will probably be different after 2-1 and split and 2-1 and slot,
but it will still be predetermined.

The idea is good, but you should apply the correction elsewhere: let the
dice be random and substract the luck from the game or match result.

This is what is called "luck adjusted result" in gnubg (and something
less obvious in XG, but the feature is available in it as well).

Winning or losing fractions of points in money games or some percentage
of the match in match play is a bit puzzling at first, though.

On December 4, 2022 at 3:36:41 PM UTC-7, Philippe Michel wrote:

> On 2022-12-02, MK <mu...@compuplus.net> wrote:

>> After the 1st player rolls and plays, and before the
>> 2nd player rolls, look at the "dice map" ("temperature
>> map in Gnubg) of equities of all possible rolls and
>> roll the numbers with the closest equity (i.e. "luck")
>> to what the 1st player had gained.

As I sat down to write an essay on leveling of luck,
Phillip's post came in. I'll postpone it and respond
to him first, since this is a more specific subject.

> This won't work very well since the rolls (with the
> exception of the first one) will not be random but
> determined by the previous rolls and plays.

Either I wasn't clear or you misunderstood. The first
player always rolls random, (he is the driver), after
each of his rolls the second player will be given an
equally lucky/unlucky roll (he is the passenger but
in the same car). I will illustrate with an example.

> For instance, after 3-1, the second will will always
> be the same. After 2-1 it will probably be different
> after 2-1 and split and 2-1 and slot, but it will still
> be predetermined.

Let's say X rolls a 31 first. The best move 8/5 6/5
gives him +0.231 equity.

To give O a roll with similar "luck equity", we look at
the temperature map and see that 33 is the closest
with +0.093 and we give O a hand-picked 33 roll.

At that point, O is still behind by +0.138 which we
will add to his next hand-picked roll.

Then X randomly rolls again, let's say a 52 giving him
a -0.116 equity. Now he is only ahead by +0.022.

We again look at the temperature map and see that
64 is the closest with +0.033 and we give O a 64.

At that point, X falls behind by -0.011 which we will
subtract from O's next hand-picked roll.

I hope you guys get the idea...?

If O had started with an opening 21 instead, giving
him only +0.003 equity by playing 24/23 13/11 as
best move. Then from the map we would pick a 61
for X which would give him +0.000 as nearest equity.

After O would roll randomly again and play, we would
add +0.003 to X's next "quity-to-be" and hand-pick a
roll from the map that would place him closest to O.

And so on...

> The idea is good,

Yes. Thanks for acknowledging.

> but you should apply the correction elsewhere:
> let the dice be random and substract the luck
> from the game or match result.

No. The two are not the same thing. In mine, players
don't wait for luck to level out after the fact when it's
already too late. Instead, luck is constantly kept as
level as possible throughout the duration of each
individual game, roll per roll. This is the beauty of it. ;)
This is what sets it apart from retroactive calculations.

You all be warned, though, that leveling luck this way
will cause serious damage to the so-called "cube skill
theory"...! (I'll explain this more in my next essay but
you all can already guess what's going to happen :).

> This is what is called "luck adjusted result" in
> gnubg (and something less obvious in XG, but
> the feature is available in it as well).
> Winning or losing fractions of points in money
> games or some percentage of the match in
> match play is a bit puzzling at first, though.

As I said above, what I'm proposing has absolutely
nothing to do with all that after-the-fact, retroactive
mathshittings... :( (I say this because in a world where
"PR-sacrificing" moves are made, making the most out
of one's luck is not necassary; there is no "wastage of
luck"! You all may want to write down this expression
that I coined. It may prove more valuable that the idea
of "wastage of pips"...)

Gnubg has that stupid "dice manipulation" feature that
looks at the temperature map to hand-pic rolls for 21
levels (no freaking less!!) of good/bad numbers.

Just trash that "absurd feature". The code you need is
already in there. Take a few lines from it, make a few
minor changes to pick the roll with the closest equity
to what is desired and throw out the the rest of the no
longer deeded code. It should be trivial for you to do.

You really have no excuses for refusing to make this
"improvement" to Gnubg, which will throw a monkey
wrench into the "jackoffski cube skill theory bullshit"
also, that I'm sure you all will greatly enjoy... :))

MK

On 12/4/2022 9:38 AM, peps...@gmail.com wrote:
> Doubles seems reasonably common in backgammon, and those are teams.

Those with more experience than I do might want to weigh in, but my
impression is that doubles are largely regarded as being "for fun,"
whereas the idea of duplicate backgammon is presumably to emphasize
skill.

Chouettes are another example of team play in backgammon, but again,
my impression is that people who play for significant stakes tend to
favor non-consulting chouettes, which largely eliminates the "team"
aspect of the game.

Anyway, the main argument is probably that the idea has been tried
a number of times before, and it has never caught on. You can read
a little more about it here:

http://www.bgonline.org/forums/webbbs_config.pl?read=48603

https://bkgm.com/variants/Duplicate.html

---
Tim Chow

On 2022-12-05, MK <murat@compuplus.net> wrote:

> Either I wasn't clear or you misunderstood. The first
> player always rolls random, (he is the driver), after
> each of his rolls the second player will be given an
> equally lucky/unlucky roll (he is the passenger but
> in the same car).

OK. This is not how I had understood it.

Still, these pairs of similarly lucky rolls would lead to some weird
situations. For instance if player 1 rolls a big doublet in a close
race, so will player 2? If player 1 rolls badly in a bearoff against a
low anchor and exposes two blots, player 2 will roll badly as well and
miss them?

I suspect that the manipulation to level out luck so quickly would be
more obvious and unnatural than you think.

On December 8, 2022 at 3:58:59 PM UTC-7, Philippe Michel wrote:

> On 2022-12-05, MK <mu...@compuplus.net> wrote:

>> Either I wasn't clear or you misunderstood. The first
>> player always rolls random, (he is the driver), after
>> each of his rolls the second player will be given an
>> equally lucky/unlucky roll (he is the passenger but
>> in the same car).

> OK. This is not how I had understood it.

What I first wrote wasn't clear at all. I'm glad you are
still interested in the subject. I've been thinking more
on it for the past few days and working on an essay
combining it with what I had in mind before, while at
the same time giving you all time to digest the idea
and pursue it further.

> Still, these pairs of similarly lucky rolls would lead to
> some weird situations. For instance if player 1 rolls
> a big doublet in a close race, so will player 2? If player
> 1 rolls badly in a bearoff against a low anchor and
> exposes two blots, player 2 will roll badly as well and
> miss them?

Such situations may arise if neither players owes any
equity to the other. Remember that a running balance
of equity will be kept and updated after every roll, so
that the "passenger" player can be given a better or a
worse roll. This is a key aspect of it. I hope you all do
undertand this.

Even if "equity balance" allows big doublets for both
players in a close race, they may not necessarily help
both players equally.

Player 2 with a low anchor is likely to get stuck behind
a block for a few rolls and "accrue" some "luck equity"
which will be given to him as soon as possible in one
of his next rolls. So, in your example above, he actually
may be given a "luck equalizing" roll that hits both blots.

> I suspect that the manipulation to level out luck so
> quickly would be more obvious and unnatural than
> you think.

I'm not so sure that what you are saying will happen
even in equally skilled bot vs bot play because there
may/will be positions where the closest equity to be
given to player 2 may still be substancially different
than what player 1 rolls.

I'm pretty sure that things will look more natural in bot
vs human play, and even more natual in human vs
human play because error moves (or PR-sacrificing
moves) will be made. Skill will emerge maybe slower
than "normal" but more visibly and decisively, which is
actually exactly what it tries to accomplish.

You guys take your times. Think about it some more.
There is no need to rush. Once you understand well
enough I think you will recognize its value as another
"mutant tool" to experiment with, especially to observe
what happens to that so-called "cube skill theory" in
the absence of excessive luck swings...

In fact, I'm tempted to give it a bg variant name such
as LevelGammon, because people may come to like
playing it, especially against bots, (considering also
its added benefit that it would limit, if not eliminate,
any possible cheating by some bots;)

As I said, the code to pick a roll from the temperature
map is already there in Gnubg. You only need to add a
variable to keep a running tab on equity. I hope you will
see that with a minimal effort, this can be a very useful
tool and/or even a possible variant.

MK

On December 8, 2022 at 7:18:48 PM UTC-7, MK wrote:

> As I said, the code to pick a roll from the temperature
> map is already there in Gnubg. You only need to add
> a variable to keep a running tab on equity. I hope you
> will see that with a minimal effort, this can be a very
> useful tool and/or even a possible variant.

I'm sure Philippe is more than capable of doing it
himself but because I would like to see this done
very much and thus motivate him to do it, I want
to contribute some effort myself also as a show
of goodwill.

In play.c there is a section that sorts the array of
equities for the 21 possible dice rolls. I think that
would be the best place to add one more line to
compare which array element is closest to the
"equity balance" of the "passenger player" and
save the array index for it in a variable.

And where it decides which nth best equity to roll,
add one more line to check if luck leveling option is
selected (i.e. 22nd level?) and use the above saved
index to roll the numbers from that array element,

You will need two more lines; one after each time
the "driver" player rolls, to add the max equity from
that roll to the "equity balance" of the "passenger"
player, and one after each time the "passenger"
player moves (or fails to move) to subtract the used
(or add back the unused) equity to/from his "equity
balance".

That should be all that is needed. Oh, of course, a
22nd option must be added to the drop-down list
where users select the level of "dice manipulation" ;)

Don't worry about whether it will lead to some weird
situations and such. Because it's so easy to do, just
do it please and let's all see what will come out of it,
whatever that may be...

MK

You can try this yourself without having to code. I did this and played
10 games this way. Here is how to do this:

- Set GNU Backgammon to manual dice.
- Roll a pair of real dice (different colours) at hand
- Determine who starts (e.g. GNU Backgammon is assigned the red die)

You can try this yourself without having to code. I did it and played 10
games this way. Here is how to do this:

1. Set GNU Backgammon to manual dice, "Expert" play, "Expert" analysis,
and start a new money session.

2. Roll a pair of real dice (different colours, e.g. gnubg has the red
one). Say, gnubg rolls a 4 and you roll a 1. By now, gnubg has become
"luck master" and you will be the "luck servant".

3. Enter this opening roll in the "Dice" dialog. gnubg will move 24/23
13/9.

4. Click on "Analyze", "Analyze Game", then "Match or Session
Statistics", then the "Luck" tab. Take note of the value for the "luck
master" and subtract the value for the "luck servant". The difference
is -0.006. (I will ignore any discussion about the values here.)

5. Click on "Analyze", "Distribution of rolls". Note the value for
"Average equity". It is +0.012. Add the above difference (-0.006) to
get to +0.006. Memorize the roll closest to this +0.006. It is 64. Try
not to peek at the correct move!

6. Enter 64 in the "Dice" dialog and complete your move.

7. Now say you roll 33 for gnubg. It moves bar/22 8/5(3).

8. As per item 4, the luck difference is now +0.166 - (-0.003) = +0.169.

9. As per item 5, the average equity is -0.132, and after adding 0.169
the closest roll to the target equity of 0.037 is 11. Move.

10. Repeat cycles from 2 to 6.

Yes, this was tedious, but kind of interesting. Some impressions:

- As long as you do not blunder, the game stays very close (the equity
oscillates just a little around 0). A pretty static tug-of-war.

- Doubles hence are rare (unless you have lost considerable equity
before)

- I won only one of the 10 games, as "luck master" with the help of
final lucky doublet in a pure race.

- It cuts down the branching factor considerably. For example, after a
42 opening (played correctly) the reply roll will always be 31. After
a big racing doublet by the master, expect the same for the
servant. After a dance on the bar by the master, expect that you will
not be able to cover your home board blot as a servant.

- Once I started to try to predict the rolls for the luck servant
(finally feeling like a dice paranoid instead of working on my game) I
was surprised quite often by how bad doublets can be, or, put more
generally, how far off my estimations were. There is a lesson to be
learned by some here.

- There are highly artificial dice sequences: I saw 8 subsequent 21s for
the servant in a race in order to compensate for some previously
accumulated luck (that due to the position could not be compensated
adequately before).

- It can only be done with a computer.

- Considering the dice and luck paranoids out there, the artificial
sequences, the even more frustrating outcome against a world class
bot, I expect that Gary Wong's official dice complaint form would need
a substantial update.

Not for me, but your mileage may vary.

Axel

On January 7, 2023 at 3:40:16 AM UTC-7, Axel Reichert wrote:

Wow! What a nice effort and contribution. I must
thank you for this. I'm anxious to add my replies
and comments, hoping that this will inspire others
to participate. So let me start.

> You can try this yourself without having to code.
> I did it and played 10 games this way.

Yes, of course. I must admit that I haven't played
complete games but I experimented a little before
suggesting that this would lever luck immediately
with the dice still "random enough in the long run".

Have you saved your games? If so, can you paste
them here in text format?

> Here is how to do this:
> 1. Set GNU Backgammon to manual dice, "Expert"
> play, "Expert" analysis, and start a new money session.

It's great that you spelled out the steps. Even after
I clarified my instructions for Philippe, they weren't
this detailed and clear.

> 2. Roll a pair of real dice (different colours, e.g.
> gnubg has the red one). Say, gnubg rolls a 4 and
> you roll a 1. By now, gnubg has become "luck
> master" and you will be the "luck servant".

I don't understand the need for colored dice but it's
no big deal. I understand that your "luck master" is
my "driver", who rolls random dice for himself but
gives "luck leveling" numbers to your "luck servant"
or my "passenger" player.

I omitted some steps which can be seen in your post.

> closest roll to the target equity of 0.037 is 11. Move.

Following these steps my numbers were a little off
but the target equity was 0.037 also. Perfect. :)

> 10. Repeat cycles from 2 to 6.

So far we are in agreement. Let's put a marker here.

> Yes, this was tedious, but kind of interesting. Some
> impressions:
> - As long as you do not blunder, the game stays very
> close (the equity oscillates just a little around 0). A
> pretty static tug-of-war.

Why not blunder? To the contrary, my intention was to
level the "luck equity" but allow the "skill equity" swing
and/or drift apart without limitations, so that whatever
skill difference may be, it would become visible quickly.

> - Doubles hence are rare (unless you have lost
> considerable equity before)

I wouldn't have expected this but maybe because you
eliminate blunders. I was worried that after one player
rolls a big double in a race, it would become a battle of
the big doubles.

> - I won only one of the 10 games, as "luck master"
> with the help of final lucky doublet in a pure race.

Only 10 games may not be telling much but it would be
interesting to see how your games enfolded.

If this get added to Gnubg, we can run long experiments
with different plies of the bot play against each other to
see what happens in 1,000 games in this variant vs. in
10,000 games in regular gamblegammon. If the results
come out similar, that would be a trophy reward for me.

Also, what about cube actions in your games? I predicted
that there would be very cautious and very very few cube
actions. This is what I'm most curious about (for obvious
reasons that you can guess.. ;)

> - It cuts down the branching factor considerably. For
> example, after a 42 opening (played correctly) the
> reply roll will always be 31.

I'm not so sure about this. Replies to opening rolls can
be memorized but only assuming best play from both
sides. Playing either roll differently won't gain anything
since it may not necessarily thwart the opponent's
subsequent correct play.

In any case, I foresee the branching factor going back
to normal after the 3rd roll. This is to be seen. Do you
think it can be determined by inspecting the games..?

> After a big racing doublet by the master, expect the
> same for the servant.

Ah, I was afraid of this but does it really matter all that
much, except the different look and feel of the games?

What's wrong with luck being leveled very fast? Other
than lack of luck swing will dull the cube activity which
I would like to see indeed as a proof that cube is fueled
by luck and not by skill...

> After a dance on the bar by the master, expect that
> you will not be able to cover your home board blot
> as a servant.

Again, I don't see these as bad things. A player will only
be able to widen the equity gap through skill, not luck.

> - Once I started to try to predict the rolls for the luck
> servant (finally feeling like a dice paranoid instead
> of working on my game) I was surprised quite often
> by how bad doublets can be, or, put more generally,
> how far off my estimations were. There is a lesson
> to be learned by some here.

I don't quite understand this paragraph. Can you expand
on it a little..?

> - There are highly artificial dice sequences: I saw 8
> subsequent 21s for the servant in a race in order to
> compensate for some previously accumulated luck
> (that due to the position could not be compensated
> adequately before).

I wonder if this will hold in the long run? If so, then it
may only mean that "dice dispersion" can have more
clusters but the "dice distribution" can still be good?
> - It can only be done with a computer.

Yes. I hope they will put this in some bots someday.

> - Considering the dice and luck paranoids out there,
> the artificial sequences, the even more frustrating
> outcome against a world class bot, I expect that
> Gary Wong's official dice complaint form would
> need a substantial update.

I don't understand this paragraph either. How do you
mean? I don't think you are being sarcastic. Are you
complaining about the dice? Why..?
> Not for me, but your mileage may vary.

If you mean you wouldn't enjoy playing this variant
gainst a bot, I can understand that but what about
using it in "bot vs. bot" and/or bot vs. "mutant bot"
experiments?

As you all know, I'm a big advocate of experiments.
I love experiments. I think this one would be a very
revealing experiment about luck and/or cube skill...

I have more to say on this but I don't want to get too
far ahead of you or others who may want to chime in.

MK

On January 7, 2023 at 2:13:35 PM UTC-7, MK wrote:

> On January 7, 2023 at 3:40:16 AM UTC-7, Axel Reichert wrote:

>> 10. Repeat cycles from 2 to 6.

> So far we are in agreement. Let's put a marker here.

After posting, I realized that there was nothing about
keeping a "running equity balance" in your numbered
steps, which is indispensable!

What you did has nothing to do with what I proposed.

The average equity in the temperature map has nothing
to with this. You look at the temperature map to find
the roll with the closest equity to the opening roll of
the "master" only once.

After that you put the difference into a "equity balance"
variable and then on you look at the temperature map
to find the roll with the closest equity to that "equity
balance" which will be updated after every move of the
"servant" and after every random roll of the "master".

Darn! :(( I got excited for nothing. What you did is all
wrong compared to what I proposed. That's probably
also the reason for your complaints about the dice.

If you want to play another 10 games properly, that
will be great. We can talk about your observations
again after that. If you don't want to do it, too bad. :(

I think I see now how you guys have such difficulty
understanding my noverl ideas. You keep going back
to what is etched into your minds... Sorry that your
efforts were in vain but I also feel like a fool to have
followed your steps without catching myself sooner. :(
So, I guess we are even... Just too bad...

MK

MK <murat@compuplus.net> writes:

> On January 7, 2023 at 3:40:16 AM UTC-7, Axel Reichert wrote:

[...]

> Have you saved your games? If so, can you paste them here in text
> format?

As you should know by now, I prefer if others get their hands dirty as
well. There is a lot to learn.

>> - As long as you do not blunder, the game stays very close (the
>> equity oscillates just a little around 0). A pretty static
>> tug-of-war.
>
> Why not blunder? To the contrary, my intention was to level the "luck
> equity" but allow the "skill equity" swing and/or drift apart without
> limitations

I understand this. But if I err/blunder (gnubg does not), then the game
does not stay close but rather crawls slowly in direction of a double
for gnubg.

>> - Doubles hence are rare (unless you have lost considerable equity
>> before)
>
> I wouldn't have expected this but maybe because you eliminate
> blunders.

I did not "eliminate" blunders, far from it. In fact I was quite busy
getting the "logistics" right and so most likely paid less attention to
the game itself. For the record: I doubled not at all (but should have,
prior to the final 33 as "luck master", which won the only game for me,
an oversight by me). In all other games gnubg doubled, I took 6 and
passed 3. So doubles are rare for the (non-expert) human.

> what about cube actions in your games?

Well neither player adapted his strategy, but from first principles I
would argue as follows:

The luck master should be careful with doubling, because there will not
be many market losers: Good rolls by the doubler will be compensated for
the luck servant. For the luck servant on roll it is different: If the
luck master rolled something good, the servant will also get a good
roll. If then the next roll by the luck master is bad (say, half of the
times), it might be a market loser. The logic for takes/passes will be
similar.

>> - It cuts down the branching factor considerably. For example, after
>> a 42 opening (played correctly) the reply roll will always be 31.
>
> I'm not so sure about this. Replies to opening rolls can be memorized
> but only assuming best play from both sides. Playing either roll
> differently won't gain anything since it may not necessarily thwart
> the opponent's subsequent correct play.
>
> In any case, I foresee the branching factor going back to normal after
> the 3rd roll.

No. The roll of the luck servant is fixed by the roll of the luck
master. Backgammon is listed with a branching factor of 250 here:

https://en.wikipedia.org/wiki/Game_complexity#Complexities_of_some_well-known_games

This boils down (21 rolls) to an assumed 250/21 = 12 moves per roll. If
one player has only one roll (servant to the master), then the
combinatorial explosion (calculated for one move each) is cut down from
250*250 to 250*12. A factor of 21 is "considerable" for me.

>> - There are highly artificial dice sequences: I saw 8 subsequent 21s
>> for the servant in a race

[...]

> I wonder if this will hold in the long run?

Anyway, this constraint was extremely annoying. The game feels crippled,
less lively. Try it yourself.

>> - Considering the dice and luck paranoids out there, the artificial
>> sequences, the even more frustrating outcome against a world class
>> bot, I expect that Gary Wong's official dice complaint form would
>> need a substantial update.
>
> I don't understand this paragraph either.

Do you think that people who dislike the amount of luck in Backgammon
(often dice paranoids) will happily join such a game variant on a
server? Do you think that weak player (who often blame the dice instead
of their game) will happily play such a variant? After all, it could
reveal how weak their play is (of course fueling further paranoia about
the dice and the luck calculations involved, which they most likely do
not even understand).

> If you mean you wouldn't enjoy playing this variant
> gainst a bot, I can understand that but what about
> using it in "bot vs. bot" and/or bot vs. "mutant bot"
> experiments?

No. I play agaist bots to prepare for play with humans. Since I would
not want to play this variant agaist humans, this is "L'art pour
l'art". Not interested.

As to your other post: I did things correctly. The luck difference
between master and servant is the amount of luck (= potential equity)
that needs to be given to the servant as quickly as possible. The
average equity in the "Distribution of rolls" dialog corresponds to an
imaginary roll with zero luck. Hence you need to add the difference from
above and pick the closest roll. All fine here. I will not elaborate
further, I believe this is all crystal clear.

My effort was not done in vain, I will not play further games either.

Axel

On January 8, 2023 at 12:58:56 AM UTC-7, Axel Reichert wrote:

> MK <mu...@compuplus.net> writes:
>
> > On January 7, 2023 at 3:40:16 AM UTC-7, Axel Reichert wrote:

>> Have you saved your games? If so, can you paste
>> them here in text format?

> As you should know by now, I prefer if others get
> their hands dirty as well. There is a lot to learn.

Fuck your holy selfishness! Especially when you
yourself have not done anything to learn from it
for your own sake... :(

>> Why not blunder? To the contrary, my intention
>> was to level the "luck equity" but allow the "skill
>> equity" swing and/or drift apart without limitations

> I understand this. But if I err/blunder (gnubg does
> not), then the game does not stay close but rather
> crawls slowly in direction of a double for gnubg.

Because you don't understand the purpose of the
experiment and you are doing it all wrong...!

>>> - Doubles hence are rare (unless you have lost
>>> considerable equity before)

>> I wouldn't have expected this but maybe because
>> you eliminate blunders.

> I did not "eliminate" blunders, far from it. In fact I
> was quite busy getting the "logistics" right and so
> most likely paid less attention to the game itself.

Okay, my comment was based on you having done
the 10-game experiment right. Since you hadn't, it
doesn't apply. Since the "master" rolls random dice,
at least he would roll an expected amount of doubles
and the "servant" can't do too much worse...

> For the record: I doubled not at all (but should have,

Ha ha ha...! This was one of my points (even if you
haven't done it right).

> prior to the final 33 as "luck master", which won the
> only game for me, an oversight by me). In all other
> games gnubg doubled, I took 6 and passed 3. So
> doubles are rare for the (non-expert) human.

I think thwy would be rare for both if you had done it
right.

>> what about cube actions in your games?

> Well neither player adapted his strategy, but from
> first principles I would argue as follows:

> The luck master should be careful with doubling,
> because there will not be many market losers.....

All this is only partially true, (i.e. partially proving my
point), again because you didn't do it right.
>> In any case, I foresee the branching factor going
>> back to normal after the 3rd roll.

> No. The roll of the luck servant is fixed by the roll
> of the luck master.

Yes but how he will play it is not fixed. Mentally ill
gamblers like you can't understand this because
you are fixated on "perfect play" bullshit...! :(

Forget about the 3rd roll, you can't even know how
your opponent will play the 1st or the 2nd roll, if he
is a human...!

You can know if it is a bot because bots are 100%
predictable and that should be an advantage for
the human player. But bott kissers can't understand
that...! :(

> Backgammon is listed with a branching factor of
> 250 here:

So what? How does that apply here is important.

Besides the fact that the branching factor does not
correlate to game complexity, as I have demonstrated
to you all in my post about "baranching factor inflation".

>> I wonder if this will hold in the long run?

> Anyway, this constraint was extremely annoying. The
> game feels crippled, less lively. Try it yourself.

It's supposed to be less lively. :) What is crippled is luck,
not skill...! This is the whole fucking point of it! And you
demonstarated that mentally ill gamblers don't like skill,
they like luck!

> After all, it could reveal how weak their play is

Yes, indeed, that's what I would like to demonstrate.

> (of course fueling further paranoia about the dice and
> the luck calculations involved, which they most likely
> do not even understand).

Let's forget about them. Let's talk about you and your
ilk. You all understand luck calculations, don't you..? ;)

>> If you mean you wouldn't enjoy playing this variant
>> gainst a bot, I can understand that but what about
>> using it in "bot vs. bot" and/or bot vs. "mutant bot"
>> experiments?

> No. I play agaist bots to prepare for play with humans.
> Since I would not want to play this variant agaist
> humans, this is "L'art pour l'art". Not interested.

Oh, wow! You almost left me speechless here. :( I have
no interest in discussing "L'amour pour l'amour" with a
whore either...!! :((

> As to your other post: I did things correctly.

No, you didn't. Why don't you make a genuine effort to
understand instead of arguing that you understand all
the while you don't?

> All fine here.

I've never been "there". So, I can't very well argue against
how things may be in your realm of sick gablers...

> I will not elaborate further,

Good decision. Why elaborate on something you don't
even have sane brains to understand...?

> My effort was not done in vain,

Actually, on second thought after my second post, I also
thought so because some of (at least my) arguments
would apply either way. In other words, if you did another
10-game experiment, there would be occasions for me
to repeat those arguments.

> I will not play further games either.

Please don't! Don't waste your time and effort on things
that you don't have the capacity to understand...

What is truely sad is tha Phillippe can impemeny this in
Gnubg in less time than it would take you to play anothe
10-games manually but unfortunately he won't, for the
reasons that I/we know very well why... :(

I will write a separate article about "LevelGammon", or
"FlatGammon", or any other suggestions? regardless
of who may understand it and/or benefit from it.

MK