IBM Backgammon

[Dave Yeo]

I've been playing backgammon quite a bit. I'm finding it quite
frustrating as the computer seems to cheat. Keeping statistics I see
things like 40 times the computer jump on my single compared to 6 for
me, usually as soon as I have a single the computer rolls the right
number to hit my single or both if I have two singles, often with
doubles. The computer can leave a single for an average of 4 rolls
before I can roll to hit it whereas the other way it is 1.2 rolls, the
singles are usually in equivalent spots.
The average game the computer will roll 8 doubles while I'll roll 2
unless I have a man off. Then, eg if the computer has 4 and 6 blocked
I'll roll double 6s, then double 4s then double 5s as now 5s are also
blocked.
12 times in a row when I had 5 out of 6 home spots blocked the computer
will roll the one unblocked number, often also rolling the other number
that will hit my single. Whereas 10 out 12 times when the computer had 3
or less blocked I couldn't roll the correct number, often getting doubles.
If I'm way in the lead I often start rolling small doubles but if I'm
slightly less in the lead, in a surprising number of times the computer
will roll multiple large doubles and pull of a surprising come from
behind win if I don't double.
I seem to be very unlucky playing this game even when following the same
strategy as the computer or it cheats.
Meanwhile playing a win3.1 backgammon game on high, the computer wins
about 60% of the time and it seems random if a single gets hit and
roughly equal.
I find it really frustrating playing a game where the player hides the
dice before showing the roll and has amazing luck game after game.
Perhaps I'm just paranoid but I'm not such a bad player that when a
2,3,4 will hit a single and I roll double ones twice in a row (by then
all singles are covered) that it is a matter of skill.
Dave

[Peter J. Seymour]

Post by Dave Yeo
I've been playing backgammon quite a bit. I'm finding it quite
frustrating as the computer seems to cheat. Keeping statistics I see
things like 40 times the computer jump on my single compared to 6 for
me, usually as soon as I have a single the computer rolls the right
number to hit my single or both if I have two singles, often with
doubles. The computer can leave a single for an average of 4 rolls
before I can roll to hit it whereas the other way it is 1.2 rolls, the
singles are usually in equivalent spots.
The average game the computer will roll 8 doubles while I'll roll 2
unless I have a man off. Then, eg if the computer has 4 and 6 blocked
I'll roll double 6s, then double 4s then double 5s as now 5s are also
blocked.
12 times in a row when I had 5 out of 6 home spots blocked the computer
will roll the one unblocked number, often also rolling the other number
that will hit my single. Whereas 10 out 12 times when the computer had 3
or less blocked I couldn't roll the correct number, often getting doubles.
If I'm way in the lead I often start rolling small doubles but if I'm
slightly less in the lead, in a surprising number of times the computer
will roll multiple large doubles and pull of a surprising come from
behind win if I don't double.
I seem to be very unlucky playing this game even when following the same
strategy as the computer or it cheats.
Meanwhile playing a win3.1 backgammon game on high, the computer wins
about 60% of the time and it seems random if a single gets hit and
roughly equal.
I find it really frustrating playing a game where the player hides the
dice before showing the roll and has amazing luck game after game.
Perhaps I'm just paranoid but I'm not such a bad player that when a
2,3,4 will hit a single and I roll double ones twice in a row (by then
all singles are covered) that it is a matter of skill.
Dave

I just found this post.
I play OS/2 Backgammon occasionally and have done for years. My
impression also is that the computer cheats. There have been such
allegations ever since the game came out. I remember such comments on
DialIBM (or whatever it was called at that point) forums in the mid
1990s. I guess the developers hadn't quite got the hang of computer game
play. It may be significant that it was supposed to be a paid-for game
but IBM quickly released it for free.

[Peter J. Seymour]

Just kind of thinking aloud about this topic. The game has a high chance
element because moves depend on the throwing of dice. The skill element
is then in deciding how to deploy the score. I am not an expert on the
subject but I wonder if in computing terms the skill element is fairly
straight forward (after you have figured out what it is). Introducing a
range of skill levels into the computerised game might then create a
difficulty if there is not much scope for variation. A simple way to
deal with this would then be to clobber the throwing of the dice so that
for increasing "skill level" the throw is progressively biased in the
computers favour. So first you determine your optimum move and then you
"throw the dice" with bias to support that move with an elevated chance
of advantage. For instance, at the simplest level, all six numbers are
included in the throw, while at the most difficult level only say three
numbers are included (included the numbers you want).

This is just speculation but it does illustrate a way in which the game
could "cheat". For the novice computer user wanting to be impressed by
the computers capabilities it might not matter, but someone more expert
is likely to notice the bias. So I submit this as a possible explanation
of what is going on in the game.

[Martin Iturbide]

This is a interesting topic.

To bad I don't play Backgammon to try it out.

Which game is? Is this one the one you think is cheating?

http://www.os2world.com/games/index.php/native-games/board/114-os-2-td-gammon

Regards
Martín.

[Peter J. Seymour]

Post by Martin Iturbide
This is a interesting topic.
To bad I don't play Backgammon to try it out.
Which game is? Is this one the one you think is cheating?
http://www.os2world.com/games/index.php/native-games/board/114-os-2-td-gammon
Regards
Martín.

Yes that is the game, but I believe there is at least a 1.1 version in
existence. I think it is on an IBM site but I don't remember where.

[Martin Iturbide]

It is very interesting.

I didn't know that that game had a Wikipedia article:
https://en.wikipedia.org/wiki/TD-Gammon

I will put the OS2World Games site picture on that article to collaborate with the Wikipedia.

Let me know if you find version 1.1. I will upload it to hobbes and the games site.

Regards
Martín.

[Peter J. Seymour]

Post by Martin Iturbide
This is a interesting topic.
To bad I don't play Backgammon to try it out.
Which game is? Is this one the one you think is cheating?
http://www.os2world.com/games/index.php/native-games/board/114-os-2-td-gammon
Regards
Martín.

I had a look around, but I can't locate where it might be available to
download.

It is interesting that there are several articles on the net extolling
the virtues of TD-Gammon. Apparently it uses a world-beating algorithm.
When I have a spare moment I will have to see if I can analyse the
apparent cheating more objectively.

[Barry Landy]

On Tue, 29 Oct 2013, Peter J. Seymour wrote:

:>On 29/10/2013 16:59, Martin Iturbide wrote:
:>> This is a interesting topic.
:>>=20
:>> To bad I don't play Backgammon to try it out.
:>>=20
:>> Which game is? Is this one the one you think is cheating?
:>>=20
:>> http://www.os2world.com/games/index.php/native-games/board/114-os-2-td-=
gammon
:>>=20
:>> Regards
:>> Mart=EDn.
:>>=20
:>I had a look around, but I can't locate where it might be available to
:>download.

click on the size part of the download line.

:>
:>It is interesting that there are several articles on the net extolling th=
e
:>virtues of TD-Gammon. Apparently it uses a world-beating algorithm. When =
I
:>have a spare moment I will have to see if I can analyse the apparent chea=
ting
:>more objectively.
:>

--=20
Barry Landy Email: Remove nospam in from address
192, Gilbert Road, Cambridge CB4 3PB

[Peter J. Seymour]

Post by Martin Iturbide
This is a interesting topic.
To bad I don't play Backgammon to try it out.
Which game is? Is this one the one you think is cheating?
http://www.os2world.com/games/index.php/native-games/board/114-os-2-td-gammon
Regards
Martín.

I have been on an OS/2 system since my previous post (well, ECS 1.2R to
be precise). That has TD-Gammon 2.0 installed.

I see that the game has six levels of hardness. That puts my variable
faced dice hypothesis out of the window as it wouldn't support that many
levels.

So, back to the suspicion of the program cheating in the harder levels.
Is there any way to investigate this other than looking for anomalies in
the dice throw statistics?

[Dave Yeo]

Post by Peter J. Seymour

Post by Martin Iturbide
This is a interesting topic.
To bad I don't play Backgammon to try it out.
Which game is? Is this one the one you think is cheating?
http://www.os2world.com/games/index.php/native-games/board/114-os-2-td-gammon
Regards
Martín.

I have been on an OS/2 system since my previous post (well, ECS 1.2R to
be precise). That has TD-Gammon 2.0 installed.
I see that the game has six levels of hardness. That puts my variable
faced dice hypothesis out of the window as it wouldn't support that many
levels.
So, back to the suspicion of the program cheating in the harder levels.
Is there any way to investigate this other than looking for anomalies in
the dice throw statistics?

The problem is that the dice seem random until a situation where rolling
one combination can turn the game around. I just played a quick game, I
was winning with most of my end blocked, got stuck having to leave 2 men
open and bam, the computer rolls a 5+2 and lands on both my men.
It may just be confirmation bias but the computer seems to be able to
consistently roll the right numbers when in a tight spot, and not just
rolling eg a seven which is pretty likely but a combination that does
maximum damage.
I can win over 50% of the games by playing very conservatively compared
to the computer but it seems that leaving a man open means getting it hit.
Dave