TD-GAMMON with OS/2 WARP 4
Lex Weaver
Can someone please tell me the version number of the TD-GAMMON package
distributed in the Family Fun Pack as part of OS/2 Warp 4?
I have OS/2 v2.1 with TD-GAMMON v1.0, and am wondering if the latest
release is any newer or better.
cheers,
Lex
----------------------------------------------------------------------------
Lex Weaver
l...@cs.anu.edu.au http://cs.anu.edu.au/~Lex.Weaver
Ph: +61 6 279 8180 (w) Add: c/- Dept of Computer Science
Australian National University
Australia 0200
============================================================================
"Often the difference between a successful man and a failure
is not one's better abilities or ideas, but the courage that
one has to bet on his ideas, to take a calculated risk - and
to act."
Maxwell Maltz
Gerry Tesauro
In article <5d3rr4$a...@flash.anu.edu.au>, Lex Weaver <l...@cs.anu.edu.au> wrote:
>Can someone please tell me the version number of the TD-GAMMON package
>distributed in the Family Fun Pack as part of OS/2 Warp 4?
>
>I have OS/2 v2.1 with TD-GAMMON v1.0, and am wondering if the latest
>release is any newer or better.
>
This was posted a few months ago:
------------------------------------------------------
Article: 16084 of rec.games.backgammon
From: tes...@ferrari.watson.ibm.com (Gerry Tesauro)
Newsgroups: rec.games.backgammon
Subject: OS/2 TD-Gammon 2.0 now available
Date: 1 Oct 1996 19:50:39 GMT
Organization: IBM T. J. Watson Research
Message-ID: <52rsmf$m...@watnews1.watson.ibm.com>
I'm happy to announce that version 2.0 of OS/2 TD-Gammon
is now shipping along with the new release of OS/2 Warp 4,
which was just announced last week. TD-Gammon is one
of over 70 applications on the Applications Sampler CD-ROM
that's included for free along with the new version of OS/2.
The new version of OS/2 TD-Gammon contains many improvements
over last year's version 1.0, including:
-- A new "Grandmaster" skill level based on 3-ply search
-- Extensively optimized code runs approximately 3-4 times
faster than version 1.0
-- New options allow the user to ask TD-Gammon to recommend
a move, and to get TD-Gammon's estimated equity of
a position
-- New "Manual Dice" option allows the user to input dice
rolls manually
-- New "Network Play" option allows two users to play
against each other, for computers connected to a common
Local Area Network (LAN).
To place an order or find out more about OS/2 Warp 4,
call 1-800-CALL-IBM, or visit the IBM Software home page
at: http://www.software.ibm.com
John Littler
Version 2.0, no idea if it's better than v1.0. It's voice enabled (may
be the only difference) it sure plays a mean game of 'gammon though !!
John
Lex Weaver wrote:
>
> Can someone please tell me the version number of the TD-GAMMON package
> distributed in the Family Fun Pack as part of OS/2 Warp 4?
>
> I have OS/2 v2.1 with TD-GAMMON v1.0, and am wondering if the latest
> release is any newer or better.
>
Peter Pollack
John Littler wrote:
>
> Version 2.0, no idea if it's better than v1.0. It's voice enabled (may
> be the only difference) it sure plays a mean game of 'gammon though !!
> John
>
> Lex Weaver wrote:
> >
> > Can someone please tell me the version number of the TD-GAMMON package
> > distributed in the Family Fun Pack as part of OS/2 Warp 4?
> >
> > I have OS/2 v2.1 with TD-GAMMON v1.0, and am wondering if the latest
> > release is any newer or better.
Okay, I'll bite. Where do we find version 2.0?
Peter Pollack
--
Johannes Gramsch
It's on the application sampler CD that comes with Warp 4
Ciao
Johannes
David D'Antonio
In article <01bc182f$5ab0d880$132ee926@dlane>, DL...@ContactPt.com says...
>
>Is a version that will work with Warp 3 available?
>
>Anybody know?
The version that one could get for fee from IBM's ftp server works fine
under Warp 3 FP17. Actually, that's no quite true as the colour of the
pieces got wierd with the ATI MACH64 drivers I was using.
Oh and there was alot of net speculation that the damn thing cheats! I'm
not a backgammon slouch but it could trash me consistently. And it always
seemed to get the die rolls it needed to cover an otherwise vulnerable
position...
>David.
DDA
--
David D'Antonio, CNE (dant...@process.com)
Some they do and some they don't and some ya just can't tell
Some they will and some they won't and some it's just as well.
-SuperTramp
David Zuidema
On 11 Feb 97 18:44:51 -0500, dant...@process.com (David D'Antonio)
wrote:
>In article <01bc182f$5ab0d880$132ee926@dlane>, DL...@ContactPt.com says...
>>
>>Is a version that will work with Warp 3 available?
>>
>>Anybody know?
>
>The version that one could get for fee from IBM's ftp server works fine
>under Warp 3 FP17. Actually, that's no quite true as the colour of the
>pieces got wierd with the ATI MACH64 drivers I was using.
>
>Oh and there was alot of net speculation that the damn thing cheats! I'm
>not a backgammon slouch but it could trash me consistently. And it always
>seemed to get the die rolls it needed to cover an otherwise vulnerable
>position...
I'm glad that someone else thinks the die rolls were too good to be
true ! I thought I was going paranoid. :-)
________________________________________
David Zuidema
http://www.physics.monash.edu.au/~davidz
Stephen Turner
David Zuidema wrote:
>
> >Oh and there was alot of net speculation that the damn thing cheats! I'm
> >not a backgammon slouch but it could trash me consistently. And it always
> >seemed to get the die rolls it needed to cover an otherwise vulnerable
> >position...
>
> I'm glad that someone else thinks the die rolls were too good to be
> true ! I thought I was going paranoid. :-)
>
Good players get more good rolls than bad players. What's new?
--
Stephen Turner sr...@cam.ac.uk http://www.statslab.cam.ac.uk/~sret1/
Stochastic Networks Group, Statistical Laboratory,
16 Mill Lane, Cambridge, CB2 1SB, England Tel.: +44 1223 337955
"Collection of rent is subject to Compulsive Competitive Tendering" Cam. City
Jim Babka
Actually, I have a theory about this. I'm not "in the know" about this - I
don't know the people who put TD-GAMMON together - but I have noticed the
rather blatant appearance of cheating (program has three unprotected pieces,
only way I *won't* hit it at least once is if I roll a 1 2, and guess what I
roll). My theory is that this is a game that was trained via massive
reiteration of playing against itself. During the course of that learning,
it also "learned" the pattern to the pseudo-random numbers used to generate
the die rolls. I understand that they probably change the seed from game to
game, but depending upon the quality of the random-number generator, there can
still be a repeating pattern, and that pattern might be built into the neural
net (or whatever other algorithm is being used) that drives the computer play.
In article <33014410...@NewsServer.cc.monash.edu.au>,
David Zuidema <David....@sci.monash.edu.au> wrote:
>On 11 Feb 97 18:44:51 -0500, dant...@process.com (David D'Antonio)
>wrote:
>
>>In article <01bc182f$5ab0d880$132ee926@dlane>, DL...@ContactPt.com says...
>>>
>>>Is a version that will work with Warp 3 available?
>>>
>>>Anybody know?
>>
>>The version that one could get for fee from IBM's ftp server works fine
>>under Warp 3 FP17. Actually, that's no quite true as the colour of the
>>pieces got wierd with the ATI MACH64 drivers I was using.
>>
>>Oh and there was alot of net speculation that the damn thing cheats! I'm
>>not a backgammon slouch but it could trash me consistently. And it always
>>seemed to get the die rolls it needed to cover an otherwise vulnerable
>>position...
>
>I'm glad that someone else thinks the die rolls were too good to be
>true ! I thought I was going paranoid. :-)
>
>________________________________________
>David Zuidema
>http://www.physics.monash.edu.au/~davidz
>
--
Jim Babka ba...@austin.ibm.com (Internet), BABKA at AUSTIN (VM)
AIX Kernel Service and Support USIB6UKS at IBMMAIL
What my employer thinks about what I say is best left unprinted...
If we could just get everyone to close their eyes and visualize
John Clements
In article <5dsm20$s...@ausnews.austin.ibm.com>,
Jim Babka <ba...@austin.ibm.com> wrote:
>Actually, I have a theory about this. I'm not "in the know" about this - I
>don't know the people who put TD-GAMMON together - but I have noticed the
>rather blatant appearance of cheating (program has three unprotected pieces,
>only way I *won't* hit it at least once is if I roll a 1 2, and guess what I
>roll). My theory is that this is a game that was trained via massive
>reiteration of playing against itself. During the course of that learning,
>it also "learned" the pattern to the pseudo-random numbers used to generate
>the die rolls. I understand that they probably change the seed from game to
>game, but depending upon the quality of the random-number generator, there can
>still be a repeating pattern, and that pattern might be built into the neural
>net (or whatever other algorithm is being used) that drives the computer play.
>
a) you're right about the technique used for building TD-gammon (last time I
heard, it was a self-trained neural network.
b) you're almost certainly wrong in your suggestion that TD-gammon acts based
upon learned knowledge about the way the dice perform.
What you're proposing (as I understand) it is that TD-gammon gives high
equity ratings to positions which, though bad, are made good by a persistent
bias of the dice, and thus are 'safe'. For this sort of persistent bias to
exist, it would appear to be necessary for the dice to be linked in some way to
the positions being seen, a near-impossibility. Given that TD's evaluation of
positions is based upon a 'generalization', if you will, of a plethora of
somewhat similar situations, I rather suspect that you could get away with
using a decidedly _non_ random die sequence for training, and that you would
still end up with a fairly valid evaluation engine, i.e. one which is not
biased towards certain risky situations.
As further evidence that no such bias exists, witness the fine performance
of neural net-based 'bots like td-one, jellyfish, and mloner on FIBS, where the
die-rolling algorithm is almost certainly not the one that these 'bots were
trained with.
How about this: I'll give you ten to one that you're mistaken.
john
Nick Cassimatis
David Zuidema wrote:
> I'm glad that someone else thinks the die rolls were too good to be
> true ! I thought I was going paranoid. :-)
>
> ________________________________________
> David Zuidema
> http://www.physics.monash.edu.au/~davidz
If it's programmed to cheat, maybe it's programmed to know when it's
been caught!
--
Nick Cassimatis
My opinions are mine alone.
Gareth McCaughan
Jim Babka wrote:
> Actually, I have a theory about this. I'm not "in the know" about this - I
> don't know the people who put TD-GAMMON together - but I have noticed the
> rather blatant appearance of cheating (program has three unprotected pieces,
> only way I *won't* hit it at least once is if I roll a 1 2, and guess what I
> roll). My theory is that this is a game that was trained via massive
> reiteration of playing against itself. During the course of that learning,
> it also "learned" the pattern to the pseudo-random numbers used to generate
> the die rolls. I understand that they probably change the seed from game to
> game, but depending upon the quality of the random-number generator, there
> can still be a repeating pattern, and that pattern might be built into the
> neural net (or whatever other algorithm is being used) that drives the
> computer play.
There are only three big problems with this theory.
1. The "patterns" in any halfway-decent pseudo-random number generator
are *way* too subtle to be spotted by a neural network. They just
aren't the sort of thing you can detect that way.
2. TD-Gammon wasn't trained on a PC running OS-2. It was trained on an
RS/6000 running (I think) AIX. It's unlikely (though possible) that
these two systems use the same random number generator.
3. What a neural network like TD-Gammon learns is an estimate of the
value of the position (I am over-simplifying here). It doesn't get
told (e.g.) the move number, or the current state of the RNG, or
anything like that.
In other words: your theory is interesting, but it is a million miles
away from being possible.
--
Gareth McCaughan Dept. of Pure Mathematics & Mathematical Statistics,
gj...@dpmms.cam.ac.uk Cambridge University, England.
ChuckCook@warp
In <5dsm20$s...@ausnews.austin.ibm.com>, ba...@austin.ibm.com (Jim Babka) writes:
>Actually, I have a theory about this. I'm not "in the know" about this - I
>don't know the people who put TD-GAMMON together - but I have noticed the
>rather blatant appearance of cheating (program has three unprotected pieces,
>only way I *won't* hit it at least once is if I roll a 1 2, and guess what I
>roll). My theory is that this is a game that was trained via massive
>reiteration of playing against itself. During the course of that learning,
>it also "learned" the pattern to the pseudo-random numbers used to generate
>the die rolls. I understand that they probably change the seed from game to
>game, but depending upon the quality of the random-number generator, there can
>still be a repeating pattern, and that pattern might be built into the neural
>net (or whatever other algorithm is being used) that drives the computer play.
>
>David Zuidema <David....@sci.monash.edu.au> wrote:
I think you may be on to something there. I have noticed that the computer's luck
with the die seems to be directlly related to the skill setting. Possibly the higher
skill levels used more of the learning you describe.
Possibly the skill levels relate to ones ability to prevail over an increasingly
dis-honest opponent?
>>On 11 Feb 97 18:44:51 -0500, dant...@process.com (David D'Antonio)
>>wrote:
>>
>>>In article <01bc182f$5ab0d880$132ee926@dlane>, DL...@ContactPt.com says...
>>>>
>>>>Is a version that will work with Warp 3 available?
>>>>
>>>>Anybody know?
>>>
>>>The version that one could get for fee from IBM's ftp server works fine
>>>under Warp 3 FP17. Actually, that's no quite true as the colour of the
>>>pieces got wierd with the ATI MACH64 drivers I was using.
>>>
>>>Oh and there was alot of net speculation that the damn thing cheats! I'm
>>>not a backgammon slouch but it could trash me consistently. And it always
>>>seemed to get the die rolls it needed to cover an otherwise vulnerable
>>>position...
>>
>>I'm glad that someone else thinks the die rolls were too good to be
>>true ! I thought I was going paranoid. :-)
>>
>>________________________________________
>>David Zuidema
>>http://www.physics.monash.edu.au/~davidz
>>
>
>
>--
>Jim Babka ba...@austin.ibm.com (Internet), BABKA at AUSTIN (VM)
>AIX Kernel Service and Support USIB6UKS at IBMMAIL
>What my employer thinks about what I say is best left unprinted...
> If we could just get everyone to close their eyes and visualize
//=======================
// Chuck Cook
// co...@clark.net
// www.clark.net/pub/cookc
//=======================
Bob Selfinger
In message <33019D...@cam.ac.uk> - Stephen Turner <sr...@cam.ac.uk>
writes:
:>
:>David Zuidema wrote:
:>>
:>> >Oh and there was alot of net speculation that the damn thing cheats! I'm
:>> >not a backgammon slouch but it could trash me consistently. And it always
:>> >seemed to get the die rolls it needed to cover an otherwise vulnerable
:>> >position...
:>>
:>> I'm glad that someone else thinks the die rolls were too good to be
:>> true ! I thought I was going paranoid. :-)
I have been suspecting that it cheats very badly at times but not always.
Maybe it looks at the date or some other thing to decide when it will
cheat. Some days I can beat it every time, but on some days it will roll
exactly what it needs again and again and it is just about unbeatable in
this mode.
:>
:>Good players get more good rolls than bad players. What's new?
:>
:>--
:>Stephen Turner sr...@cam.ac.uk http://www.statslab.cam.ac.uk/~sret1/
:> Stochastic Networks Group, Statistical Laboratory,
:> 16 Mill Lane, Cambridge, CB2 1SB, England Tel.: +44 1223 337955
:> "Collection of rent is subject to Compulsive Competitive Tendering" Cam. City
Bob Selfinger
******************************************************
Views expressed here may not actually be views.
******************************************************
Stephen Turner
John Clements wrote:
>
>
> How about this: I'll give you ten to one that you're mistaken.
>
Only ten? I'll give you 100.
Scott Allen
** Reply to note from dant...@process.com (David D'Antonio) 11 Feb 97 18:44:51 -0500
> Oh and there was alot of net speculation that the damn thing cheats! I'm
> not a backgammon slouch but it could trash me consistently. And it always
> seemed to get the die rolls it needed to cover an otherwise vulnerable
> position...
I've played it this way many times using real dice. It "cheats" just as
much when *I* roll the dice as when it's internal generator is used! ;-)
** Scott Allen (warp on FIBS) Scarborough, Ontario, Canada [Team OS/2] **
*** Internet: sco...@idirect.com Fido: (1:250/636) RelayNet: ->1347 ***
Hugh B. McNeil
Said it before, and we can say it again:
"Good players make their own luck."
doesn't mean they are cheating, just that they are maximising
their chances of getting a good roll. Even if it means only
one roll in thirty six.
And as for missing a triple shot, haven't you ever fanned
against a three point board? Often? Same thing right?
As for TD Gammon predicting the outcome of the dice because
it has "learned" the algorithm... interesting. This suggests
that people can do the same thing (subconsciously) by observing
how the dice are picked up, placed in the cup, shaken, how the
dice were facing when they were picked up, etc...
BTW, the corollary of the above maxim is:
"Bad players make their own luck."
ie they are minimising their number of good rolls, and
maximising their opponent's number of good rolls. Bad players
seem unlucky because they maximise their opportunity to be
unlucky.
IMHO
John Clements
In article <5dsvjv$u...@rtpnews.raleigh.ibm.com>,
Nick Cassimatis <NickC@!ibm.net> wrote:
>
>Fight back! Put it on Manual Dice Entry whenever the computer cheats!
>
Hold on... if you get to re-roll the dice whenever the computer gets a really
really good roll, shouldn't the _computer_ get to re-roll the dice whenever
_you_ get a really good roll?
(BTW, think motive... what possible motive could there be for a programmer
to write a backgammon program that cheats?)
john
Gerry Tesauro
In article <1997Feb11....@delta.process.com>,
David D'Antonio <dant...@process.com> wrote:
(snip)
>
>Oh and there was alot of net speculation that the damn thing cheats!
The following is a copy of some recent correspondence
I've had on this subject. (As an anti-spam measure,
e-mail addresses have been deleted. If you want to
send me e-mail, delete "ferrari." from the address in
the header above.)
=========================================================================
Date: 28 January 1997, 18:23:40 EST
From: Gerry Tesauro
To: Douglas Metcalf
Re: TD-Gammon
Ref: Your note of 28 JAN 97 07:10:39 CST
Thanks for your comments. If you can manage to get a copy
of OS/2 Warp 4, it includes an Applications Sampler CD
which contains the latest version of TD-Gammon, which
has among other features an option for Manual Dice Entry.
It's been pretty interesting to see people's feedback
about the dice rolls-- people are suspicious not only of
TD-Gammon, but basically all programs that are good enough
to wins lots of games. Apart from going into a dissertation
on statistics, what I can say is the following:
(1) I'll give you my personal word of honor that the dice
rolls are honest. The dice generator has no idea whose turn
it is or what the state of the board is when it rolls.
(2) There's an elegant proof that the dice rolls can't
depend on the state of the game, based on the fact that
the random number generator is seeded by the time of day.
What this means is that, if you start two copies of the
program at the same time, you will get the exact same
sequence of dice rolls on both copies. This will hold even
if you are playing on different levels and making completely
different moves, leading to radically different game states.
Best wishes,
-- Gerry Tesauro
=========================================================================
Date: 29 JAN 97 07:52:06 CST
Subject: re: TD-Gammon
To: Gerry Tesauro
MIME-Version: 1.0
X-Mailer: 2.3.5 ZOOMIT X.400/SMTP Dual Stack
X-Complete-Subject: re: TD-Gammon
Message-ID: <0007oqkuicww...@igate.sprint.com>
Thank you for the reply. I have Warp 4, and have now installed the newer
TD-Gammon.
I can hardly wait to get a pair of dice and see how Backgammon is supposed
to be played.
I'll bet your responses have been interesting. Although I have only played
one game (in Grandmaster mode) with the new program, I would like to relate
the following incident to you, although after reading your e-mail, I'm
quite sure it is just a coincidence/bad luck.
I was ready to get my men home, with only one roll that could cause me
trouble (6,2) and the only thing that would cause me trouble after my
rolling 6,2 was the program rolling any 2. Guess what. I rolled 6,2 and
the program followed with 2,2. I think the odds of that happening are
almost 60-1, certainly not in the realm of the impossible, but even after
only one game, it is a little suspicious. I sure hope that's just my bad
luck. But I will find out for sure when I input the rolls myself, and have
played more games to see if my luck is consistently bad.
Oh, and by the way, the changes to the program, ie, resignation, manual
dice, etc, make this one of the finest (backgammon) games available. (And
it only works on my favorite operating system - too bad for those MS
folks!) Thanks for the fine work, and if I'm convinced my luck is no good
after several weeks, I am going to attempt to document my 'bad luck' with
some statistical proof, which I will forward to you.
Thanks again,
Doug Metcalf
=========================================================================
Date: 30 JAN 97 06:53:37 CST
Subject: re: TD-Gammon
To: Gerry Tesauro
MIME-Version: 1.0
X-Mailer: 2.3.5 ZOOMIT X.400/SMTP Dual Stack
X-Complete-Subject: re: TD-Gammon
Message-ID: <0007fzbbzphz...@igate.sprint.com>
Mr. Tesauro,
My apologies. TD-Gammon wins even if it lets me roll the dice! The manual
entry convinced me! Of course, now I'm wondering why I play so badly. Which,
of course, makes me wonder if perhaps someday, TD-Gammon might tell me when
I'm making an obvious mistake... Users, huh? Never happy...
Oh, well. At least I can brag to my "Windows" friends that at least I have a
real Backgammon opponent whenever I want.
Thanks, and once again, my sincere apologies sorry for doubting your
program's integrity.
Doug Metcalf
Gerry Tesauro
>don't know the people who put TD-GAMMON together - but I have noticed the
>rather blatant appearance of cheating (program has three unprotected pieces,
>only way I *won't* hit it at least once is if I roll a 1 2, and guess what I
>roll). My theory is that this is a game that was trained via massive
>reiteration of playing against itself. During the course of that learning,
>it also "learned" the pattern to the pseudo-random numbers used to generate
>the die rolls. I understand that they probably change the seed from game to
>game, but depending upon the quality of the random-number generator, there can
>still be a repeating pattern, and that pattern might be built into the neural
>net (or whatever other algorithm is being used) that drives the computer play.
>
This is an interesting idea that has come up before.
It's hard to imagine how this could happen, since
TD-Gammon doesn't "see" the dice rolls as part of
its input -- it only sees transitions from one board
state to the next. Nevertheless, to be thorough,
I did do the experiment of changing the dice generator,
and found that and TD-Gammon plays just as well against a
benchmark opponent as when the training dice generator
is used. This shows that TD isn't getting an edge by
adapting to the statistics of the dice generator used
in training.
-- Gerry Tesauro
Lex Weaver
In article 3...@news1.ee.net, self...@ee.net (Bob Selfinger) writes:
>
> I have been suspecting that it cheats very badly at times but not always.
> Maybe it looks at the date or some other thing to decide when it will
> cheat. Some days I can beat it every time, but on some days it will roll
> exactly what it needs again and again and it is just about unbeatable in
> this mode.
If there are days when you nearly always beat TD_gammon, and others
when it nearly always beats you, perhaps this is just a natural
variation in the die?
Perhaps someone, or better several people, could play a number of games
against TD-Gammon using firstly the automated die, and secondly rolling
real die themselves (v2.0 allows you to enter the die manually), and
report the results (scores and level at which TD-Gammon was playing)
here.
geo...@ms5.hinet.net
In <5dtqgv$3...@news1.ee.net>, self...@ee.net (Bob Selfinger) writes:
>In message <33019D...@cam.ac.uk> - Stephen Turner <sr...@cam.ac.uk>
>writes:
>:>
>:>David Zuidema wrote:
>:>>
>:>> >not a backgammon slouch but it could trash me consistently. And it always
>:>> >seemed to get the die rolls it needed to cover an otherwise vulnerable
>:>> >position...
>:>> I'm glad that someone else thinks the die rolls were too good to be
>:>> true ! I thought I was going paranoid. :-)
>:>>
>
>Maybe it looks at the date or some other thing to decide when it will
>cheat. Some days I can beat it every time, but on some days it will roll
>exactly what it needs again and again and it is just about unbeatable in
>this mode.
>
>:>Good players get more good rolls than bad players. What's new?
>:>
>:> Stochastic Networks Group, Statistical Laboratory,
>:> 16 Mill Lane, Cambridge, CB2 1SB, England Tel.: +44 1223 337955
>:> "Collection of rent is subject to Compulsive Competitive Tendering" Cam. City
>
bal...@ludens.elte.hu
Does anybody know if there is an official backgammon association in Hungary,
and how they may be contacted? All informations are welcome.
Boldizsar Fejervari (bal...@ludens.elte.hu)
ChuckCook@warp
In <01bc182f$5ab0d880$132ee926@dlane>, "David Lane" <DL...@ContactPt.com> writes:
>Is a version that will work with Warp 3 available?
>
>Anybody know?
>
>
>
Yes, It is avaliable as a free download from
http://www.software.ibm.com
Warning: It is a serious player!
bal...@ludens.elte.hu
gow...@ibm.net
So I cheat right back using manual dice!
Rob Hutchens
Daniel Murphy
gow...@ibm.net writes:
>You're right the damn thing cheats! I'm a fairly competent backgammon
>player and the system has the most unbelievable luck! So I cheat right
>back using manual dice!
TD-Gammon's creator, Gerry Tesauro, has affirmed publicly, and more than
once, that the TD-Gammon program does not cheat. That's good enough for
me.
Considering that Gerry's program plays backgammon on an expert level, it's
not surprising that a merely "fairly competent" player would have to cheat
to beat it.
No one I've ever known has been able to *examine* his or her game and
become a good player without first learning to blame poor play, not bad
luck, for their losses. Some never learn this. That's ok. I happen to
like fish.
--
_______________________
Daniel Murphy
San Francisco
rac...@cityraccoon.com
David D'Antonio
In article <raccoon.856599064@shellx>, rac...@best.com says...
>
>gow...@ibm.net writes:
>
>>You're right the damn thing cheats! I'm a fairly competent backgammon
>>player and the system has the most unbelievable luck! So I cheat right
>>back using manual dice!
>
>TD-Gammon's creator, Gerry Tesauro, has affirmed publicly, and more than
>once, that the TD-Gammon program does not cheat. That's good enough for
>me.
I'm sure it doesn't "cheat" (or at least I'm sure Gerry didn't write it
to cheat), but it gets the most amazing luck to pull it out of weak positions.
And I never seem to be able to get that level of luck going for me.
>Considering that Gerry's program plays backgammon on an expert level, it's
>not surprising that a merely "fairly competent" player would have to cheat
>to beat it.
No, it doesn't ALWAYS play at expert level (at least the version I have).
But it still gets lucky. :-)
>No one I've ever known has been able to *examine* his or her game and
>become a good player without first learning to blame poor play, not bad
>luck, for their losses. Some never learn this. That's ok. I happen to
>like fish.
A good backgammon player can still be defeated by the dice. It just seems
that TD-GAMMON always gets the dice on its side. And a good backgammon
player will NEVER lose if they have the die on their side!
>Daniel Murphy
John E. Abraham
David D'Antonio wrote:
>
> In article <raccoon.856599064@shellx>, rac...@best.com says...
> >
> >gow...@ibm.net writes:
> >
> >>You're right the damn thing cheats! I'm a fairly competent backgammon
> >>player and the system has the most unbelievable luck! So I cheat right
> >>back using manual dice!
> >
> >TD-Gammon's creator, Gerry Tesauro, has affirmed publicly, and more than
> >once, that the TD-Gammon program does not cheat. That's good enough for
> >me.
>
> I'm sure it doesn't "cheat" (or at least I'm sure Gerry didn't write it
> to cheat), but it gets the most amazing luck to pull it out of weak positions.
> And I never seem to be able to get that level of luck going for me.
I bought a real (actual, physical) backgammon board to see if I'd have
better luck against a real (actual, physical) person. The actual person
cheats just as much as TD-GAMMON does! There's no way anyone can throw
that many double sixes!
--
John Abraham
Urban Modelling and Transportation Planning
j!abr...@acs.ucalgary.ca
(the "!" isn't part of my email address - be sure to remove it before
replying.
it's just there to keep the spammers at bay.)
ChuckCook@warp
I thought I was a good Gammon player, so when TD-Gammon started consistanly
beating me ( real bad too! ). I of course concluded that it was cheating. I accepted
that part of the challange was beating a dishonest opponent.
However after playing at the expert level for a while I have changed my mind.
It is just a very good player!
Watch the moves it makes at the various board positions. I changed my style of play
to match the machine and now I win about half the time.
Chuck