gnubg's dice
Pete
and beat it. In the post-match analysis, in the Luck rating, it gave me the
"Good dice, man" message. This, form a program that was behind by 31 pips in
a race (67 to 98 total pips), and four rolls later I can't take the
redouble. It rolled 66, 62, 33, and 66.
This is not the first time I have seen this program get more than its normal
share of doubles in a race, and other jokers also. Anybody else think there
is something fishy here? And what exactly is the criteria for "Rolls marked
very lucky", and "Rolls marked lucky"?
Just musing outloud.
Thanks,
Pete
Murat Kalinyaprak
Pete wrote ut6j7ol...@corp.supernews.com
> .... Anybody else think there is something fishy here?
> And what exactly is the criteria for "Rolls marked very
> lucky", and "Rolls marked lucky"? Just musing outloud.
I don't think there is anything fishy here; I think there is
something "jellyfishy" here... :))
Well, Pete, welcome to the land of the rigged servers,
cheating bots, scum sucking sick bastards, shit eating
fat faggots, their ass kissers, $19.95 cube skill books,
tourists from planet Moro, etc... :)
They complain about what we write here but we never
complain about the sick gamblers anonymous people's
going on and on about subjects worthless and irrelevant
to bg like what happens when a player uses the cube in
a crawford game, etc. etc...
Unfortunately, playing against cheating bots and rigged
shit-pits (itself!) or even writing here could amuse a guy
only so far... After that, there has been times I wished I
were wealthy enough to set up my own display-booth at
one of those world-class competitions and shove it all up
their dumb fucking asses in front of their own cameras...
But then again, I never paid for a whore to listen to me...
And not because I wouldn't pay them but because they
wouldn't listen to me... Apparently whores get paid to
fuck you and not listen to you... :((
Futhermore, what's the use of trying even for the ideals
since even the Greek, Chinese, etc. philosophers were
never able to help them herds of dumb fuckers...?? :((
MK
-----= Posted via Newsfeed.Com, Uncensored Usenet News =-----
http://www.newsfeed.com - The #1 Newsgroup Service in the World!
-----== 100,000 Groups! - 19 Servers! - Unlimited Download! =-----
Albert Silver
It's not fishy, though if you don't know how it calculates luck, your
comments are more than understandable. Luck isn't based on a single
roll or two that are marked, but on ALL rolls. The ones that are
marked (see in the Settings menu and then Analysis for the thresholds)
are merely the ones where there was the biggest equity swing cause by
the roll. The other rolls which gave a certain amount of luck or lack
of, will merely be added up. The luck is calculated according to the
quantity of equity the roll forcibly wins or loses. In the end it is
all tallied and you get your luck rating. Thus you could have 10 rolls
which gained 0.200 equity but are not earmarked as the default
threshold is 0.300 to be highlighted, and I could get 3 rolls marked
Very Lucky at a 0.600 gain, and you would still be judged the luckier
of the two.
Albert
Back4U2 BBL
normal
> share of doubles in a race, and other jokers also. Anybody else think
there
> is something fishy here?
Nothing 'fishy'.
You can change the Random Number Generator if you like
( Settings / Options. Look at the Game&Dice column)
or just set it to Manual Dice and use your own set of (precision) dice to
roll.
Nardy
JohnK
>
> This is not the first time I have seen this program get more than its
normal
> share of doubles in a race, and other jokers also. Anybody else think
there
> is something fishy here?
Lucky rolls happen.
I'm a long time palm owner and there are a number of BG programs for that
O/S that are of quite limited strength. (Slapgammon is the best of the lot,
by the way) ALL but the worst of these have been accused of cheating. The
ones that don't have any complaints play at about a 1200 level, IHMO. I
assure you that if you did as well against GnuBG as you did you could easily
wipe the floor with most of what is available for the Palm OS. I invite you
to go look at the user reviews of these programs at www.palmgear.com
The complaints are always the same. It gets too many doubles; the only time
I get them are when I'm on the bar; it always gets exactly what it needs, ad
infinitium, ad nauseum. My point is why would anyone write a program that
cheats? NO one has EVER shown ANY verifiable proof.
Furthermore, this applies all the more so for GnuBG because it is part of
the Gnu Project which is a collaborative effort and is open source. So um...
is the cheating code buried so deep that no one else can ever find it? Are
perhaps the cheat portions of the code remarked so that if you find it you
can send away for a hush money reward? What motivation would a group of
Unix-based programming aficionados have to write a program that cheats? A
group who have little or no profit motive, who more or less believe software
should be free to all, who publish their source code, etc, etc.
(By the way, don't feed the troll.)
Pete
bother quantifying one side's good luck or extremely good luck, since it
also can be expressed as the other side's bad luck or extremely bad luck.
IOW, why not just have a single set of numbers to reflect that a single roll
or a series of rolls by one side has caused an inordinate amount of equity
shifting.
Thanks,
Pete
PS And BTW, I don't thnk gnubg noted that its four rolls of 66, 62, 33, and
66 were extremely lucky, or extremely unlucky for me. I'll double-check if I
can.
Gregg Cattanach
> Your explanation is enlightening. But since this is a zero-sum game, why
> bother quantifying one side's good luck or extremely good luck, since it
> also can be expressed as the other side's bad luck or extremely bad luck.
>
> IOW, why not just have a single set of numbers to reflect that a single
roll
> or a series of rolls by one side has caused an inordinate amount of equity
> shifting.
>
> Thanks,
>
> Pete
The total luck factor for a match in Snowie is expressed as a single number,
and is reported as a positive number for the 'lucky' player and the negative
of that number for the other. (i.e. if my luck factor was 10.123,
opponent's will be -10.123.) GnuBG splits it out at one more level, adding
up all the positive or negative luck every time I rolled, and a separate
number for every time the opponent rolled. If both numbers are positive,
then both players rolled better than average, with the bigger number for the
more lucky player. If both numbers are negative, then both players rolled
worse than average, with the smaller number for the less lucky player.
However, if you add these two number together, you'll get the same idea as
the single number Snowie reports.
Gregg C.
Pete
single unified "luck" factor?
"Gregg Cattanach" <gcattana...@prodigy.net> wrote in message
news:dA6B9.722$HZ7.20...@newssvr17.news.prodigy.com...
Gregg Cattanach
> OK, thanks for the clarification, but does gnubg also have the concept of
a
> single unified "luck" factor?
>
Gregg Opponent
Luck rate (total) -2.973 (-16.022%) +1.752 (+17.970%)
Luck rate (pr. move) -0.020 (-0.109%) +0.012 (+0.121%)
The luck rate per move is similar to Snowie's value, (Snowie multiplies it
by 1000 just to make it easier to read). So for this match, I had -0.020 on
my rolls, Opponent had +0.012 for his rolls, so the total luck for the match
would be -0.032 for me, (or about -32.000 in Snowie's terms).
Gregg C.
Pete
> O/S that are of quite limited strength. (Slapgammon is the best of the
lot,
> by the way) ALL but the worst of these have been accused of cheating. The
> ones that don't have any complaints play at about a 1200 level, IHMO. I
> assure you that if you did as well against GnuBG as you did you could
easily
> wipe the floor with most of what is available for the Palm OS. I invite
you
> to go look at the user reviews of these programs at www.palmgear.com
>
but in any case, they bored me to tears because of their playing capability.
> The complaints are always the same. It gets too many doubles; the only
time
> I get them are when I'm on the bar; it always gets exactly what it needs,
ad
> infinitium, ad nauseum. My point is why would anyone write a program that
> cheats? NO one has EVER shown ANY verifiable proof.
You misunderstood my complaint. It wasn't that I am accusing the gnubg
number generator of anything. It was that I was complaining about what I
perceived to be a very nice sequence of rolls in a non-contact position
where I am way ahead, and four rolls later, I can't take the redouble. I was
complaining about its analysis of the luck factor and its comment that I had
good rolls without acknowledging its fantastic reversal of fortune. I also
noted that this wasn't the first time I had noted gnu to get jokers in key
situations (bearing off, coming in from the bar with a hit and covering
blots left and right). Some of the complaints have been resolved by the
explanations of how luck is calculated over a sequence of rolls. Some of the
issues are probably related to perception - one tends to note the opponent's
good rolls more than their own.
> Furthermore, this applies all the more so for GnuBG because it is part of
> the Gnu Project which is a collaborative effort and is open source. So
um...
> is the cheating code buried so deep that no one else can ever find it? Are
> perhaps the cheat portions of the code remarked so that if you find it you
> can send away for a hush money reward? What motivation would a group of
> Unix-based programming aficionados have to write a program that cheats? A
> group who have little or no profit motive, who more or less believe
software
> should be free to all, who publish their source code, etc, etc.
>
I got nothing against the program (with the following caveats). It is the
best bg program I have seen, and it has pointed out numerous leaks in my
play; though I don't like its propensity to make the 1- and 2-points so
quickly. The way I learned the game you (almost) NEVER made the 1-point
unless you were absolutely forced to, and you tried like hell to avoid
making the 2-point either. I do not like UI insofar as the display of the
statistics and evaluations - I think that a user interface engineer could do
serious improvements in that area.
> (By the way, don't feed the troll.)
I have no idea what this means. Sorry.
Pete
JohnK
Okay, sorry about that, I guess I misread it entirely.
> I do not like UI insofar as the display of the
> statistics and evaluations - I think that a user interface engineer could
do
> serious improvements in that area.
Absolutely! This is somewhat common for the Gnu project. The primary reason
is that the Windows UI is a secondary effort. The people working on the
programming do so gratis and (I would guess) many of them don't even use
Windows.
>
> > (By the way, don't feed the troll.)
>
> I have no idea what this means. Sorry.
>
A Troll is a term used for someone who is typically abusive and inflamatory
in their posts and often do so for no other reason than to irritate people.
See further:
http://www.winternet.com/~mikelr/flame1.html
Murat Kalinyaprak
Back4U2 BBL wrote 3dd3b4fd$0$190$ba62...@news.skynet.be
>> Anybody else think there is something fishy here?
> Nothing 'fishy'. You can change the Random Number Generator
> if you like ( Settings / Options. Look at the Game&Dice column)
> or just set it to Manual Dice and use your own set of (precision)
> dice to roll.
OK, guys, I'm starting a new contest: "Can you be a dumber ass
than this guy"...??
Under gnudung's setting options we have the following selections:
ANSI
BSD
ISAAC
MERCENE TWISTER
WWW.RANDOM.ORG
USER
Now, if that list had said:
ANSI TWISTER
BSD TWISTER
MURAT'S TWISTER
MERCENE TWISTER
WWW.I-SWEAR-TO-GOD-RANDOM.ORG
ABUSER, etc...
What would have changed...??
I have been telling and I will keep telling dumb-fucking tourists
from planet Moro here to go take a few philosophy classes at
some community college before they try to debate with people
like me, in public forums like this one...
They pay hundreds of dollars for bg books, software, web site
membership, etc. but they won't pay $10 per unit to go take a
few classes which could improve their whole lives and not just
their bg skills but the dumb fuckers keep resisting... :( Sigh... :((
Murat Kalinyaprak
JohnK wrote 3dd42400$1...@corp-news.newsgroups.com
> The complaints are always the same. It gets too many
> doubles; the only time I get them are when I'm on the bar;
> it always gets exactly what it needs, ad infinitium, ad
> nauseum.
"Arguments" on this subject have gone beyond "complaints"
for a long long time by now...
Many people who aren't smashed by or even claim to beat
certain extra-terrestrial-scums also complain that they are
cheating...
Get a little educated on the issues before you open your ass...
> My point is why would anyone write a program that cheats?
Is this a genuine question or an assuming answer...?? Either
way, it shows that you are too stupid to find a generic answer
to such a question...
For example, why did the so-called "sniper" shoot and kill all
those innocent people...?
If you don't know the answer, you may just conclude that the
bastard was perhaps sick in the head or in the ass or both...
Now, try to apply the same to backgammon... And let me give
you a head start by letting you in on the secret that there are a
scientifically statistically proportionate number of sick bastards
here in the bg world as there are among the general population
out there in the universe...
> NO one has EVER shown ANY verifiable proof.
Just look at the source code...
> (By the way, don't feed the troll.)
I've been writing in this group for 6+ years and I am not a troll by
any stretch of definition or imagination...
However, I do understand some dumb Moronians' needs to call
people like me "trolls" in order to bury their heads in their asses...
I'm sure your head would fit fine but I would like to know how did
my arguments fit your ass...? If they didn't fit right, you can always
return them for a full refund... I guarantee 100% satisfaction... :))
Pete
matches total).
In the 70's I played bg in the clubs 4-6 hours a day almost every day.
That's a lot of money games. Let's put it this way. If I found myself
playing against a human that could roll perfectas as well as gnubg does, I
would have gotten up and ran out of there a long time ago. I am talking
about "doubles off the bar to bring in two men, cover a blot, and hit" type
of doubles; easily get out of a 5-pt prime; starting the game where every
roll makes a point (it never rolls a 5 unless it's a set of them); hardly
ever dancing more than once; I can name a few more if needed.
I know, I know. I hear the counter-arguments. 1). Don't play it if you don't
like it. 2). Hey, it's free. 3). Why would they mess with the dice - it's
not to their benefit.
I am not smart enough to know or to claim I know what's going on. I only
know that this thing's rolls are unconscious. Pretty much consistently.
Pete
Pete
Albert Silver
> I have been playing against gnubg for the past couple of weeks (maybe 50
> matches total).
>
> In the 70's I played bg in the clubs 4-6 hours a day almost every day.
> That's a lot of money games. Let's put it this way. If I found myself
> playing against a human that could roll perfectas as well as gnubg does, I
> would have gotten up and ran out of there a long time ago. I am talking
> about "doubles off the bar to bring in two men, cover a blot, and hit" type
> of doubles; easily get out of a 5-pt prime; starting the game where every
> roll makes a point (it never rolls a 5 unless it's a set of them); hardly
> ever dancing more than once; I can name a few more if needed.
>
> I know, I know. I hear the counter-arguments. 1). Don't play it if you don't
> like it. 2). Hey, it's free. 3). Why would they mess with the dice - it's
> not to their benefit.
The counter-arguments are a bit different actually, except maybe for
number 3. To my knowledge, every single top program, of which GNUBG is
a part, has been accused of manipulating the dice. Here is an
interesting link on the subject so you may see:
http://www.bkgm.com/rgb/rgb.cgi?menu+computerdice
There you will see a variety of old posts accusing every single
program under the sun (a slight exaggeration of course) and the
various replies. My favorites are the ones for TD-Gammon
(http://www.bkgm.com/rgb/rgb.cgi?view+10), the father of all
neural-net programs, which made programs such as Jellyfish, Snowie,
and of course GNUBG possible. In a nutshell, the programs are now
simply on a par of the very best players in the world. Really.
In this newsgroup you'll find lots of people commisserating with you
and agreeing it cheats, however I have yet to hear of a SINGLE top
player doing so. Basically, the bots (short for robots), as they are
referred to, have revolutionized backgammon strategical theory in the
last couple of decades, and continue to do so. You CAN, in order to
satisfy your curiousity, set the dice to manual BTW. Go to the
Settings menu, click on Options, and in the window, check Manual dice
under the heading Dice Generator.
More importantly though, GNUBG isn't just a bot to beat the pants off
of you. And me. And so many others. :-) It can help you learn and
improve. After your match is finished, have it analyze it by clicking
on the Analyze menu --> Analyze match (Analyze game will only analyze
a single game). When it is done, click on Match statistics. Do not be
disheartened if the grade it gives is far below your expectation.
Instead, click on the Window menu and select both Game Record and
Annotation. The Game Record displays the moves with the commentary. a
?? means a blunder was played, a ?! equals a dubious move, etc. The
green arrows indicate going to the next error. In the Annotation
window, you can see the moves, their grades (the Diff. indicates the
size of the error compared to the top move), and see what the
recommended move was.
I wrote a detailed tutorial which was published in Gammonline
(http://www.gammonline.com) last month. It will be released at the
GNUBG site in a few days for free. I would have done this earlier, but
my motherboard fried, and I don't have the files here on my laptop.
Albert Silver
Peter Schneider
> If I found myself
> playing against a human that could roll perfectas as well as gnubg does, I
> would have gotten up and ran out of there a long time ago.
[...]
> I know, I know. I hear the counter-arguments. 1). Don't play it if you
don't
> like it. 2). Hey, it's free. 3). Why would they mess with the dice - it's
> not to their benefit.
It's only the doubts with your own suspicion and your sensible language that
make me answer at all ;-).
I have 3 more counter arguments.
1. The first is a psychological one. I call it "The Red Traffic Light
Syndrom": You just forget the green ones so quickly. (I.e. analyze the
games, check the luck factor, and I bet you'll see that you have in fact
been luckier than you perceived, because it seems to lie in the human nature
to notice bad luck much clearer than good luck, which is easier attributed
to ones skills.)
2. The second is a technical one: gnubg is open source; go and check the
random number generators (you have a choice with gnubg), or write your own.
3. The third is a bg argument, and it's not new to you, I guess; I just want
to emphasize it: gnubg, as an excellent player, chooses the move which gets
(to its not-too-bad knowledge) the best results with all the possible rolls
for the next three moves (if playing 2-ply), weighted by their respective
probability. Even if some human players can theoretically do so in
sufficient exactness, they will probably rarely do so in *every* *single*
move, without *ever* making *any* technical mistake, *exactly* being aware
of the respective match equities etc.
The result of these moves which make a pretty much perfect use of all
possible dice rolls may, perhaps, in effect look like lucky dice, don't you
think so?
Regards,
Peter aka the juggler
Zorba
> I have been playing against gnubg for the past couple of weeks (maybe 50
> matches total).
>
> In the 70's I played bg in the clubs 4-6 hours a day almost every day.
> That's a lot of money games. Let's put it this way. If I found myself
> playing against a human that could roll perfectas as well as gnubg does, I
> would have gotten up and ran out of there a long time ago. I am talking
> about "doubles off the bar to bring in two men, cover a blot, and hit" type
> of doubles; easily get out of a 5-pt prime; starting the game where every
> roll makes a point (it never rolls a 5 unless it's a set of them); hardly
> ever dancing more than once; I can name a few more if needed.
It's common, especially if it's your first experience with a bg
program, to have the perception that it rolls lucky. What all the
causes are for this, I don't know exactly, but I guess it has
something to do with the fact that the bots play at a "world class"
level, so yoiu might be playing a (much) stronger opponent than you've
ever played before. Also, the speed of play can make bad luck seem to
happen more often. Some players mighyt also start to play worse
against a bot if they lose a lot of matches against them, because of
frustration.
It's certainly possible that GNUBG's dice would be "unfair", however
since all the source code is readily available to ANYONE, and you can
compile your own GNUBG from it if you wish, everyone's invited to
check the source code and see if the random dice generator is fair or
not.
There haven't been any complaints so far.
Also, you can set GNUBG to manual dice, roll your own favorite dice
and enter the rolls in GNUBG.
--
_
/
_ orba
Murat Kalinyaprak
Pete wrote uub94g3...@corp.supernews.com
> I know, I know. I hear the counter-arguments. 1). Don't play it if you
> don't like it. 2). Hey, it's free. 3). Why would they mess with the dice
> - it's not to their benefit.
None of these are valid counter-arguments...! And I'm pretty sure
that NONE of the people who offer the argument that gnudung's
source code is public has ever looked at its source code... They
are simply assuming-ass-kissers...
> I am not smart enough to know or to claim I know what's going on.
Why are you putting yourself down...? Are you intimitated by a herd
of dumb bluffers...?
Demonstrating that gnudung cheats is a child's play and I gave step
by step instructions more than once here on how one can prove it to
himself...
The piece of dung is incapable of doing 1-ply luck analysys but has
no problem with 1-ply cheating... :))
What this means is that after having already done a 1-ply analysis,
when push comes to shove, it will roll the numbers that best counter
your top two (or sometimes more) possible moves... If you honestly
don't know what I'm talking about, just ask and I will be glad to give
examples to illustrate and further explain this...
To anybody who can claim to be a world-class (or, god forbid, an
extra-terrestrial) level player, this should become obvious rather
quickly and then it gets simply boring... :((
What some of us wonder is how long this sick game will continue
and/or when there will be a non-cheating version of gnudung will
be released...??
Al Walker
> "Pete" <pe...@garlic.com> wrote in message news:<uubbpsq...@corp.supernews.com>...
>
> > I have been playing against gnubg for the past couple of weeks (maybe 50
> > matches total).
> >
> > In the 70's I played bg in the clubs 4-6 hours a day almost every day.
> > That's a lot of money games. Let's put it this way. If I found myself
> > playing against a human that could roll perfectas as well as gnubg does, I
> > would have gotten up and ran out of there a long time ago. I am talking
> > about "doubles off the bar to bring in two men, cover a blot, and hit" type
> > of doubles; easily get out of a 5-pt prime; starting the game where every
> > roll makes a point (it never rolls a 5 unless it's a set of them); hardly
> > ever dancing more than once; I can name a few more if needed.
>
complaints I had read about other bots were true. With BGBlizt you can
download matches played by the so call experts or world class players.
I have examined over 500 of the matches and was surprised to see that
all the big name folks got the same dice that I did. Multipy 6-6s. 6-1
from the bar hitting and putting their opponent on the bar against a
closed board.
Then I went back and re-read Magriel, Robertie and Woosley's stuff and
improved my game and actually started winning about 50% of my matches
against BGBlitz. I was also able to get a rating in the upper 1600s on
Fibs (I am still under 400 experience) . I never have been over 1600
on any site that I have played on.
Once I learned to diversify and duplicate I started to get a lot more
luckier. So I no longer believe that the Bots cheat, I am truely
conviced that people are not as good as they think they are.
Gregg Cattanach
is an independent web site created just to generate random numbers. Pick a
different one.
The truth of it all is the RNG has NO idea what the position is. It just
picks two numbers and that's it.
One possible explanation of the endless complaints that bots/servers cheat
with the dice is that you can't 'see' the dice being rolled. It's as if you
sit down to a table to play backgammon, but instead of rolling your own dice
a third person rolls for both players in a shoebox and then announces what
the roll is. When your opponent gets a joker, it would be pretty easy to be
suspicious and ask "Are you SURE that's the number??"
Gregg C.
"Pete" <pe...@garlic.com> wrote in message
news:uubbpsq...@corp.supernews.com...
Ron & Ann Barry
I have been a computer programmer, software engineer, and designer, developer, and maintainer of both systems and applications programs since 1961. I am proficient in C language software, in which gnubg is written. I have looked at the source code, in great detail, and I can state categorically that your claims regarding gnubg's cheating are entirely without merit. If you yourself are, or have been, a programmer, I challenge you to read the source code yourself, and point out that portion of it which incorporates the "cheating" in specific detail. If your current knowledge of C language is insufficient for the task, there are many good books and online tutorials on the subject. Surely a man of your intelligence and dedication should have no difficulty in accomplishing this assignment.
Regards, Ron Barry.
Murat Kalinyaprak wrote:
*** post for FREE via your newsreader at post.newsfeed.com ***Pete wrote uub94g3...@corp.supernews.com
> I know, I know. I hear the counter-arguments. 1). Don't play it if you
> don't like it. 2). Hey, it's free. 3). Why would they mess with the dice
> - it's not to their benefit.None of these are valid counter-arguments...! And I'm pretty sure
that NONE of the people who offer the argument that gnudung's
source code is public has ever looked at its source code... They
are simply assuming-ass-kissers...
<< additional text deleted >>
Kees van den Doel
Ron & Ann Barry <r-ab...@telocity.com> wrote:
>I have been a computer programmer, software engineer, and designer,
>developer, and maintainer of both systems and applications programs
>since 1961. I am proficient in C language software, in which gnubg is
>written. I have looked at the source code, in great detail, and I can
>state categorically that your claims regarding gnubg's cheating are
>entirely without merit. If you yourself are, or have been, a
>programmer, I challenge you to read the source code yourself, and point
>out that portion of it which incorporates the "cheating" in specific
>detail. If your current knowledge of C language is insufficient for the
>task, there are many good books and online tutorials on the subject.
>Surely a man of your intelligence and dedication should have no
>difficulty in accomplishing this assignment.
The way it works is I think that a certain set of seemingly normal C
statements activates a secret part of the GNU gcc compiler, which then
generates the cheating code without it ever showing explicitly in the
source.
Kees (You only being paranoid, Serdar, dearest, could hear thrashing and
were crossposted SPAMS.)
Pete
them.
I don't know what's going on. I am pretty convinced that none of the bots'
programmers did anything - there is little reason to. I also do know that
gnubg consistently rolls more perfectas than anything I've seen in live
games. It has nothing to do with it being at a world class level. It has to
do with the (perceived, at the very least) ability to roll what is needed in
hitting situations, getting off gammons, escaping, you name it. Conversely,
I have trouble coming in on its 2- and 3-point boards, miss double shots
left and right, etc., etc.
One final point on gnubg. I already do the post-match analyses. I have yet
to see gnubg mark any move (either gnubg's or the user's) as good or very
good. Are these features actually implemented or are they just place
holders? How would these be defined?
Again, thanks for your comments.
Pete
"Albert Silver" <silver...@hotmail.com> wrote in message
news:f9846eb9.02112...@posting.google.com...
Pete
> 1. The first is a psychological one. I call it "The Red Traffic Light
> Syndrom": You just forget the green ones so quickly. (I.e. analyze the
> games, check the luck factor, and I bet you'll see that you have in fact
> been luckier than you perceived, because it seems to lie in the human
nature
> to notice bad luck much clearer than good luck, which is easier attributed
> to ones skills.)
No doubt there is a psychological effect. But I am not saying that "gnubg is
beating me, so it must be luckier". I am talking about the preponderance of
individual rolls that are beyond the norm: the number of 17-1 shots it
makes, the number of 3-point boards I don't enter, etc.
> 3. The third is a bg argument, and it's not new to you, I guess; I just
want
> to emphasize it: gnubg, as an excellent player, chooses the move which
gets
> (to its not-too-bad knowledge) the best results with all the possible
rolls
> for the next three moves (if playing 2-ply), weighted by their respective
> probability. Even if some human players can theoretically do so in
> sufficient exactness, they will probably rarely do so in *every* *single*
> move, without *ever* making *any* technical mistake, *exactly* being aware
> of the respective match equities etc.
>
A related point is that gnubg's analysis generally says that the user has
more lucky rolls than gnubg. This is due to how it calculates lucky rolls,
if I understand it correctly. It calls a roll lucky when it significantly
increases one's equity. Since gnubg is able to look ahead better than the
user, generally its equity swings are more smoothed out than the user's, and
hence gnubg will have less sudden changes, and therefore less rolls that it
would call "good rolls". I'll concede that this would contribute to one's
preception that something is not right.
Thanks again,
Pete
Volsano
> Subject: Re: 1-ply cheating... (was Re: gnubg's dice revisited)
> The way it works is I think that
> a certain set of seemingly
> normal C statements activates
> a secret part of the GNU
> gcc compiler, which then
> generates the cheating code
> without it ever showing
> explicitly in the source.
Hmm, that's certainly possible. But how could we prove it?
Not by looking at the publicly available source for the gcc compiler -- that
may deliberately withhold the "bg cheat" code.
So maybe we could compile the gcc compiler with itself to produce a compiler
that doesn't contain the secret bg cheat code? No, they'll have thought of
that. We need an executable that hasn't been tainted by those GNU guys at all.
We could try an entirely different compiler -- maybe a commercial one to
compile GNU-BG. But I just bet that would give us cheating dice too. That'd be
useful further evidence that the whole world is conspiring to make Murat lose
at backgammon.
So let's get serious here. We need to write a C compiler from scratch (not that
difficult, really) in a clean-room environment so none of the Men in Black who
insert the BG cheat code can sneak in and corrupt the project.
But how could we be sure They hadn't -- somehow -- got in? There is no way we
can prove a negative -- the only real evidence we will ever have is if Murat
continues to lose a lot.
Indeed, even if we could prove to ourselves (and Murat) that out code did not
contain any BG cheat modes we'd still have to explain why he loses a lot.
Perhaps backgammon cheating is actually quantum-mechanically woven into the
very fabric of our universe. This means that no matter what we do, the dice
will fall badly for us all. (Except Kit Woolsey and a few others who must,
following this line of reason, have paranormal powers that affect dice).
If that is the case, then maybe the best thing any the rest of us can do is
take up chess.
Colin.
Jim Segrave
>to see gnubg mark any move (either gnubg's or the user's) as good or very
>good. Are these features actually implemented or are they just place
>holders? How would these be defined?
They are placeholder at the moment. Gnubg evaluates moves and ranks
them according to their equity, under this scheme moves can only be
the 'correct' move or an error of some degree.
I suppose you could define a move as good/very good if it is a move
that in 0 or 1 ply evaluation was not the best move, but with deeper
evaluation became the favourite. The 'goodness' could be a measure of
how bad the move looks in terms of 0 or 1 ply evaluation compared to
what it's deeper ply evaluation would yield. The question is if this
is actually a useful thing to measure. It makes sense to report the
equity cost of less than optimal moves, but reporting the equity gain
of what amounts to looking ahead is not actually the same.
--
--
Jim Segrave j...@jes-2.demon.nl
Jim Segrave
...
>A related point is that gnubg's analysis generally says that the user has
>more lucky rolls than gnubg. This is due to how it calculates lucky rolls,
>if I understand it correctly. It calls a roll lucky when it significantly
>increases one's equity. Since gnubg is able to look ahead better than the
>user, generally its equity swings are more smoothed out than the user's, and
>hence gnubg will have less sudden changes, and therefore less rolls that it
>would call "good rolls". I'll concede that this would contribute to one's
>preception that something is not right.
A minor point - a roll is marked lucky if it *could* significantly
increase one's equity (or if it causes significantly less loss of
equity when your game is going downhill) compared to other rolls. The
luck evaluation is only done at 0 ply, so there are definitely
situations where the luck calculation can be significantly wrong for a
roll. But your point that gnubg, because it plays a stronger game,
will tend place itself in situations where as many rolls as possible
will be favourable and as few as possible will be disasterous, you
will probably see a closer to zero overall luck rating for gnubg.
Gus
. //\ \___,,,_ _n_‚‚‚
. // (____\\_)=\`ůll|
Jive Dadson
counting system and the conditions of contest, using both theory and simulation.
Jive Dadson
>
> The way it works is I think that a certain set of seemingly normal C
> statements activates a secret part of the GNU gcc compiler, which then
> generates the cheating code without it ever showing explicitly in the
> source.
>
> Kees (You only being paranoid, Serdar, dearest, could hear thrashing and
> were crossposted SPAMS.)
Hahaha. I assume you are making a funny. It is possible of course. The Woz once gave a lecture in which he detailed such possibilities. The audience was rather purplexed. They had expected a talk about near term advances in technology or something.
If you really think the gnu compiler secretly miscompiles gnubg source code (and its own source code too, presumably) -- which I know your really don't, but play like -- then take the gnu compiler source code and compile it using a Microsoft or Borland compiler. Or just compile gnubg using a different compiler. Sure it might take a little work to port the code, but think of the benefit to humanity! Or is the language development tools division of Microsoft in on the big backgammon bot conspiracy too?
Jive