Jellyfish dice bug!?
Murat Kalinyaprak
(3.01 player/trial) that I'm using. If other users can
duplicate it on their computers, JF's developers may want
to do something about it pretty quick as it seems to be a
little more than a harmless bug and may have some serious
implications.
Here is what happens. You start a new game, with the
seed=777777777 and counter=0. Player2 starts with 31,
followed by 32, 54, 51 and then player2 rolls 44. I was
player2 of course, and at that point I clicked on my
piece on the 24 point. When I saw it move eight spaces
at once and hop over to the 16 point, I thought I was
hallucinating. I took the move back and tried a few more
time to see that this is what was happening indeed.
I thought the program code in memory might have been
corrupted by a Win95 bug or something. So I closed JF,
reopened and tried it again. The same thing happened.
Then I thought maybe Win95 was loading the program from
the cache instead of the hard drive and that maybe the
program code in the cache was corrupted. So I closed
down Win95, rebooted my computer, reloaded JF and tried
again, to see that it was still happening.
Then I decided to play on and see what else weird would
happen. But I could play only 3 times 4 and JF woudn't
let me play the 4th one. So, I figured I should let JF
play this roll and see what it would do. If you start
a new game at counter=1, player2 gets the 32 and a few
rolls later JF gets the 44. Lo and behold, JF moved a
piece from the 1 point to the 9 point, moved another
one from the 5 point to the 9 point and stopped there.
But now since it's JF's move, you can click on the
"show last move" option to see what it rolled and how
it played it. The little window shows that it rolled
an 84!
And JF happily went on to make the best out of that
roll, like any other roll. If you didn't follow its
moves closely, you might not have noticed that it
only played 3 times 4 and stopped there, because the
dice icons were showing 44...!
I played hundreds of games by now and never felt the
need to look at his moves. I just assumed that it
moved his pieces around properly for a given roll,
without making mistakes. Actually, even in this case
it played the 84 properly, but the dice icons were
showing a 44. So, if I hadn't noticed this because
it first happened when I tried to move one of my own
pieces, I probably would have not noticed the missing
fourth 4 in JF's moves...
But this is not the end of the story yet. Just to see
what else may happen, I played on. Two rounds later,
it rolled another 44 for itself. This time it took a
piece from the 12 point, moved it happily to the 2
point, hitting my piece there, and covered it with
another piece moved from the 6 point. When I clicked
on "show last move", I saw that it had rolled a 114!
Since we are now up to 2 digit numbers per die, it
will be necessary to use a separator in between, so
that you can tell this was an 11-4, not an 1-14... :)
What happens if you keep playing on? Another odd move
happens at counter=21, it looks like it rolls a 41 but
you see a piece flying accross the board, all the way
from the 12 point to the 24 point. It had actually
rolled a 1-11 (or 11-1:)... I didn't play any further
than that to see what else would happen later on.
It looks like a section of the code generates a random
number, another section converts it to an invalid dice
numbers, somehow the section that displays the dice
icons interprets them as valid and displays some valid
dice pair anyway, and finally, the section that evaluates
the board position and makes the "best" move for a given
dice pair accordingly, has no problem at all with playing
an 8-4, 11-4, 1-11, etc...
So far, I haven't thought any deeper about the possible
implications of this. I was more concerned about making
sure that I can duplicate the problem enough times and
in enough different ways, in order to convince myself
that this was indeed a problem with JF and not my
computer, before mentioning here. Now I would like to
hear from other readers if they can duplicate this on
their own computers, with whatever version of JF they
may have (but especially v3.01). If others can duplicate
this with the same seed/counter settings (or even with
other combinations), then it can perhaps be officially
declared a bug. And if that happens, then I'll enjoy
sitting back and reading what everybody will have to
say about it...
MK
Casual_Observer
Running 3.0. At that seed it performed
the same sequence with the same result.
Have a Marvelous Afternoon,
Richard
Jeremy P. Bagai
But you fail to mention the obvious advantages of this
"undocumented feature": Now we can finally expand our
best reply to opening rolls tables to include 8-4 and
the elusive 1-11.
Wow.
Jeremy Bagai
Rodrigo Andrade
only does it hop over 8 pips at once, but it also completely ignores the
blocked 8 point on its board.
RODRIGO
spurs
The mind boggles.....
Casual_Observer wrote:
> Murat,
>
> Running 3.0. At that seed it performed
> the same sequence with the same result.
>
> Have a Marvelous Afternoon,
> Richard
>
> Murat Kalinyaprak wrote:
> >
> > There seems to be a serious bug in the Jellyfish version
> > (3.01 player/trial) that I'm using. If other users can
> > duplicate it on their computers, JF's developers may want
> > to do something about it pretty quick as it seems to be a
> > little more than a harmless bug and may have some serious
> > implications.
> >
> > Here is what happens. You start a new game, with the
> > seed=777777777 and counter=0. Player2 starts with 31,
> > followed by 32, 54, 51 and then player2 rolls 44. I was
> > player2 of course, and at that point I clicked on my
> > piece on the 24 point. When I saw it move eight spaces
> > at once and hop over to the 16 point, I thought I was
> > hallucinating. I took the move back and tried a few more
> > time to see that this is what was happening indeed.
> >
my_...@hotmail.com
mu...@cyberport.net (Murat Kalinyaprak) wrote:
> There seems to be a serious bug in the Jellyfish version
> (3.01 player/trial) that I'm using. If other users can
You're right. If you set the seed to 777777777 (nine sevens),
and the counter to 4, the computer rolls, (increasing the counter
to 5) the dice will show 4 4, but it allows you to play as
though you've rolled 8 4. With the same seed, if you set the
counter to 20 and then roll, the dice display 4 and 1, but you
get to play a roll of 11 and 1. This appears to be true
regardless of how the checkers are positioned.
This isn't the first outrageous bug in Jellyfish I've read
about in this newsgroup. No one has ever accused it of having
an even passable user interface. But I think this bug is
only user interface related and doesn't affect the neural net
evaluation, or support a claim that the program cheats.
-----== Posted via Deja News, The Leader in Internet Discussion ==-----
http://www.dejanews.com/rg_mkgrp.xp Create Your Own Free Member Forum
Murat Kalinyaprak
>Same results with my copy of 3.0.
>But you fail to mention the obvious advantages of this
>"undocumented feature": Now we can finally expand our
>best reply to opening rolls tables to include 8-4 and
>the elusive 1-11.
>Wow.
Don't forget the 11-4, 9-2 and 4-0... :)
MK
Murat Kalinyaprak
But that's ok. As I explained before, it had rolled
an 8-4 and was playing an 8-4. With an 8, you can
even jump over a solid 7-block... :)
MK
Murat Kalinyaprak
>In <6pp062$gnd$1...@news.chatlink.com> Murat Kalinyaprak wrote:
>> There seems to be a serious bug in the Jellyfish version
>> (3.01 player/trial) that I'm using. If other users can
>You're right. If you set the seed to 777777777 (nine sevens),
>and the counter to 4, the computer rolls, (increasing the counter
>to 5) the dice will show 4 4, but it allows you to play as
>though you've rolled 8 4. ......
If you poke around, you can find many more such weird
rolls with all kinds of seed/counter combinations...
>This isn't the first outrageous bug in Jellyfish I've read
>about in this newsgroup. No one has ever accused it of having
>an even passable user interface. But I think this bug is
>only user interface related and doesn't affect the neural net
>evaluation, or support a claim that the program cheats.
Not so fast... I'm not sure if this can be just
shrugged off so easily.
What do you mean "only user interface related"? It
seems to go right to the heart of the program. It
generates invalid dice numbers. If it had rolled
4-4 and played 4-4 but the dice icons errenously
showed 8-4, then you could call it a "user interface
problem". It's the reverse that is happening here!
Where are all the people who put me through "RNG-101"
just a few days ago? Let's hear what they have to
say about this? Whether dividing integers by 6 or
36 and adding 1 to the remainder, or interpreting a
sequence of bytes 6 bits at a time (as Gary had
explained I believe), I would think that the task of
generating dice numbers would be trivial. Shouldn't
this part of the program be no less than rock-solid
in a software of JF's claimed caliber? Could it be
that something more "awkward" is tried (for whatever
reason) and not succeeded at smoothly enough, etc.?
Maybe we should give JF's developers a fair chance
and enough time to respond/explain, before we go on
to speculate much further.
MK
Fredrik Dahl
> Maybe we should give JF's developers a fair chance
> and enough time to respond/explain, before we go on
> to speculate much further.
>
Thanks :-).
There should have been a magnitude check for seed<65535.
Sorry!
At least the seeds JF creates automatically satisfy this,
which probably explains why it took so long for so many
to stumble into it.
Weeding out all bugs of complex software is hard.
In this case it appearently took tens of thousands of user hours
to find it.
Fredrik Dahl.
Rodrigo Andrade
problem.
The interface is also mediocre!!!! I've asked once here if there was a
better interface to download somewhere on the Net. Later I found out that
even the Analyzer uses this poor interface too.
RODRIGO
Chuck Bower
Murat Kalinyaprak <mu...@cyberport.net> wrote:
>There seems to be a serious bug in the Jellyfish version
>(3.01 player/trial) that I'm using. If other users can
>duplicate it on their computers, JF's developers may want
>to do something about it pretty quick as it seems to be a
>little more than a harmless bug and may have some serious
>implications.
>
>Here is what happens. You start a new game, with the
>seed=777777777 and counter=0. Player2 starts with 31,
>followed by 32, 54, 51 and then player2 rolls 44. I was
>player2 of course, and at that point I clicked on my
>piece on the 24 point. When I saw it move eight spaces
>at once and hop over to the 16 point, I thought I was
>hallucinating. I took the move back and tried a few more
>time to see that this is what was happening indeed.
(snip)
And if you think THAT is bizarre, you ought to see
what happens with seed = 666! (WARNING!! Children under
the age of 102 should be protected from seeing this. WARNING!)
When I plugged this into JF, my monitor did an amazing 360 degree
rotation and then spat a nasty pea-grean soup looking goo all
over my face....
Chuck
bo...@bigbang.astro.indiana.edu
c_ray on FIBS
Chuck Bower
Murat Kalinyaprak <mu...@cyberport.net> wrote:
(snip)
>Maybe we should give JF's developers a fair chance
>and enough time to respond/explain, before we go on
>to speculate much further.
I must say that in all of the ten bazillion lines of commentary
written by Murat in the past two or three weeks, this one appears
to be the most sensible. Now, if he had just entertained this idea
early on before accusing JF of cheating (with its dice)...
Casual_Observer
is a legitimate problem with seeds over
65,534.
Keep Smilin',
Richard
Chuck Bower wrote:
>
> In article <6pp062$gnd$1...@news.chatlink.com>,
> Murat Kalinyaprak <mu...@cyberport.net> wrote:
>
> >There seems to be a serious bug in the Jellyfish version
> >(3.01 player/trial) that I'm using. If other users can
> >duplicate it on their computers, JF's developers may want
> >to do something about it pretty quick as it seems to be a
> >little more than a harmless bug and may have some serious
> >implications.
> >
> >Here is what happens. You start a new game, with the
> >seed=777777777 and counter=0. Player2 starts with 31,
> >followed by 32, 54, 51 and then player2 rolls 44. I was
> >player2 of course, and at that point I clicked on my
> >piece on the 24 point. When I saw it move eight spaces
> >at once and hop over to the 16 point, I thought I was
> >hallucinating. I took the move back and tried a few more
> >time to see that this is what was happening indeed.
> (snip)
>
> And if you think THAT is bizarre, you ought to see
> what happens with seed = 666! (WARNING!! Children under
> the age of 102 should be protected from seeing this. WARNING!)
> When I plugged this into JF, my monitor did an amazing 360 degree
> rotation and then spat a nasty pea-grean soup looking goo all
> over my face....
>
Murat Kalinyaprak
> Murat Kalinyaprak wrote:
>>Here is what happens. You start a new game, with the
>>seed=777777777 and counter=0. Player2 starts with 31...
> And if you think THAT is bizarre, you ought to see
> what happens with seed = 666! (WARNING!! Children under
> the age of 102 should be protected from seeing this. WARNING!)
> When I plugged this into JF, my monitor did an amazing 360 degree
> rotation and then spat a nasty pea-grean soup looking goo all
> over my face....
The problems that I had observed/reported have been
duplicated by other users and acknowledged by the
developer. I myself couldn't duplicate the problem
you described above. Unless it happened to some other
users also and/or is acknowledged by the developer,
you may want to consider the possibility that it may
be a personal problem between you and JF...
MK
Murat Kalinyaprak
>In <6pujjm$q24$3...@news.chatlink.com> Murat Kalinyaprak wrote:
>>Maybe we should give JF's developers a fair chance
>>and enough time to respond/explain, before we go on
>>to speculate much further.
> I must say that in all of the ten bazillion lines of
> commentary written by Murat in the past two or three weeks,
Should I take this as your complaining? If so, why?
Do you think what I had written were uninteresting,
off-topic, useless to this newsgroup? These threads
seem to have attracted participation from a larger
number of readers than a lot of other threads...
> this one appears to be the most sensible.
If you had looked for them, you probably would
have detected many other signs of my trying to be
as responsible/sensible as I could in my previous
writings also. Notice that I had gone as far as
reinstalling JF from the ZIP file, in order to
eliminate all possibility that the problem I had
reported wasn't limited to my computer, before I
talked about it here publicly. Even then, I put
a "?" after the subject header and commented in
my article that this should be declared a bug
only after other users duplicate it and/or the
developer acknowledges it. I may be criticized
for being this and that, but I would like to
believe that I'm a reasonably fair person...
> Now, if he had just entertained this idea early on
> before accusing JF of cheating (with its dice)...
But I had done that, at least to a certain degree
I had considered reasonable enough. If you remember
or reread my first article on this subject, you will
see that the basis I offered for my impression/opinion
about JF's cheating was that I felt I could predict
the dice rolls to come (not because I was feeling sour
that it always beat me, to the contrary I thought that
I was beating it as often or more often). I had tried
to validate this impression to myself by making
dry-runs at guessing dice rolls and compiling some
statistics about it. While I was mentioning such things
in this group, I never denied that it may all be just
my imagination and solicited opinions on the subject
from other readers. How much more would you expect
that someone considers before expressing an initial
impression/opinion on such a subject?
MK
Murat Kalinyaprak
>> Maybe we should give JF's developers a fair chance
>> and enough time to respond/explain, before we go on
>> to speculate much further.
> There should have been a magnitude check for seed<65535.
Fredrik, thanks for the effort but such a concise
explanation may not be obvious enough for some of
us who may not be sufficiently intelligent and/or
knowledgeable on the mechanics of dice generation.
Would you care to eleborate a little more by any
chance?
> Sorry!
> At least the seeds JF creates automatically satisfy this,
I guess the above magical number implies something
being limited to 16 bit? And what would "satisfying
a magnitude check" lead to in this situation? Not
allowing players to enter a seed value greater than
65535? If that's the case, would that give any sort
of basis/validity/justification for my previously
wondering whether JF's allowing 9 digit seed values
wasn't beyond the real scope of its dice generator?
> Weeding out all bugs of complex software is hard.
I surely sympathise with this statement. However,
I would like to express my opinion that I would
expect the complexity of such software to be found
in its sections that evaluate positions/make moves,
etc. In fact, I don't think that the dice rolling
function is even an essential part of the software.
This view is supported by your and other writers'
efforts (i.e. arguments based on seed/counter) to
segreagate JF's playing strenghts/weaknesses from
the dice (whether rolled manually or automaticly).
This being the case, would you consider making
JF's algorithms that generate random numbers and
derive dice rolls public? This surely would be
much more effective in putting peoples minds at
ease on this topic than all those much repeated
seed/counter based arguments.
> In this case it appearently took tens of thousands of
> user hours to find it.
In my case, just "tens" without the "thousands"
hours... My poking around to see if I could find
some patterns in the dice rolls must have speeded
it up. I have to admit that I wasn't expecting to
run into dice numbers greater than 6 or less than
1 (in fact some are negative). If someone had asked
me to develop an algorithm that would produce valid
dice numbers 99+%? of the time and some invalid
numbers on purpose, in a manner related to the seed
having to do with being less or grater than 65535,
I probably would have to scratch my head for a very
long time.
On top of this, I had played 100 games with the seed
value 999999999 and another 100 games with seed value
888888888, and no invalid dice numbers had occurred.
Yet, with the seed value 888888887 one runs into
invalid dice numbers just after a few rolls. So, I
hope you will be understanding of my curiosity not
being satisfied yet (by your brief explanation) about
how this happens in JF...
> Fredrik Dahl.
MK
PS: I have absolutely nothing aginst JF or you for
any particular reason. I got sucked into this mostly
because I was irked by some comments putting down
people as "lesser players", etc. just because they
happened to express their opinion about JF's "apparent
luck", etc. Such impressions from players can be and
perhaps should be taken as compliments further proving
JF's worth. But when you add to these comments like no
"respectable/world-class/etc." player had criticized
JF for being "exceptionally lucky", etc. it all leads
to a false logic that says: "If a player doesn't suspect
JF of cheating, he must be a good player and an expert
on the subject. If he does, then he must be a lesser
player". I just felt a need to defeat such erroneous
deductions and receiving sharp/daring responses got me
even deeper into these discussions.
I also just found out that you made JF player available
now as freeware. I think this is a great contribution
on your part, which will hopefully make JF even more
popular that it seems to be already. It will probably
make most of us feel owing something back too. I hope
that I myself will eventually be able to contribute
something positive to JF in the future...
Fredrik Dahl
> > There should have been a magnitude check for seed<65535.
>
> Fredrik, thanks for the effort but such a concise
> explanation may not be obvious enough for some of
> us who may not be sufficiently intelligent and/or
> knowledgeable on the mechanics of dice generation.
> Would you care to eleborate a little more by any
> chance?
>
Sure,
The problem is that the random number generator does not work
properly if the seed is greater than 65535.
It has to do with how integer numbers are represented
in the pc.
The program should have tested the size of the seed given,
but appearently this was forgotten.
Fredrik Dahl.
Joseph B. Calderone
--
Joe Calderone <joe...@roanoke.infi.net>
"Don't force it, Use a bigger Hammer. Well,
No sense being pessimistic. It wouldn't work anyway"
Murat Kalinyaprak
Murat Kalinyaprak previously had written:
"Maybe we should give JF's developers a fair chance
"and enough time to respond/explain, before we go on
"to speculate much further.
Fredrik Dahl previously had responded:
"Thanks :-).
"There should have been a magnitude check for seed<65535.
"Sorry!
"At least the seeds JF creates automatically satisfy this,
"which probably explains why it took so long for so many
"to stumble into it.
"Weeding out all bugs of complex software is hard.
"In this case it appearently took tens of thousands of user
"hours to find it.
Fredrik Dahl wrote:
>Murat Kalinyaprak wrote:
>> Fredrik, thanks for the effort but such a concise
>> explanation may not be obvious enough for some of
>> us who may not be sufficiently intelligent and/or
>> knowledgeable on the mechanics of dice generation.
>> Would you care to eleborate a little more by any
>> chance?
>Sure,
>The problem is that the random number generator does not work
>properly if the seed is greater than 65535.
>It has to do with how integer numbers are represented
>in the pc.
>The program should have tested the size of the seed given,
>but appearently this was forgotten.
Thanks again, and this time I also apologize for
"making you repeat yourself" (as the person using
the name "my_...@hotmail.com" had put it, in his
article talking about Tiger Woods, golf balls and
pine cones; otherwise I wouldn't have used such an
expression myself)...
For a moment, I even wondered whether I shouldn't
feel really stupid for expecting anything more as
answers to such dumb questions I had asked in the
first place. I try to be not so curious/eager to
know/willing to learn, but I just can't help it...
For example, after reading (in your first answer)
that "the seeds JF creates automatically satisfy
this (i.e. magnitude check for seed<65535)", I had
wondered if I had ever noticed seeds created by JF
even coming close to such magnitudes like 65000+?
I couldn't remeber for sure but didn't think I had.
So, I started playing with the Windows clock, upping
the minutes and hours, and kept restarting JF to
see how high the seeds JF generated would go. At
7:19 and some seconds, it reached 31987 but at 7:20
and some seconds it went back down to 33. If I wasn't
worried about the sarcasm of some writers here, I
might have now asked even more dumb questions such
as whether that function of JF doesn't do a magnitude
check based on a different magic number (like 32768)?
But I won't ask, because even I:) can understand that
a magnitude check for <100 automatically satisfies a
magnitude check for <100000000:) and also because I
started having nightmares about getting more replies
from my_...@hotmail.com & Co...
Last night I woke up by a soft voice saying: "It's ok
dear, it's ok. You were just having a nightmare". There
it was, this anonymous reply on the screen that said:)
"A computer is a little black box with tiny midgets
"inside. When you click on the dice icons, they roll
"dice and paint the numbers on the screen. One of
"the midgets is a mischievous one. Once in a while he
"substitutes the doubling cube for one of the dice .
"That's why you get rolls like 84. And the negative
"numbers are of course, explained by how 5-bit signed
"integers are stored in the computer."
Since I was soaked in sweat, I took a cool shower
before going back to sleep. (Does this satisfy the
"sweaty ... scene, followed by cool shower scene"
requested by "Casual_Observer <divdesm...@home.com>"
as an obligatory passage in my "novella"...?:)
Anyway, enough of this "irrelevant" stuff. I wish
Fredrik "double sixes" in fixing this little bug...
MK
PS: May I coin "wishing someone double sixes" to mean
"wishing someone good luck" (if not already done by
others)? BTW, the word "fixing" in the last sentence was
used with no intentions other than to mean "correcting".
Stephen Turner
>
> PS: May I coin "wishing someone double sixes" to mean
> "wishing someone good luck"
Not if they're on the bar.
Serious question. Suppose I change the probabilities on my dice (not my
opponent's) to give p double sixes instead of 1/36, and (1-p)/35 of each of
the other rolls. We both play according to the optimal strategy for regular
dice. How does my equity change as p varies?
--
Stephen Turner sr...@cam.ac.uk http://www.statslab.cam.ac.uk/~sret1/
Normally: Statistical Laboratory, 16 Mill Lane, Cambridge CB2 1SB, England
Until 12/98: Dept of Math & Stats, 585 King Edward Ave, Ottawa K1N 6N5, Canada
"Ad infinitum, if not ad nauseam." (Interviewee, BBC Radio 4)