B
Sent: Monday, June 05, 2000
Subject: Request for Raw Data
Greetings,
Several weeks ago, I found your web site. I was immediately intrigued
by the prospect and challenge of predicting roulette outcomes using
sound scientific principles.
Since then, I have read Eudaemonic Pie and Robert Nathans' patent
document on his roulette betting calculator. I have also perused
several elementary texts on electronics and secured a copy of
Electronics Workbench Multisim ver 5.0c.
I am in the process of designing a roulette timer device similar to
Nathans' roulette calculator. I need some data on ball rotation
times (expected max, min, and mean) for a wide range of wheels and
ball types. Your data would save me a LOT of time otherwise spent
in casinos gathering this basic information.
Kind regards,
Dan
Sent: Saturday, March 09, 2002
Subject: RE: data receipt
I just received a copy of Newtonian Casino by Thomas Bass. I am reading it this
weekend. I have also checked out some electronics texts and am looking
into purchasing an OOpic (see
www.oopic.com).
My head is full of ideas; but, to realize them I must educate myself in
electronics and acquire the necessary tools and
equipment to construct the prediction device. This will take time.
I will check back next week.
Dan
Dear patsy,
Try this AMAZING CARD TRICK! Its' Fun, It's Fee:
nowscape.com/blk/card-trick/index.htm
patsy here is the card trick i found on the wb
Signed... brent
Sent: Friday, August 14, 1998
Subject: Roulette data
Hi
I fished your web page out of a Metcrawler search about an old passion of mine -
predicting roulette. I have planned for years to finally get around to coding up
a neural net prediction system running on an embedded Intel chip. Neural nets
have the amazing ability to divine linear and non-linear correlation functions
from two data sets where *some* kind of correlation exists.
I remember reading about Doyne Farmer and crew's methods of hand (or foot)
tuning the five variables in their equation manually to each wheel. I also
chatted with a gentleman a few years back who claimed that a 50 millbar jump in
pressure would throw fixed predictions off. A neural based system can be built
that continually retunes itself from all the previous predictions up to the
present. It will fit itself custom to each wheel and adjust dynamically as
conditons, such as air pressure change. I think such a system is probably the
only practical one under variable playing conditions.
If you think this is a worthy project I would love to get a copy of your data to
throw an untrained NN at and see what it comes up with.
Sincerely,
James
Stockholm, Sweden
Date: 15 August 1998
Subject: Re: Roulette data
Hello, James --
Yes, of course you are welcome to our data .. in fact I think we have a sample
and description of format of the data on the web page you saw. I too have for a
long time now thought that really the best way to attack this problem is with
neural nets. And yes, I've read about Doyne Farmer's work in the book Eudaemonic
Pie. In fact, for some reason I was think thinking about this fellow just today,
wondering what ever happened to him... I think he also did some work on Fractals
/ Stock market prediction.
I'm not sure we have enough cases for you to train your net AND verify its
predictions. All data sets are from different roulette wheels -- but you are
welcome to try.
What brings you all the way to Sweden? You seem to be a native speaker of
English.
Sincerely,
Sent: Saturday, August 15, 1998
Subject: Re: Roulette data
I'm looking forward to playing with your data. I may need a little deeper
explanation of just what you guys measured. The 'ideal' data I am after consist
of wheel speed (ws), ball speed (bs), elapsed time between the when bs and ws
measurements were made, plus the final pocket the ball lands in.
The main problems with neural nets are the required floating point (or fixed
point if you want to write your own math routines), long training times, and
optimal net structure. The training times can be lessened by either using a
partially 'pre' trained net or by selecting a net architecture that trains
quickly. The optimal net structure can be found empirically, built into the net
algorythm itself, or found using an evolvable architecture (ie. variants of so
called genetic algorythms).
I spoke to Doyne Farmer in New Mexico about four years ago. He was running The
Prediction Company, Inc. and not at all interested in Roulette any more. He
apparently had found backers for funding a project predicting currency exchange
rate fluctuations using parts of chaos theory.
I have been in Sweden for fours years now and ended up here after meeting a
Swedish girl while mining gold in Honduras - which is a rather long story. Kan
du lite svenksa kanske? Originally I hail from about 15 miles North of San
Francisco over the Golden Gate Bridge.
Cheers,
James
Sent: Monday, January 26, 1998
Subject: Re: Raw Data - Roulette
Like I mentioned in a previous email, I have been studying
roulette for some
time. I am interested in your raw data. I have several
questions regarding
the information presented on the web page.
1. The data shows the time the ball landed. Is this the time
it first hit the
rotor, or the time is stopped?
--- The tine it first hit.
2. Was the data taken in a lab, or in casinos?
--- Casino, Although I believe we also have one file if about
700 trials
that were done in the lab with a normal roulette wheel (which is
for sale, BTW:)
3. Were you ato win?
--No. At the time we gave up, I thought the thing looked
promising, but
my partner did not. (We did predict the correct slot twice)
4. What is your method of calibrating the wheel?
--- We gathered ball, wheel and rotor times ahead of play, on
the pocket
computer. The we downloaded the data to an Apple II computer
(This was some time ago,
5. What is the Y axis in the POS vs TTT graphs? Is it in
1/1000ths of a
spin?
--- I believe it's time (I'm working from memory here) I
think the time
is thousands of a second, but I think it's explained on the web
page. POS is the cloc
6. Do you know if these wheels were level?
--- We don't know. We were able to determine, from our data,
that some
wheels were definitely not level. The "uphill" travel of the ball
was clearly visible
7. The graphs showing POS vs TTT seem fairly erratic given
528 spins. Do you
know why? Is this strictly due to timing errors? Something
doesn't seem right
with this. I have a great deal of data, which seems to
produce a closer fit
than this. I haven't studied your data numerically, though.
--- You may be right, but I don't think so. If I remember
correctly we
verified our ability to do consistent data entry. Nevertheless it was
all done by manual
8. Have you calculated confidence intervals on your
calibration methods? How
many spins do you need to measure to get a good calibration?
--- I don't remember that we did any stats on that.
9. Have you taken tilt into account?
--- No. We assumed the decay of the ball to be a second
degree
polynomial, and our data showed that in most cases this is a good
assumption.
10. You refer to not taking information on atmospheric data.
Do you have
calculations of the coefficient of drag on the ball, and the
rolling
resistance?
--- No. But my guess is that there is great variability in
friction
losses. One can hear it. I think a great deal of energy is converted
to sound in many wh
11. What shape is your wheel in and how much do you want for
it?
--- Our wheel is like new. But YOU don't want it. Our data
gathering on
is shows it is biased. I suppose it could be fixed, but it would
take a lot of work.
12. Your graph showing the predicted vs actual pocket implies
that you didnt
get the system to work. Is this correct? I have much better
results than I
see in your graph.
--- You're right. It would be interesting to see what you did
and HOW you
did it!!
Look forward to hearing from you.
Dave K.
==================================
Response - continued
NOWSCAPE and I did a great deal of development over several
years. It
was an entertaining, frustrating, enlightening experience. NOWSCAPE
built
several versions of communication devices - including radio (too
unpredictable, the frequencies were too busy with background noise,
the time required to pick up a signal varied by many milliseconds,
etc.). We built machines that picked up false switch presses because
of skin moisture, and which did not operate properly for a
variety of reasons.
The system as finally constructed worked as follows: I wore the
computer, switches, speech synthesizer, and infrared transmitter. I
entered the times that "00" passed a fixed position on the wheel for
a number of presses. I would continue to enter "wheel" times, even
after the ball had been thrown, until the time that the ball had
slowed to a given speed. At that time, I would begin entering
"ball" times as it passed the fixed position. The computer determined
the time to broadcast a predicted number when the algorithm determined
that "touchdown" was some predetermined time away. This predetermined
time gave enough time to predict the outcome, broadcast it to
NOWSCAPE ,
and have him place the bet before the dealer stopped all bets. This
was three or four seconds before touchdown.
In response to your questions, NOWSCAPE has answered
them in the same
manner I remember. In addition,
#3 - Nevada made the use of electronic devices for predicting
roulette a felony at about the time we had a viable system (1985?).
This would have been difficult to enforce, but the casinos could
have made life difficult and expensive for us. It didn't seem
worth continuing.
#5 - Yes, time was in 1/1000's. We did not have a clock on the
computer, but we spent alot of time calibrating instruction
times under various temperatures. We intentionally calibrated into
thousandths.
#6 - We don't really know if they were level. We collected some
data that suggested a sine wave superimposed over the normal decay
curve of the ball. We assumed that the ball times that
resulted in a smooth curve were from a flat wheel.
#7 - I don't have the data in front of me. I think this graph
represents several TTT for each POS. Thus, both a 1-second and a
2-second TTT could both represent the same pos. The graph
(if I am thinking of the right one) shows several TTT for
the same POS gathered from the same timing trial.
#8 - We measured our ability to time an object passing a fixed point,
and we collected the same data in a computer. It seems to me that we
had quite good numbers - plus or minus less than 10 ms of the actual
event. It was a long time ago, but we were concerned very early about
whether we could hit an actual event. I believe we were quite accurate.
#9 - Tilt should show up in the graph of the ball times. It did in the
wheel we used for our own collection, as I recall.
#10 - We collected for some time without realizing that the balls were
of different material and weight. Also, we did not consider barometric
pressure and I'm not sure it would have had a material difference. It
would have been worth testing in retrospect. It seems to me that any
prediction would have to be based on the same size and weight of ball.
#12 - I agree with NOWSCAPE - keep us posted.
Let me know if you need more info -
It appears NOWSCAPE and I were one of the first attempts
to actually do
this. We were really hampered by lack of good equipment and programming
tools. Much of what we did was bootstrap [hamstring? -Ed.]
work with
primitive tools. It would be a much easier task today.
Sent: Friday, April 03, 1998
Subject: roulette prediction
I found your page with MetaCrawler. Very interesting work.
Looks like we had very similar projects underway at just about the same
point in history. Only you had some skill in electronics and I had to
contract out for my stuff.
Back in the early 70s I was on the verge of buying components to have a
concealed roulette computer built. I also purchased a Radio Shack PC2 and
felt I could have it adapted if it became necessary but I thought it was
pretty bulky to use the whole device. My main use for the PC2 was to be to
get calibration data from real casino wheels. I think the think set me back
about $700. I probably still have it out in a box in the garage. OUCH!
I had rented 4 different used casino wheels (of inconsistent qualities) and
videotaped thousands of spins with markers affixed to the wheel. The RCA
video camera had a stopwatch in the picture. Handy, eh?
I calibrated ball and wheel deceleration rates and developed the software
to put this all together in real-world simulations. I conducted specific
'human click reaction time' tests.
Then Nevada came up with the device law and my plans evaporated faster than
cat spit on a hot radiator.
Coincidentally the original idea for wheel clocking came to me when I was
counting cards in Binion's Horseshow in the early 70s. At that time there
were no computers outside of mainframes. My concept was to quickly mentally
recite the alphabet for 3 wheel spins; then for 3 ball spins; and then use
the values obtained to do a mental lookup operation on a huge memorized
matrix.
When computers came along it enabled me FIRST to validate the concept.
Chief among my concerns was: "are the unavoidable human reaction time
errors of sufficient magnitude and scattering so as to make the concept of
clocking the wheel with a computer impractical"?
I found out that the errors were definitely NOT an impediment to accurate
prediction.
I found out that I would have an overall edge over the casino of about 26%.
This is HUGE! It means that if I could play unencumbered by security I
could have literally owned Caesars Palace in a few days! I felt that with
proper camouflage I could pull $3000 a weekend out of the Vegas casinos.
And, unlike blackjack, the busy weekends are the best times to play, since
the crush of the crowds would diffuse any attention that might be focused
on the clocker as an observer.
As far as the information passed between the clocker and the player, I kept
reducing and reducing the amount of info to be passed until I finally
realized that it could simply be reduced down to:
a) BET ANY NUMBER IN CERTAIN DENSE CONTIGUOUS REGION ON THE WHEEL.
or
b) bet any number OUTSIDE that region.
After having done all my work I found Allan Wilson's "Casino Gambler's
Guide". This is a masterful work. Wilson anticipated computer wheel
clocking and discusses it!
Then I found Thomas Bass's "Eudaemonic Pie". Great book. Frustrating
hardware problems sank the project as I recall.
I was at a conference at Jet Propulsion Laboratories in Pasadena in the
early 80s. Al Hibbs, the then director, mentioned that as a (brilliant)
college student he had managed to come up with a way to beat roulette and
made quite a lot of money with it.
I never found out what this system was but I suspect it was a concealed
analog computer.
I wonder what success concealed-computer roulette clockers have had. I
suspect that there have been some truly huge scores made with roulette
computers, particularly in Europe. But I think that the most successful
teams would have probably kept a VERY low profile and their exploits may
forever remain hushed whispers and rumors.
I wouldn't rule out non-computer clocking ...BUT there are marked
differences between one wheel and the next; same thing with balls. The
differences are enough to even throw off computer clocking unless you do
calibration runs for each particular ball-wheel combination. So since you
can't use a computer for calibration runs your data is going to be lower
quality. This is a problem of "extreme dependence on initial conditions".
If those measurement of those initial physical conditions contain more than
a certain threshold of error...you do NOT have a positive expectation.
Plain and simple.
We're talking about a major amount of effor to even attempt non
device-based wheel clocking. I dunno if people could drill themselves into
becoming consistantly accurate high speed human metronomes.
It would take a couple of total monomaniacs to put it into practice I
think.
What do you think?
Fred
Canoga Park CA
Sent: Monday, February 08, 1999
Subject: Roulette Data
Dear Sirs,
Firstly congratulations on an extremely informative and useful website. I
am quite sure it has already been of use to me. I am writing to you to
request a copy of your data files, and any other relevant information you
may be able to send me. I am doing some research into roulette as part of
a project at university (mainly looking at the statistics and probability
side of it, although I have a friend looking at aerodynamic effects,
different balls etc.). We hope to use some fairly high tech computer
equipment to hopefully incorporate some of the non-linearity of the
system, although your quadratic relationship seems to produce some very
impressive results!
I would be most grateful if you could send any information (particularly
the data sets and any equations of motion or algorithms), as it is fairly
difficult to get any accurate data due to the strict gaming laws in the UK
(and I can't quite afford my own wheel yet!).
I hope to hear from you soon, and would like to keep in contact throughout
my project to keep you informed of my progress.
Yours in anticipation
Nathan L.
Sent: Thursday, April 22, 1999
Subject: Roulette Computer
Hello,
Finally we found someone willing to give out some information
about roulette computers and prediction. Thank you.
Few months ago we started to think about making a roulette
computer and first thing we did was get few books, make some
internet search and such. Next step was to order a 32 inch
roulette wheel which should be coming from manufacturer in next
week. Anyways, would you be willing to help us our project?
I'd like to know how well did the roulette computer predict numbers
and how well does second-degree polynomial correspond to actual
number. Also, was the calibration of computer easy or did it take
hours to make to computer produce decent results like 20% edge over
house?
We are Finnish students and we believe that its still possible to use
roulette computers in Europe
Thank You,
Ilmari
Sent: Sunday, March 19, 2000
Subject: roulette question
Hi I was wondering if you have read Mathematics of Gambling by Dr. Edward O.
Thorp---he has his own formulas for calculating ball revolutions and i was
wondering how it compared to what you have done. For example for ball
revolutions he would time the first revolution of the ball spinning and put
that time into this equation: 1/(e^(3T/20)-1)-10/3 (if T was .5 then
the answer would be 9.51 revs) I am just guessing that if you did all or
your tests on a different size wheel then Dr. Thorp then these formulas
would be different ---or is there a casino standard size that they all use,
I have seen different size wheels for sale from gaming distributors and I
always wondered if they all bought the same size wheel.
Thanks,
Steve
Sent: Sunday, October 24, 1999
Subject: what's error?
What is dependence between error of foot-timing the ball and error of the clock
position in which the ball landed? I think small error in timing can give us big
error in prediction of clock position in which the ball landed.
Michael.
==================================
Sent: Monday, October 25, 1999
Subject: Re: what's error?
I try to repeat your project today. I don't know what for formula was in your
project.
I have now 3 formulas. It means 3 ways for realising your project. I have
modified equipment, use some new technologies.
If you interesting in my project, let me know. Of course i need some comments
for my ideas.
Michael
P.S. And by the way, what for price for your quality double-zero roulette.
==================================
Date: 25 октября 1999 г.
Subject: RE: what's error?
Yes, I think that is exactly correct. However, it the error of timing is more or
less constant, then knowing that would help a little, I think.
==================================
Date: 26 октября 1999 г. 4:17
Subject: RE: what's error?
Our formula was y=ax^2+bx+c. We collected date for each wheel BEFORE play; these
data were analyzed to determine the a, b, and c. It would be interesting to see
what your equipment and new technology is.
Thank you for writing!
-- NOWSCAPE
==================================
Sent: Tuesday, October 26, 1999
To: KRP
Subject: Re: what's error?
I understand how you collect data and analyze. But I don't understand exectly
what is "The clock position in which the ball landed" and Estimated-observed
"would-have-landed" if no bounce???", what the different?
It means you estimate: 1.revolutions time
2. time or distance from the ball start falling into the wheel and landing into
the wheel
3. distance from the ball landing to win number
or not?
If you show correlation between POSition and REVolution time, POSition includes
revolution you timing or you count only next revolutions.
What is the checksum?
What dependence between timing the wheel and predicted number?
About my project:
I connect with my computer trough D-AMPS mobile phone. The result i get not
trough earphone, I get it trough 6 motors, vibrating boolen numbers from 1 to
63.
Formulas:
I have 2:
1. POS = ((2 * pi * R / t + a * t / 2) ^ 2 - V ^ 2)/(4 * pi * a * R) (including
timing revolution)
where pi=3,1415926, R=radius of the wheel (constant), t = revolution time, a =
ball acceleration (constant), V - ball speed, ball starting fall (constant).
R, a, V constants analyzed using curve fitting routines
2. POS = (t2 ^ 2 + 2 * t1 * t2 - t1 ^ 2) ^ 2 / (4 * t1 * t2 * (t2 ^ 2 - t1 ^ 2))
- V ^ 2 * t1 * t2 * (t1 + t2)/(16 * pi ^ 2 * R ^ 2 * (t2 - t1))
or POS = (t2 ^ 2 + 2 * t1 * t2 - t1 ^ 2) ^ 2 / (4 * t1 * t2 * (t2 ^ 2 - t1 ^ 2))
- K ^ 2 * t1 * t2 * (t1 + t2)/(16 * pi ^ 2 * (t2 - t1))
(including 2 timing revolution)
where pi=3,1415926, R=radius of the wheel (constant), t1 = revolution time of
first revolution, t2 = revolution time of next revolution), V - ball speed, ball
starting fall (constant), K = V / R (constant).
R, V or K constants analyzed using curve fitting routines.
Clock starting when the ball near 0 and stop or click when the ball near 0.
My project is now in stage of creation.
I'm interesting in any data for simulating and proving my formulas and in advice
that help me in realization of my project.
I answer for question you have.
Michael
==================================
We choose a spot on the outside of the wheel and called it 12 o'clock. This is
the same, I think as what you call the 0 position. If the ball lands -- at the
end of the trial -- directly opposite, for example, that is 6 o'clock.
Sometimes, when the ball is just about to fall into its final position, The ball
bounces to another position. This is because it can hit the partition between
pockets, or it can hit the metal deflectors (we sometimes call them "diamonds").
This too, causes skip or bounce. If the ball does not skip, then the "landed"
position is the same as the "would-have-landed" position. If the ball skips, or
bounces, then the clock position it was aiming towards is the would-have-landed
if no bounce position. This, of course is only important during the
data-collecting phase, before we try to predict. The prediction phase happens
later, after the data have been analyzed with curve-fitting. The curve fitting
gives the parameters, A, B, and C, for the second-degree polynomial Ax^2+Bx+C.
This of course, is a parabola. This we chose from ASSUMPTION -- it is the
equation for a falling body. The ball, too is a falling body, much like cannon
ball, or an earth satellite. When we plot the data, we find that a second-degree
polynomial is a good fit. For wheels that are not level, I think now that a
third-degree polynomial might be better. We did not try that. It's better to not
play at a wheel that is not good.
In the playing phase -- the predicting phase, we ignore the bounce, or skip. On my web page I have a graph that shows
this 'skip'. Skip seems to be
distributed as a Gaussian curve, bell shaped, but I do not have much data to
know this for sure.
We estimate TTT -- Time To Touch down: the length of time the ball will remain
in orbit. We ignored the time or distance from the ball start falling into the
wheel and landing into the wheel. We considered the whole time the ball remains
in travel (after timer starts:) as TTT. The motion of the ball for the last few
milliseconds before it lands, or touches down, Is included in TTT.
I think I understand your telephone connection -- Cellular telephone?
We considered and investigated some sort of vibrating motor (actually, small
relays), but thought that it would be difficult to interpret the output. The
computer to voice-synthesizer connection was not easy to figure out, or to make,
but my partner (the smarter one) could make it work. By the way, I have been
sending copies of our communication to him, and you can ask him questions any
time.)
I'm sure your formulas are good -- I did not take the time to think through
them. Because of the pi, I assume you are using angular velocity to do the
computation.
Please look at the description of our data on our web page. If it is clear and
you can understand and make use of it, I will be happy to send it to you.
Do you have any pictures of your equipment?
-- NOWSCAPE
==================================
Sent: Thursday, October 28, 1999
Subject: Re: what's error?
And what the difference between "position in which the ball landed" and "The
winning number"?
I know that using of vibrating it's more difficult, but give more security, it's
most important.
I can't send photo of equipment now, because it's not ready, now in producing,
after it's ready (end of november), of course i will send it to You. The very
interesting thing in my project: all details I order anonymously in Internet.
And nobody knows, what's he doing: equipment one person, math other. I have idea
and people develop part of it for me. Now I want order neuroanalysis for your
data. If You send me more data, I send You result.
Michael.
Sent: Monday, March 11, 2002 4:11 PM
Subject: Re: what's error?
I'm finished my work, but I'm not trying it.
Why You ask.
Before trying I decide to solve One problem. What is error between ball clocking
in the middle of the spin and error when ball lands.
Estimation was too bad, to try it on practice. A little error in the middle of
the spin make a big error at its end.
Now it is very interesting for me how its work by You.
Michael
Sent: Sunday, August 06, 2000
Subject: roulette
hi,
Are you still beating roulette with your computer? Did you play in the US? The
device laws always frightened me; 10 years in Nevada. I always played abroad.
Now I just play by eye. Do you?
I like the infrared link. We used UHF. Bad idea; harmonic bleed through to
casino break room radio in London. Do you play the new wheels?
Rose
Sent: Saturday, June 02, 2001
Subject: Roulette Time Data
I'd be interested in getting a copy of your timing data on the roulette
wheel.
Also a copy of the algorithm/basic programs you used if possible.
As for what I plan on using the for.. Why to build my own roulette computer
of course. I just got Eudaemonic Pie yesterday and am about 20% thru it.
No mention of specific algorithms, but it's a good read on how they went
about doing the research, etc.
Were you able to get anywhere close to the 44% they claimed in the book?
Neural nets would be an interesting way to tackle the problem.. My only
concern would be speed.. I'm not sure a portable computer would have the
speed necessary to run the neural net in real time. And retraining for a
new wheel would be extremely time consuming.
I'm look at this embedded controller as my base
http://www.uclinux.org/ucsimm
. It's prebuilt, small and has all the IO
pins you'd need. The C compiler support would make programming it fairly
simple as well.
I've also done some looking around for roulette wheel heads, decent ones are
fairly expensive it seems.. Do you still have the wheel in your pictures for
sale?
-Domar
==================================
The algorithm, of course was to time the ball and the rotor, and to predict
where the rotor would be when the ball drops out of orbit. We don't have the
code anymore. Out implementation did not get better than chance... we did not
refine the work and we quit. In my opinion, neural net prediction would be quite
fast; It's the training of the net that takes time. Yes we still have the wheel;
it's available, I'll put new pictures of it on the net soon.
Attached is our data, good luck!
Sent: Sunday, August 26, 2001
Subject: roulette
i shelved a very similar roulette project to yours about 10 years ago. why did
you stop work on it ? is there some fundamental hole in the theory?
i started working on the problem while i was dealing blackjack in a local casino
to put me through university. i was doing astrophysics and applied maths - you
can see how i had the idea. i got very similar results to you.
i recently dusted off my notes and bought some nice toys. you seem to be the
only people ive heared of in the past 0 years who know what they are talking
about.
are you still working in it?, using it already? id like to talk
David Griffiths
Sent: Sunday, January 14, 2001
Subject: Raw data request
Hi
I am looking for raw data preferably ball and rotor timing for between 200
to preferably 500 spins. We are testing a Neural Net prediction method and a
single board computer. Our casinos here in Africa use single 0 wheels. But
obtaining raw data from them over a long period of time is difficult because
of the extremely high minimum bets and a slow "European" playing style which
results in relatively few spins per hour.
Thank You
Regards
Martin
Sent: Thursday, September 27, 2001
Subject: Info re data processing
Hi
We spoke earlier this year regarding the roulette data and it's analysis.
We have had success using a probabilistic Neural Net in predicting the ball
drop point and time.
I am currently working on the rotor prediction setup. I see you mentioned
the deceleration appears to be almost Linear. Since we have no roulette
wheel handy (Possession of them being illegal here) I'm afraid I have to ask
you about this. A number of articles I have read including Thorp's say it
should simply be regarded as a constant or at best approached with a
standard linear equation for angular deceleration.
Vf = Vi + at
I personally would have thought it was exponential.
Vf = Vi exp (-t/b)
Did you guys find curve fitting necessary for the wheelheads?
I would be greatly appreciative of your comments on this.
Martin
==================================
We timed a rotor-decay once for a period maybe 5-7 minutes, as I recall and,
and found the decay to be an almost straight line. I can't remember how
that was implemented in the algorithm, but I suppose, yes, we used
linearity.
I'm interested in your success. Do you have any output to show us?
Good luck,
-- NOWSCAPE
==================================
(Martin)
Just got back from Turkey and have not had a chance to respond until now. We
used an exponential curve to fit the wheel decay, which was quite predictable.
As I recall the decay was about 10ms per revolution, and varied a little from
one wheel to another.
Sent: Tuesday, October 30, 2001
Subject: RE: Roulette prediction
Sir,
I have a few questions about roulette prediction.
1. I would imagine it is possible to predict the quarter of the wheel the
ball will land in and nothing more. Is this true?
2. When using your system, do you have 2 timers - one for decaying ball
time in one direction and one for decaying wheel time in the other
direction. If so, how do you relate these two times to predict the ball
landing point?
-- Yes, there were two timers, Simon.
3. Could one simply time the period of the ball passing zero (maybe twice)
to predict the landing slot? Even if the wheel was travelling at a
different speed each time and the ball also, the relative speed would be
proportional, meaning similar landing positions.
--Regarding prediction of the landing place and thus the landing TIME of the
ball: yes one can time one or more zero-passing events (passing some fixed
reference point on the stationary rim of the roulette wheel); of course the more
time measurements are made the better the prediction will be. The speed of the
ball, of course is independent of the speed of the wheel.
--Regarding the timing of RELATIVE positions of the ball and the wheel -- for
example considering a time interval when the ball passes the zero of a rotating
wheel (not some outside, fixed reference point): I believe it's possible to make
a valid prediction this way, but my friend and collaborator on this project
thinks that making such a prediction with only this relative time information is
not possible.
I know these questions are very simplistic, but I would appreciate some kind
of feedback. Please excuse my lack of intelligence in this matter.
Regards,
Simon
Sent: Tuesday, November 06, 2001
Subject: Ceiling stain
If you could briefly explain the reaction that precipitated the big bang of
your electrolysis of heavy water please?
Didn't the Utah boys burn a hole in
their table with the same experiment? Does the electrode load up with
hydrogen and then release it rapidly or where is the burst of heat coming
from?
Thanks from a curious novice,
Mark
|
==================================
Sent: Tuesday, November 06, 2001
Subject: RE: Ceiling stain
Importance: High
Mark,
We didn't have a Big Bang, except when the deuterium exploded once. Unlike other
experimenters, we tried to reclaim the heavy water by igniting the deuterium
with sparks. As I remember, yes, in the Pons and Fleishman experiment, a hole
was burned trough a sink, or wall, or something. The electrode loads up slowly,
over a period of a week or so. I don't know the rate at which hydrogen is
released. It is not released suddenly.
-- NOWSCAPE
Hello, I stumbled across your web page when I was
researching for a
mathematical modeling project I am conduction. I am a graduate student and
part of the requirement for a class of mine is to model a mechanical
device.
I chose the roulette table but I am having a hard time gathering
information.
I was hoping you could supply some info like ball dimensions, wheel
dimensions, materials, ball release speed, wheel speed, etc.
Thanks,
LETTER FOR ROSE:
ROSE
Dear Rose:
I have read your letter posted into the web "Roulette Prediction", and I'm
faster interested in your work.
You can to help me to understand how is your "modus operandi" for visual
prediction?.
I would to appreciate any info about prediction formulas.
Doesn't abound too much information to the respect. Neither into Internet
nor between the books. I only have information about the Laurance Scott
system.
We can to exchange information about the subject.
Doesn't matter if you can reply me with delay, even if have passed moths or
years, because my interest is great about this field.
I hope your nice collaboration. Thank you very much in advance.
I remain.
Pere
-Spain (Europe)-
LETTER FOR DAVE:
DAVE
Dear Dave:
I have read your letter posted into the web "Nowscape-Roulette Prediction",
and I'm faster interested in your work.
You can to help me to understand how you created your tables of prediction?.
I would to appreciate any info about prediction formulas.
I knowledge any mechanism for visual prediction, but not the same translated
into readable tables and a simplified formula as you are saying.
We can to exchange information about the subject.
Doesn't matter if you can reply me with delay, even if have passed moths or
years, because my interest is great about this field.
I hope your nice collaboration. Thank you very much in advance.
I remain.
Pere
-Spain (Europe)-
E-MAIL FROM DAVE:
Sent: Monday, January 26, 1998
Subject: Re: Raw Data - Roulette3
In a message dated 98-01-26,
you write:
"<<
I have no idea how you were able to predict roulette, even if you can count
accurately.
>>"
"Well - given the same equations for a computers predictions, if you can
determine the ball time, the rotor time, and the relative relation between the
two - and you have the calibration coefficients for the wheel, you can predict
the outcome. Same as with a computer."
"My first system only required the rotor and ball times, as the ball time began
at the same time the rotor time ended. I created lookup tables for the
predicted pockets. They had ball times vertically and rotor times
horizontally. I was able to reduce the table down to having to memorize 24
4-digit numbers. I had to decode the numbers on the fly, but that wasn't too
bad. The problems were two-fold. First, I still needed an accurate
calibration of the wheel - thus I needed a computer. Second, I had to
rememorize the table for each new wheel. This was tough."
"I have a much better approach at this point, but am not inclined to discuss
it. Suffice it to say, it eliminates the need for a computer, and reduces the
memory work involved."
"Regards
Dave"
E-MAIL FROM ROSE:
Sent: Sunday, August 06, 2000
Subject: roulette
hi,
Are you still beating roulette with your computer? Did you play in the US?
The device laws always frightened me; 10 years in Nevada. I always played
abroad. Now I just play by eye. Do you?
I like the infrared link. We used UHF. Bad idea; harmonic bleed through to
casino break room radio in London. Do you play the new wheels?
Rose
Subject: CONTACT TO PREDICTION ROULETTE
Date: Thu, 17 Nov 2005
Dear Director of Nowscape:
I desire to e-mail the Sir Dave and Madam Rose, what have posted e-mails into
your Forum.
Please, It's important for to exchange information with people with similar
interest.
Thank you very much for your nice and irreplaceable help.
Pere
-Spain (Europe)-
Subject: Re: CONTACT TO PREDICTION ROULETTE
Dear Pere --
I don't keep a record of the people who email about roulette. This is for
security and privacy. So, sorry, I cannot help you.
-- NOWSCAPE