Re: Software roulette statistics

Gentlemen,

Many thanks for this.....I guess what you're saying is that I probably
hit upon a lucky streak, take the money and run........(just in time
for Christmas).

I shall leave you to your world of statistics and return to my world of
engineering, slightly wiser and thankful for my lucky streak, and still
wondering if someone, somewhere really ever did put that 'force the
zero' into that software...

Regards,


Cath


Gordon Sande wrote:
On 2006-11-28 15:39:37 -0400, "Reef Fish"
<large_nassua_grouper@xxxxxxxxx> said:


David A. Heiser wrote:

< snip lengthy beginning portions of thread>
<cunchy64@xxxxxxxxxxx> wrote in message
news:1164709891.405678.305250@xxxxxxxxxxxxxxxxxxxxxxxxxxxxxxx
Reef Fish Bob,

Thanks for your comments, all of which are perfectly understandable and
understood.
Cath

My belated "You're welcome, Cath" since I had nothing else to add.


++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++++
Software gambling routines depend on random number generators (RNGs). The
majority are on the single "seed" type in which the next RNG comes from
computer (digital) operations on the previous number. The earlier methods
were of the Linear Congruential form (Knuth), that generated succeeding
sequences of random numbers. These have been shown to be inherently
defective, and as Marsaglia found out generate distinct patterns and
repeating sequences. A spectral test is a way to find out about these
patterns. These patterns may be evident to a player, and the "owner" may be
oblivious of them. Some of the newer RNGs do not have this characteristic.

Marsaglia's "Random numbers fall mainly in the plain" about the
congruential
type of generator lying on "lattice" points in TWO or higher dimentions
made
it "defective" for simulating points in HIGH dimensions, but not
necessarily
defective for one-dimensional problems such as generating U(0,1) random
numbers to simulate univariate distributions.

For that reason, there's no reason to suspect (a) the gambling machine
designers are NOT aware of the RANDU or RNG defects that's well-
known since 1968 <date of Marsaglias' paper> or know about "modern"
generators, and (b) that the defective RNGs adversely affect the
software roulette machines.

-- Reef Fish Bob.

The cycle length of the circa 1968 common generators was short, 2^31 or so,
because they were single computer words and used multipliers of simple
bit structure so tended to exhibit sub-cycle behavior. "Mainly in the
planes ..." in the catchy and apt phrase. This does not generalize to
all multipliers of linear confuences being "defective" in two dimensions.

If you do uniform k-dimensional binning at fine enough resolution then
all cyclical numerical sequences (i.e pseudo random number generators)
will fail to be uniform. The uppers bounds for this are classical geometry
of numbers results and are the basis for the "spectral test" for random
number generators. Multipliers which are plausibly not "defective" can be
found but this assumes that the limitations implied by the geometry
of numbers and cycle length are understood.

In practice one should use multiple precision (multiple computer words)
sequences with recurrances which may be polynomial rather than just linear.
Knuth is a good introduction. A widely available example of the current
technology is the "Mersenne Twistor".

The whole basis is for the "owner" to get as cheap a "machine" as possible
and to "pay" for software that does not give a fair play. The display and
packaging is much more important than a fair play. With computer software
driven gambling, there is no way that the player can find out if it is a
fair play of a biased play. One cannot in a finite amount of $ obtain a
large enough of a data base to determine if a fair or foul play is going on.
We are talking about 30 million numbers for a test series, and even these
have a "variability", so it is hard to arrive at any definite conclusions..
Obtaining the software instructions is not a guarantee that what you play is
specifically from that software routine. Its all a "pea in the shell" game,
and the famous quote "a sucker is born every minute" and "it's your buck
(not mine)" still holds. Just as in a carnival, lights, colors, noise,
music, nudes etc.bring in the suckers.

Nothing has changed, only that the process has gotten "high tech".

The Indian tribes here have found out that gambling brings in so much more
money "profits" than any other legitimate business, they will do anything to
get them built in or near our towns and cities.

David Heiser

.

Relevant Pages

  • Re: Software roulette statistics
    ... Software gambling routines depend on random number generators. ... designers are NOT aware of the RANDU or RNG defects that's well- ... all multipliers of linear confuences being "defective" in two dimensions. ... and to "pay" for software that does not give a fair play. ...
    (sci.stat.math)
  • Re: Software roulette statistics
    ... Software gambling routines depend on random number generators (RNGs). ... designers are NOT aware of the RANDU or RNG defects that's well- ... and to "pay" for software that does not give a fair play. ...
    (sci.stat.math)
  • Re: Software roulette statistics
    ... Software gambling routines depend on random number generators. ... designers are NOT aware of the RANDU or RNG defects that's well- ... and to "pay" for software that does not give a fair play. ...
    (sci.stat.math)
  • Re: Software roulette statistics
    ... Software gambling routines depend on random number generators. ... all multipliers of linear confuences being "defective" in two dimensions. ... period of the LCG is AT MOST M, ... better and faster than most other LCGs but suffer the same DEFECTS ...
    (sci.stat.math)
  • Re: Software roulette statistics
    ... The "old" prng used in SPSS, IMSL, and more recently SAS was ok for case selection in survey sampling etc. ... click (left pane) ... all multipliers of linear confuences being "defective" in two dimensions. ... all cyclical numerical sequences (i.e pseudo random number generators) ...
    (sci.stat.math)