On Feb 19, 2006, at 8:11 AM, Jed Rothwell wrote:
Horace Heffner wrote:
Here's a thought about use of pseudo-random number generators in
gambling machines. The sequence of random numbers is
predetermined once a seed is chosen. It is thus possible to
select a seed in advance which guarantees no payoffs until a
certain point. In other words, it can be possible to rig the
payoffs so there is no risk at all to the machine owner provided
he resets his seed value periodically.
These machines are very carefully monitored by experts who are
state employees. The machines have to pay out a certain percent and
they have to keep a record. If a given machine did not pay out
regularly, this anomalous performance would be spotted. The
software in them is carefully vetted.
A winning pay-out is *expected*. Keeping records thus won't show a
thing, because the expected is always happening.
See <http://mtaonline.net/~hheffner/Gambling.pdf>
The advantage to a pseudo-random number generator, especially one
with a small cycle, like those which are commonly in use on windows
programs hindered by 32 bit registers, is that it can be pre-tested
to *guarantee* no run of bad luck for the machine owner. That is
even true regardless the starting seed because the entire sequence
can be tested. There are RNG algorithms that can use 32 bit
registers to generate sequences of length near 2^32 instead of 2^16,
but for some reason they don't seem to be in common use. I don't
know what kind of RNG is in common use in gambling machines however.
I do not know the details, but if they do use pseudo-random numbers
I am sure they are seeded by the real time clock, which is the
standard procedure to ensure "real" randomness starting midway
through the sequence, obviously. I think the seed is based on the
current millisecond, so that the actual time of day cannot affect
the performance.
No, a timer based seed is only selected once at power on, if then.
If there is an external source of random numbers then the additional
transformation provided by a pseudo-random number generator provides
no additional randomness. I've read that use of this fact was made
by some enterprising individuals that bought a video poker machine
and reverse assembled the ROM (though I have no reference so it could
be an urban legend?) They found out what RNG was used and then
figured out how to determine where in the cycle the RNG was by
observing play of a normal machine in a casino. From that point on
they could predict every hand.
Slot machine software and hardware is much better regulated and
much more tightly controlled than voting machine software. Voting
machines are so badly designed, and the software is so sloppy and
amateur they are a joke. They are standing invitation to any half-
wit hacker to steal an election. They are so bad that during a
test, computer experts at Johns Hopkins was able to break into the
machine over a phone line, alter the results of a dummy election,
and erase all trace of their activities . . . in about five
minutes. See, for example:
http://avirubin.com/vote/response.html
- Jed
I could not agree more. Our young men and women bleed and die for
freedom and then the press and government officials frivol it away
under the pressure of vendors hocking their wares, and in the cause
of expediency, looking good, meeting news deadlines and party
celebration schedules. There is no substitute for physical ballots
that can be verified by the voter before going into the ballot box,
and, if need be, later manually recounted by election workers
overseen by neutral observers. I think most of the blame for this
problem lies with reporters who are either too stupid or too lazy to
understand these issues. I can tell you from first hand experience
that few people understand these issues, including almost all
election officials, but it is the reporters who are *responsible* to
make the effort to understand and report these issues, and most of
all, to not be the principle part of the problem. It is a sacred duty.
Horace Heffner
