Re: random number generator
Hi Jules
----------
> From: Jules Davy <guest at RapidEuphoria.com>
> To: EUforum at topica.com
> Subject: Re: random number generator
> Sent: 9 jul 2004 y. 11:44
>
> posted by: Jules Davy <jaydavy at onetel.com>
>
> Pete Lomax wrote:
> >
> > On Thu, 08 Jul 2004 09:12:35 -0700, Jules Davy
> > <guest at RapidEuphoria.com> wrote:
> >
> > >I'm writing a program to test roulette systems (million+ "spins") and
> > >was wondering whether eu's rand() function is adequate for this or
should I use
> > >a more advanced RNG such as the mersenne twister in the archive?
> > Erm,
> > I can safely say that I learnt almost everything I know about roulette
> > from watching James Bond movies. Not quite sure how you expect anyone
> > to guess whether something is adequate for a program which has not yet
> > been written. (sorry if that sounds a bit harsh, it wasn't meant to
> > be) It may be that rand() is ok, afaik, all reports so far indicate it
> > would be fine for a million+ ints in the range 0..33[?].
> >
> > You may as well try that mersenne thing and/or the new getRandInt() in
> > win32lib, and/or any other you can find, in the fullness of time, but
> > it sounds like you should go with rand() for now.
> >
> > Regards,
> > Pete
> >
> >
> > Hi Pete,
> Thanks for you reply. It did come over as a little harsh - I know nothing
of how RNG's work or probability distributions, so even after writing
> the program I would be none the wiser as to whether rand()is
> adequate or not. I just want to use the best RNG available
> to simulate a roulette wheel over an extended period.
This task is very complicated. Think please, you have to
simulate the microdust's action, the tabacco smoke's
action, the air's temperature, damp and velocity,
the friction factor's changeability, the cleanliness
of casino man's fingers etc etc etc ... And do not
forget - there are many top secrets in this business.
These secrets are really top and there is no RNG
in the world to simulate them.
> The mersenne twister in the archive is described as "very good",
> implying that perhaps rand() is not so good? but then,
> it all depends on what you want to use it for, hence my question.
The Monte Carlo method requires the careful testing of all
RNGs used in your task. Say, the factors above (air, dust ...)
have *different* not flat probability distributions.
Then you must to build all these special RNGs from the flat
rand() or mersenne_twister() and have to include them
in your task after special statistic testing only.
> I'll go ahead and use rand() as it's the simplest option.
Yes, it is simplest, it is flat and may be the
basic RNG for all the special ones.
> cheers,
> Jules
Good Luck, but I do not think the EU or another
model can help in casino. Casino wins, I think.
Regards,
Igor Kachan
kinz at peterlink.ru
|
Not Categorized, Please Help
|
|
