Re: A83: Help (Random)

[Prev][Next][Index][Thread]

Re: A83: Help (Random)


Linus Akesson wrote:
> 
> On 22-May-98, John Kugelman wrote:
> 
> >Linus Akesson wrote:
> >>
> >> First of all, deterministically speaking, there are NO random numbers. If I
> >> asked you to give me a number, any number, then you are forced to come up
> >with
> >> a certain number, depending on what electrons are where in your brains.
> This
> >> also means that computers cannot produce random numbers either. However
> >> cryptic an algorithm you use to convert a seed number (f.ex. a timer) into
> a
> >> pseudo-random number, there will always be a mathematical formula (that
> >> probably can be simplified too).
> 
> >Of course quantum physics says that nothing can be predicted completely
> >accurately since particle motion is random.
> 
> Define random. Besides, I never said anything about predicting anything...

Random as in unable to be predicted accurately under any circumstance. 
Over all behavior can be, though.  Numbers humans pick tend to be very
NON-random, as we often have biases to certain numbers or ranges of
numbers, such as 42. ;-)

> >> What I am trying to say is that there's no use in writing complex
> >> random-number-generating routines, as the numbers they generate wouldn't be
> >> _really_ random anyway. Actually, adding 1 to the last "random" number is
> >> often sufficient! But then the user would spot the algorithm too easy. I'd
> >> recommend rotating the previous "random" number with some constant, and
> then
> >> adding another constant. For an 8 bit number:
> >>
> >> random:      ld   a,(prevrand)
> >>              rla
> >>              rla
> >>              rla
> >>              add  a,31h
> >>              ld   (prevrand),a
> >>              ret
> >>
> >> Want a 16-bit number? Call the function twice!
> 
> >Just because truly random numbers can't be generated doesn't mean you
> >can't try.  I would hate to see you write a gambling program, or perhaps
> >some kind of Monte Carlo simulation.  The randomness of a particular
> >generator can in fact be determined by many factors, including its
> >period, or the time it takes to repeat itself.  I would expect this type
> >of algorithm to have a very short period, probably less than 1000.  The
> >numbers won't be very "random" either.
> 
> Yeah, you're probably right. But in most situations it really IS enough just
> to add 1 to the previous random number.

If you want *different* numbers, then it's okay, but that's definitely
non-random.  Of course, the original request was for a magic 8-ball
program (isn't ASM overkill?), in which case I suspect it doesn't really
matter.

> >Here's a web site on one of the newest and best random number
> >generators, and its quite speedy, too.
> 
> >http://www.math.keio.ac.jp/~matumoto/emt.html
> 
> Regards, I'll check it.

-- 
John Kugelman.  kugelman@mnsinc.com

I believe we can change anything.
I believe in my dream.
    - Joe Satriani
References: