[Coco] Is the RND command really generating a random number?

Mathieu Bouchard matju at artengine.ca
Fri May 19 10:18:51 EDT 2017


Somehow, my original reply at 10:00 didn't go through the list. It seems to have 
been dropped. For some reason, my mail client put the coco list in "Cc" instead 
of "To", and this might be the reason the mail has been blocked by Mailman, 
except I don't remember Mailman ever doing that with mailing lists I managed in 
the past...???


Le 2017-05-19 à 10:13:00, Mathieu Bouchard a écrit :

> Actually, I'm not sure that my experiment was really about QuickBASIC's 
> RND(). Now I remember that RND() was annoyingly slow and I tried writing my 
> own generator. The simplest textbook examples of such generators caused the 
> visual patterns I described, and I think RND() was already better than that. 
> I wrote that about 22 years ago and I haven't retried since.
>
>
> Le 2017-05-19 à 10:00:00, Mathieu Bouchard a écrit :
>
>> "Truly random" is hard to define, if not hardly definable. It's at least as 
>> hard to test. What we dare call "truly random" is merely what we have no 
>> clue about. Of course, RNGs have various degrees of ability to "appear 
>> random" (whatever that means). In the case of BASIC, even if you seed the 
>> RNG with a realtime clock, the algorithm can still suck in different ways, 
>> such as how likely it is to produce a number after another given one. I 
>> remember an experiment I made in QuickBASIC with RND() producing numbers 
>> from 0 to 65535 inclusively. I divided them by 256, plotted the floor of 
>> the quotient as y, and the remainder as x. Various temporary visual 
>> patterns appeared (such as dotted diagonal lines). This 256x256 square also 
>> got filled completely much quickly than according to probability theory. 
>> With slight variants of the same (linear-congruential) algorithm, 
>> theoretical randomness can be faked a lot better than RND(), but that's not 
>> what Microsoft chose to use.
>> 
>> 
>> Le 2017-05-18 à 19:44:00, Dave Philipsen a écrit :
>> 
>>> Very few things in the computer world are truly random.  A better term 
>>> might be 'pseudo-random'.  Think about it.  In theory if you turn on two 
>>> identical CoCos at the same time and ask for a random number at the same 
>>> time they should both provide the same random number.  So, in reality, 
>>> it's not random.
>>> 
>>> Dave
>>> 
>>> On 5/18/2017 7:39 PM, Rietveld Rietveld wrote:
>>>> I used the RND command to generate a random game play option in the 
>>> cocoflash, but it doesn't seem to be very random. Some titles come up more 
>>> than the laws of probability would suggest is possible.   instead I am now 
>>> using a small math equation that uses a PEEK of the clock value.   This 
>>> has resulted in a far more random looking selection
>> 
>> ______________________________________________________________________
>> | Mathieu BOUCHARD --- tél: 514.623.3801, 514.383.3801 --- Montréal, QC
>> 
>
> ______________________________________________________________________
> | Mathieu BOUCHARD --- tél: 514.623.3801, 514.383.3801 --- Montréal, QC
>

  ______________________________________________________________________
| Mathieu BOUCHARD --- tél: 514.623.3801, 514.383.3801 --- Montréal, QC


More information about the Coco mailing list