🚀 go-pugleaf

In article <1096976264.1684.1096967182@bayman.org>, david.williams@bayman.org (David Williams) wrote:
>-> > RND is not truly random, of course, but it's a VERY good simulation of 
>-> > randomness. Almost all programs that allegedly show RND to be 
>-> > non-random are in fact badly-written programs that really show nothing 
>-> > of the kind. 
> 
>-> Marsaglia's 1968 paper showed exactly the opposite, so I would be interested
> 
>-> to know which programs you are referring to. VB and QB have LCRNGs of 
>-> exactly the type Marsaglia examined - see 
>  
>I did say *almost* all. Yes, Certainly, anyone who knows how a 
>RND-generator works can write a program that shows it not to be truly 
>random. The little thing I wrote yesterday, re-seeding the generator 
>with the same number and showing it then produces the same RNDs is just 
>one example. 
>  
>But the vast majority of little BASIC programs that allegedly show RND 
>to be non-random have been written by amateurs who really don't know 
>what they are doing. For example, they re-seed the generator over and 
>over again from TIMER, in such a short interval of time that it has the 
>same value. 

I dudded myself in a similar way 30 years ago using FORTRAN on a DEC 
System-10.  I decided to print out a series of random number sets for 
use in the field when I escaped once more from academia, and when I 
came to use the sets a year or two later I discovered they were all 
the same sequence of numbers!  Another fine example of the 
consequences of ignoring the injunction to RTFM.  8-)

Another trap I caught myself with as a result of more confidence than 
competence was in randomising a set of 10 individuals of each of 30 
species into a fully random field design (i.e. no blocking).

I allocated each randomly chosen species/individual to a sequential 
field position and wondered why I ended up with a preponderance of one 
or two species at the end of the paddock!  In its simplest form, what 
I *should* have done was allocate a randomly chosen field position to 
each species/individual combination.

>(Anecdote coming up:) Many years ago, a friend of mine ran a small 
>Bulletin Board System. Just for fun, he incorporated a lottery game 
>into it. Every day, users "bet" on sequences of numbers. Each night, 
>the computer generated its own sequence, using RNDs, and then rewarded 
>whoever had got the most numbers right with an extended membership of 
>the BBS. 
>  
>There was just one problem. The machine kept coming up with the same 
>numbers! My friend, who had written the code himself, tested it many 
>times, and couldn't find anything wrong. After a *long* time, he 
>realized the problem. At the start of his code, he had written: 
>  
>X=RND(-TI) 
>  
>(This was in Commodore BASIC, in which TI was a timer that ran in 60ths 
>of a second, starting at midnight. The above line was the standard way 
>to "randomly" seed the RND generator.) 
>  
>But, when it was running automatically, the BBS computer was executing 
>the lottery code at exactly the same time, by its own clock, every 
>night. So TI always had the same value. That's why the numbers the 
>computer chose were always the same. 
>  
>Simply removing the re-seeding line and replacing it right at the 
>beginning of the program, where it was executed only once as the BBS 
>was set up, fixed the problem. 


Cheers, Phred.

-- 
ppnerkDELETE@THISyahoo.com.INVALID