RANDOM.THD  --- Copyright 1988 by Phil Wheeler

An original compilation of Compuserve Model 100 Forum messages for use by
Forum members only.

These messages discuss "a RANDOMizing subroutine to READ F$, that will give a
different word each time", an interesting programming problem introduced in
the first message.  The resulting replies give some interesting insights in
to potential and valid solutions to this interesting problem.

Message range: 173577 to 173603
Dates: 8/28/88 to 8/28/88


Sb: #RANDOMizing routine!
Fm: Norm Leveillee 72327,1025
To: Anyone!

TO ANYONE! I need a RANDOMizing subroutine to READ F$, that will give a
different word each time. I have a routine now, but it repeats some of the
words at least twice during the looping. Can you help? Merci bien. Norm


Fm: Bill Brandon [DPTRAIN] 76701,256
To: Norm Leveillee 72327,1025

Norm, why don't you either a) delete a word from the universe when it's used,
or b) have the program test each word as selected: if it's already been used,
then go back and select another. That's as good as I can do, given the limited
description of your objective.

By the way, in a truly random selection from a limited population, where
selection does not remove an item from future consideration, you are very
likely to see the same item pop up more than once. That's how the folks who
write programs to improve your chances with state lotteries design their
programs, isn't it?


Fm: Tony Anderson 76703,4062
To: Norm Leveillee 72327,1025

There is no "randomizing routine" that will not, at some point or other, return
a number that has previously been used, or repeat a certain group of numbers in
the same sequence. It has to do with the "random" element, where it is entirely
possible for a number to repeat, or even a group of numbers to repeat.  The
classic mathematical example is a pair of dice; given only 11 numbers, but in
an ideal distribution of 36 possible combinations.  The odds are a little over
3 to 1 that there will be repeats.  That's why you hear about gamblers making
several "passes" in a row (winning on the number 7 or 11 when they are thrown
as the first try in a new "game").  But that may be a little too much for you
to work with.

They have always had similar problems with other computers, so no routine is
perfect.  The way they solved it in some of the mainframes, where a
non-repeating series of numbers was required, was to construct an array giving
the number of instances you would be using, sort of like a table, and as each
of the possible numbers was used, to "check it off" in the table.  If a number
repeated, it was skipped, and the random generator called again.  The bug in
this approach is that the more numbers that were used from the table, the
longer it takes to generate a new "unique" number.  The answer to that was to
select a stop point, and when a certain percentage of the numbers available had
been used, to stop, and start over.  Selecting 50%, as an example, if you need
15 unique numbers, then the table would start with 30 numbers, and when 15 had
been used, stop.  The smaller the percentage selected, the less likely for
delays based on trying to find a new unique number.

Some card game simulations work by starting with a table of 52 possible
numbers, representing an entire deck of cards, and as each "card" is used, the
number is checked off.  When the "deck" is half used, they are "shuffled",
(table is cleared) and the game starts over with a "fresh deck" (new table).

So, I would think the answer in your case would be to have a pool of words
available much larger than you actually plan to use, and provide a table to
indicate when a specific word had been used, allowing you to select another
word.  Personally, I'd start with a word file three times what I needed... for
example, if you want to select 20 words, my master word list would be 60 words.
Although you give no real details of what you're doing, if the project was
extensive enough, you might consider using the spelling checker dictionary that
is available in Lib. 9.  It has 20,000 words in alphabetic order, and that
would make a hell of a selection pool, randomly selected, if you were working
on a spelling-type program.

I believe there are also 100 and 1500 word lists of the most often used words
in English.

I assume you know how to construct an array table of x elements, and mark each
element with a 1 if the matching word is used (the array is set to zeros when
dimensioned).  A simple BASIC IF statement can test to see if an element has
been used, in True/False reading of the zero or one.  This might be one method
of doing it, using a 100 element table, and selecting 20 words from the list.

100 DIM N(100)
110 DATA Blah, Blah, Blah, etc... (100 words)
.  (more words)
.  (more words)
230 FOR X=1TO20
240 (Random Generator, Return "Y")
250 IF N(Y) THEN 240 ELSE N(Y)=1
260 RESTORE:FOR Z=1TOY
270 READ F$
280 NEXT Z
290 (Use F$ word)
300 NEXT X


Fm: Norm Leveillee 72327,1025
To: Bill Brandon [DPTRAIN] 76701,256

DEAR BILL, THANKS FOR YOUR HELP. AMITIES, NORM


Fm: Norm Leveillee 72327,1025
To: Tony Anderson 76703,4062

Dear Tony, Your response was very helpful. I will work onyour suggestions.
Again, merci. Norm