the rnd() function

[gatsby]

Why is the final distribution always skewed towards lower values?  Is it my coding or is the RND() function faulty?

' A program to make lots of numbers between 0 and 9,

' sum them, and show the results

dim z(10)

for x=1 to 1000000
y=int(rnd(1)*10)
z(y)=z(y)+1
next x

for x=0 to 9
print x;"=";z(x)
next x

print:Print "The End"

[plus]

Yep rnd() is not the best. Over about 10000 not too bad. Lower numbers specially 0 do seem preferred.

Do you want some ideas for shuffling to make better?

[tsh73]

Yes RND function is known to be faulty (more correct word "biased").
Probably OK for action games and graphic demos but not fit for rigorous numerical simulation.

You can try to roll your own - but it will be slower
like this
rosettacode.org/wiki/Linear_congruential_generator#Liberty_BASIC

[tanger32au]

Plays = 0



[RunProcess]
print ""; Plays
Plays = Plays + 1
NumRMD1 = INT(RND(1) * 10000) + 1

open "Log.txt" for append as #OutP1
print #OutP1, NumRMD1
close #OutP1
if NumRMD1 < 4999 then BigNumber = BigNumber + 1
if NumRMD1 > 5000 then SmallNumber = SmallNumber + 1
if NumRMD1 < 1000 then TinyNumber = TinyNumber + 1


If Plays = 1000 then gosub [EndProcess] else gosub [RunProcess]

[EndProcess]
print "BigNumbers"; BigNumber
print "SmallNumbers"; SmallNumber
print "TinyNumbers"; TinyNumber
notice "All done"

[plus]

Here is a nice slow rand() function:


dim digits(9)
for i = 0 to 9 ' load
digits(i) = i
next

'call shuff
'for i = 0 to 9
'print i, digits(i)
'next

'for i = 1 to 10
'print rand(0)
'next


dim z(10)

for x=1 to 100000  ' <<<<<<<<< you might want to reduce to 10,000 or 1000 for sampling and save time
scan
y=int(rand(1)*10)
z(y)=z(y)+1
next x

for x=0 to 9
print x;"=";z(x)
next x

print:Print "The End"

function rand(dummy)
r = rnd(0) * 1000000
amt =  len(str$(r))
for i = 1 to amt
call shuff
b$ = b$ + str$(digits(5))
next
rand = val("." + b$)
end function


sub shuff
for i = 9 to 1 step -1
t = digits(i)
r = int(rnd(0)*(i+1))
digits(i) = digits(r)
digits(r) = t
next
end sub