Randomizer not random

[jon93]

Hello everyone,
I am playing with a meaningless simple program in which a ball bounces randomly inside a box trying to find the quickest way to a target.
I have found something strange about "randomize". 
I want to start ball bouncing sequence with different randomize seeds and then select which one works best.
So at the start of each run I increase the randomize seed.
But when I increase the randomize seed the pattern of bouncing of the ball in the box is almost the same as the previous one.
It drove me a little crazy for several hours. I assumed I was making a mistake.
But now I see that if the randomize seed is increased by only a very small amount like .0000001 the numbers generated by one sequence of rnd are almost the same as the previous one. 

If the rnd sequence runs through many iterations (5000) the two sequences are different.

Have others noticed this before?

Jon, in Ottawa, Canada.

RandomizerSeed = .1
HowMuchToIncreaseRandimizerSeed = .0000001  ' if this is much bigger the sameness effect does not happen.
for i = 1 to 5
   RandomizerSeed = RandomizerSeed + HowMuchToIncreaseRandimizerSeed
   randomize RandomizerSeed
   print
   'for j = 1 to 500 'run this loop to go further along the rnd sequence; different sequences of rnd then become different.
   'dummy = rnd(1)
   'next
   print "RandomiserSeed", RandomizerSeed
   for k = 1 to 5
       print k, rnd(1)
   next k
next i


[plus]

Plenty of complaints about JB's random number generator but what it has to do with bouncing balls is beyond me!

You can look into building your own random number generator but first are your expectations realistic and properly reasoned?

Randomize seed  ' will produce same sequence of numbers from start to however long it goes

If the sequence values aren't random enough, store them all in a container and shuffle it.

'repeatable sequences
seed1 = .5065
seed2 = .7077
randomize seed1
Dim rn(5000)
for i = 0 to 5000
rn(i) = rnd(1)
next
randomize seed2
for i = 5000 to 1 step -1
r = int(rnd(1)*(i+1))
t = rn(i)
rn(i) = rn(r)
rn(r) = t
next

for i = 1 to 100  ' go ahead and write the numbers down and run this again to compare
print rn(i)
next


[Rod]

I am not sure how there can be a “best random sequence” but interested to see what you mean. Yes there are issues with Just BASICs pseudo random number generator but no more so than any other. Frequent restarting and small changes to the seed have been mentioned before as well as a slight bias. There are discussions to search out

I think b+ has a solution for the starting problem but then the sequence can never be repeated.

[plus]

I think b+ has a solution for the starting problem but then the sequence can never be repeated.

Dang! I was hoping to show you can repeat a double seeded randomized shuffle.

Some of my favorite works depend on repeatable "random" sequences because it saves so much time and space from having to store data.

[tsh73]

Here 7 set
of 10 random numbers,
starting from seeds 0.1+i*h with different steps,
from h=0.0000001 to h=0.1

Obviously it is pretty linear with small h.
But some decent h + discarding some spare RNDs makes you pretty random, I think.


nomainwin
K=7
N=10
seed=0.1
'h=0.0000001
h=0.1
open "test" for graphics_nsb_nf as #gr
#gr "trapclose [quit]"
#gr "down"

x0=10
y0=310
scale=300
#gr "\\";"    h=";h
#gr "size 2"
'#gr "home"
'#gr "line ";;" ";;" ";;" ";
#gr "line ";x0;" ";y0;" ";x0+scale;" ";y0
#gr "line ";x0;" ";y0;" ";x0;" ";y0-scale
#gr "north; turn -90"   'up


for k = 1 to K
   seed1=seed+k*h
   randomize seed1
   #gr "color ";word$("pink red brown yellow green cyan blue", k)

   for i = 1 to N
       hh=scale/N
       xx=x0+i*hh
       yy=y0-scale*rnd(0)
       #gr "line ";xx+3*k;" ";y0;" ";xx+3*k;" ";yy
   next
next

#gr "flush"
wait

[quit]
   close #gr
end



[jon93]

Thanks for shuffle idea and info
I was thinking of writing my own rnd generator using prime numbers .
Re "what it has to do with bouncing balls" you have to understand the utterly trivial, physically impossible and absurd nature of my ball box.
I write it for brain challenge. I wrote code for a ball bouncing in a box. boring. make the bounce random directions (physically absurd). Then see how long it takes to reach a target. Then select the best route. You will grasp the idea once you realise how pointless it is! If you try to find meaning or usefulness in it you will just be confused.

Jon

[tsh73]

Really, JB RND() usually good enough.
Just in case it is not - there are some random number generators at Rosetta code

rosettacode.org/wiki/Linear_congruential_generator#Liberty_BASIC