0

there are conformity tests to tell whether a particular RNG method, programmable or mechanical is close to so called true random numbers https://en.m.wikipedia.org/wiki/Statistical_randomness

it is based on the hypothesis that, true random numbers should not have any patterns and regularities. I guess in a short they are trying to ensure "equal distribution" in a large set of data but an homogeneous or equal distribution is another sort of regularity ( A pattern ), wouldnt that be a violation of original randomness theory?

I have made a stock chart (trading) using a random number function (javascript), it draws a chart very similar to natural stock market charts, I am wondering what implications would have when using this technique?

i am not worried about "true randomness" (if there is anything as such) but the conformity tests used to approve those algos maybe a problem, the chart is range bound, it has short trends, noises and everytime something new design when i refresh.

my code is like this:

initialize share price lets say 100.00, then run a loop 10k times, and call random function ( returns between 0 to 1 ) if the value is less than 0.5, i add 0.1 to original price or deduct 0.1, i have also tried randomizing 0.1

duckduckgo
  • 1,280
  • 1
  • 18
  • 32
  • 1
    there is something as "true randomness". Though a TRNG must be implemented as hardware. –  Feb 29 '16 at 12:27
  • interesting, is it using mechanicalness or electron partical movements to get random data? – duckduckgo Feb 29 '16 at 12:30
  • 2
    Apparently, random.org uses "atmospheric noise". – mostruash Feb 29 '16 at 12:34
  • There are quite a lot of implementations ranging from sth like 5$ up to whatever you're willing to pay. A pretty interesting and cheap concept is based on an image sensor placed in a light-dense can, which produces images consisting of noise that can be converted to a random number. Another approach uses radiation-emitting elements and a geiger-counter. Basically anything that behaves in a truly random way that can't be predicted can be used. –  Feb 29 '16 at 12:35
  • wow, do PCs have any chip for tbis purpose? also do random set from atomospheric noise can become benchmark for algo based RNG. – duckduckgo Feb 29 '16 at 12:38
  • AFAIK Intel already realized the approach with the radiating elements. IDK if the approach with the image-sensor has any use in economy yet. But in general TRNG are used rather seldom due to their expensiveness compared to simple PRNG. –  Feb 29 '16 at 12:47
  • @paul, these RNG you mentioned are indeed true random but i think one should be careful match a generator for specific use case, for example a light source method maybe good for simulating light source in a virtual reality game -- but these micro techniques may eliminate the chances of short time regularities which are present in some case -- for example stock chart. – duckduckgo Feb 29 '16 at 12:49
  • 1
    @AbhishekK I think you misunderstand the concept. The light source method doesn't follow any system. The basic concept is to simply use the fact that image sensors can't catch perfect black as such and instead produce an image that only consists of noise, which is converted into a bitstream. TRNGs don't have **any** patterns, if implemented properly. And your stock chart example doesn't have any regularities as well - at least none that would be observable without large mathematical effort. PRNGs like javascripts `random` try to get as close to real randomness as possible. –  Feb 29 '16 at 14:05

1 Answers1

1

"Equal distribution", which is in your case probably "uniform distribution" means that alghoritm is stateless and every single value has same probability to appear, therefore there is no regularity itself.

The real problem for you can be that uniform distribution is just one of many distribution, it is not the "right" one or the real "random" one, this is just one.

As far as I know, trading is not "stateless" - the selling or buying of some products changes the value of that product, which changes the way people buying it.

One way is using uniform distribution for randomize input "uniformly" and then use some kind of weights to decide, what happens next (i.e. if product goes more and more expensive, the higher number (i.e. 0,8 and above) is needed to buy product)

The other problem is about "real randomness" you mentioned... it can be problem for some extremely sensitive data, otherwise the current random methods are "random" enough for most purposes.

libik
  • 22,239
  • 9
  • 44
  • 87
  • trading in long term is slightly stateless because there are trends/corrections – duckduckgo Feb 29 '16 at 12:52
  • In my opinion any large body based RNGs for example coin flip would have inherent biases always present-- so that bias is the part of it, and yes I agree about randomness for practical purposes -- though they are application specific. – duckduckgo Feb 29 '16 at 12:56