Re: Simulating the rolling of dices to produce truly random numbers?

[Dennis Lee Bieber <wlfraed@ix.netcom.com>]

On Wed, 10 Jan 2018 14:54:44 -0800 (PST), Mace Ayres <mace.ayres@gmail.com>
declaimed the following:

>On Wednesday, January 10, 2018 at 7:50:21 AM UTC-8, Mr. Man-wai Chang wrote:
>> Is it possible? Any existing algorithms or published papers?
>
>Are we assured that our human mind/intelligence can know if a number is truly random, or if a algorithm will generate a true random number? If algorithmic, then deterministic? Is the idea of a truly random number an idea, but not necessarily a number that can be proven to be random? How random is good enough?
>

	Pick an algorithm... Run many samples... Perform statistical analysis
between expected result vs actual...

	Many random number generators out there -- some have reasonable
randomness, some not... A generator that is using some processor clock may
not provide randomness if invoked in a loop as the loop is a fixed duration
and could keep reading the clock at a determined interval.

	Your subject says "dice" (note: "dice" IS the plural, not "dices" -- a
single one is a "die"). How many dice per "roll". How many sides per die?
(I bring this up as role playing games use 4-sided, 6-sided, 8-sided,
12-sided, and 20-sided [substituting for 10-sided by modulo], and some
novelty dice purport to be 100-sided (percentile dice -- more commonly done
by rolling 2 20-sided to provide 0-9/0-9).

	For a single die, random means one should have close to an even number
of occurences for each value. For a 6-sided die, 6000 rolls should produce
1000 occurences of each value -- since reality won't be that pure, you have
to analyze if the differences still constitute random.

	If rolling multiple dice at a time, the results should fit a bell-curve
-- and again you should have a "truth" value for the bell curve against
which to test the actual results.

	No test will state that something is truly random -- statistical tests
rely on probability that the result may not be random... IE: is the
difference between "truth" and experiment within 95% of the test parameter.

>If I choose n among the set 1..500, based on my own sense of randomness, is that random?

	500 samples is rather small... More samples means statistical
differences will be more precise.
-- 
	Wulfraed                 Dennis Lee Bieber         AF6VN
    wlfraed@ix.netcom.com    HTTP://wlfraed.home.netcom.com/