thornahawk

as for the algorithm, i guess i'll use either

 E+round((3^16-10)fPart((2^16-1+E+round(fPart(((331((Sum of Original String+E))/2)/(2^32-1)))(2^32-1),0)2-E)/(3^16-10)),0→F 

or

 E+round((3^16+26)fPart((2^16+1+E+round(fPart(((331((Sum of Original String+E))/2)/(2^16+1)))(2^16+1),0)2-E)/(3^16+26)),0→F 

but i dont know which one to chose. both are good, from what i have seen

Some advanced machines can record static noise or radiation and use those for the purposes of generating random numbers. Another method is the timing of user input. Yet another approach (something that I have actually seen used in a computer program before) is where a text box appears and the user is instructed to bang on the keyboard for a few seconds. If that's not a successful RNG, then I don't know what is. (Kidding, of course.)

so any ideas on how to write the actual equation for a good rng?

To reiterate… an RNG with the characteristics you say are not practical; the point of having a seed is to be able to generate the same sequence associated with that particular seed across different platforms.

"LCGs should not be used for applications where high-quality randomness is critical."

In any event, that LCG's are not "random enough" (more precisely, their periods are too short, among other weaknesses) are the reason why you have things like the "Mersenne twister" or "Blum-Blum-Shub"; go search for these topics on Google Scholar.

There's also a discussion of RNGs in Numerical Recipes (http://www.nrbook.com/ ), check it out.

thornahawk

wiki:

LCGs should not be used for applications where high-quality randomness is critical.

On a TI-84+, you can use the clock like so: startTmr→rand

Otherwise, a dirty trick on the 83+ is to seed rand or a function of your choosing with the user's reaction time:

The first line is the dummy getKey covered here (just above Error Conditions).

i know there are equations like

but what are good numbers for the constants for equations like this?

also, my main problem is that only one random number is needed everytime the progam is run

So, if there exists a function that produces a sequence with one input…

ƒ(x) = a_n

…yet goes on to produce another sequence from the same input…

ƒ(x) = b_n

…but you require that the two sequences be different…

a_n ≠ b_n

…then you can make the substitutions as follows…

a_n ≠ b_n

ƒ(x) ≠ b_n

ƒ(x) ≠ ƒ(x)

…showing that the function itself can't be in agreement with itself at any given instant.

What you are asking for is an arbitrary function. Or, something that will compute <insert sequence> given <insert function>. Unless you can get it any more well-defined than that, you're going to have a hard time programming it in.

Simpler answer: You pick one seed, you write it into your expression on paper, and then you solve it by hand. If you carry out the same procedure every time, then you should get the same answer repeatedly. Only when you try a new approach, whether it's changing the function or that pesky input, would the end result turn up as something else. This modification from the earlier test… it's called a "variable." If one number were chosen to replace it, and then all answers followed suit, then the initial input might be dubbed a "seed."

Alternate ending: You discover that by looking at your function weirdly enough, you are able to inexplicably force it to take on a new behavior from all of its previous trials. Alas, the logic is inescapable; the manner in which you looked at your function is also a variable – a veritable seed in the budding, and a "seed" in the dubbing.

Nutshell answer: Any expression with a constant form isn't going to trade results with the void. An effect requires a cause, and in the case of a RNG, that "push" is going to be the seed – that's what it's there for.

im trying to increase the randomness of my hash2x program but i cant think of any way to use any of the variables without compromising the entire algorithm

or can someone give me the actual formula for a random number generator (not just randInt())