----------------------------------------------------------------------------- "Efficient and Portable Combined Random Number Generators", Pierre L'Ecuyer, Communications of the ACM, June, 1988 (v31 #6) "Random Number Generators: Good Ones Are Hard To Find", Park, Stephen K., and Miller, Keith W., Communications of the ACM, October, 1988. (see below) Conclusions :- horrible random (sic) number generators are used by people who don't know better, while very good ones take less than 20 lines of Pascal! Amoung the horrid generators are some that come with certain systems or are presented in textbooks! R. Salfi: *COMPSTAT 1974*. Vienna: 1974, pp 28 - 35... If the outcome of a random generator is solely based on the value of the previous outcome, like a linear congruential method, then the outcome of a shuffle depends on exactly three things: the shuffling algorithm, the input and the seed of the random generator. Hence, for a given list and a given algorithm, the outcome of the shuffle is purely based on the seed. Many modern computers have 32 bit random numbers, hence a 32 bit seed. Hence, there are at most 2^32 possible shuffles of a list, foreach of the possible algorithms. But for a list of n elements, there are n! possible permutations. Which means that a shuffle of a list of 13 elements will not generate certain permutations, as 13! > 2^32. ----------------------------------------------------------------------------- White Noise Random Number Generators... I believe it was in Andrew S. Tanenbaum's book "Computer Networks" that a great random number generator was proposed - amplify the kT noise in a resistor, sample the voltage every so often, and digitize it. You'll get a nice random sequence... --- David F. Skoll dfs@doe.carleton.ca I read some years ago about some people who tried to do this as part of some research project, and carefully analyzed the results. Their conclusion was that it is just about impossible to get flat distribution (white noise). All noise in nature is either colored one way or another, or else the noise is not, but the *measuring* instruments interfere with sampling in such a way as to give a colored distribution again. --- Doug Merritt doug@eris.berkeley.edu Incidentally, most noise-based generators can be used in combination with computed pseudo-random sequences, by XORing a sample from each source. It should be pretty hard to find patterns after that. --- Norman Diamond diamond@jit081.enet.dec.com Another way (which is equivalent) is to have a precise function generator generating sawtooth waves, which signal is then sampled (at arbitrary time) with an A/D converter. The randomness is introduced in the time domain of the sampling. --- John Porter 1292az@gmuvax2.gmu.edu ----------------------------------------------------------------------------- Simple (bad) Psuedo Random Number Generator (Sic) The low bit typically just toggles between calls. random() { seed = ( seed * mulitiplier + increment ) % modulus; return seed; } Table of Good values Multiplier Increment Modulus 25173 13849 65536 9301 49297 233280 A really bad one that was widely distibuted at one point (called RANDU) 65539 0 65536 ----------------------------------------------------------------------------- /* ** Dr. Park's algorithm published in the Oct. '88 ACM ** "Random Number Generators: Good Ones Are Hard To Find" ** His version available at ftp://cs.wm.edu/pub/rngs.tar ** This is called a Lehmer Generator */ static long Seed = 1; /* This is required to be 32 bits long */ long random() /* ** Random number between 0 and 2147483647 (2**31 - 1) inclusive **/ { a = 48271 m = 2147483647 r = 3399 /* r = m mod a */ q = 44488 /* q = m div a */ if ((Seed = Seed % q * a - Seed / q * r) < 0) Seed += m; return Seed - 1; } /* From xlockmore utils.c */ --- Or in pascal ---- function Random : real; { Initialize seed with 1..2147483646 } { maxint must be greater than or equal to 2**31 - 1 } const a = 16807; m = 2147483647; q = 12773; { q = m div a } r = 2836; { r = m mod a } var lo, hi, test : integer; begin hi := seed div q; lo := seed mod q; test := a * lo - r * hi; if test > 0 then seed := test else seed := test + m; Random := seed / m; { make floating point - important } end; --- Seeds --- Seeds you can use (calculate q and r as above) a = 48271 m = 2147483647 16807 2147483647 ----------------------------------------------------------------------------- Here is the Portable Combined Generator of L'Ecuyer for 32-bit computers. It has a period of roughly 8.12544 x 10**12. EG: two psudo-random generators working together. s1 : int := seed Function Uniform : real; { Initialize s1 to 1..2147483562; s2 to 1..2147483398 } var Z, k : integer; begin k := s1 div 53668; s1 := 40014 * (s1 - k * 53668) - k * 12211; if s1 < 0 then s1 := s1 + 2147483563; k := s2 div 52774; s2 := 40692 * (s2 - k * 52774) - k * 3791; if s2 < 0 then s2 := s2 + 2147483399; Z := s1 - s2; if Z < 1 then Z := Z + 2147483562; Uniform := Z * 4.656613e-10; end; ------------------------------------------------------------------------------- Cliff Random Generator http://mathworld.wolfram.com/CliffRandomNumberGenerator.html A random number generator produced by iterating new_X = | mod( 100 * ln X (mod 1) | for a seed X = 0.1 This simple generator passes the noise sphere test for randomness by showing no structure. However there are 'fixed points' Gawk Implementation BEGIN { _cliff_seed = 0.1 } function cliff_rand() { _cliff_seed = (100 * log(_cliff_seed)) % 1 if (_cliff_seed < 0) _cliff_seed = - _cliff_seed return _cliff_seed } Example usage... gawk --include cliff_rand.awk \ 'BEGIN { for(i=0;i<100;i++) print int(cliff_rand()*100); }' WARNING from John D. Cook https://www.johndcook.com/blog/2019/08/11/fixed-points-cliff/ There is an infinite number of fixed points. If the generator ever reaches one of these points, it gets stuck on those points forever. However these points are irrational numbers, and as such in floating point precision it will never get truly stuck. But as there are more of these irrational points near zero, the generator made a slight bias toward zero. -------------------------------------------------------------------------------