How to improve a random number generator (Q1971517)

Numerical experiments with random walking indicate the sensitivity of the results on the autocorrelation of the used (pseudo)random numbers. The author experiment with the double randomization technique. This technique uses one generator to produce a large array of random numbers. A second generator (not necesarily a different one) samples numbers from the array. Experiments are performed with various combinations of generators. The results of random walking experiments using the numbers are used as the indicator of the quality of such tools. The conclusion, not surprising, is that the quality depends mainly on the quality of the generator filling the array. One combination of generators has very good results. The paper presents no formal tests. All its conclusions are based on the diagrams showing graphically the results of the random walk experiments.