Dear Samir, I think there is not a "best" statistical approach to check a mathematical conjecture; if the conjecture is true, you can work for about 10^10...0 tests/experiments with positive results but you have not mathematically proven the conjecture; if it is false, you can work for about 10^10...0 tests/experiments with positive results but the (10^10...0 + 1)th test (which for some reason you have not tested) is negative. See histories about some past conjecture (Fermat Last Theorem, Poincaré, Goldbach). Gianluca
Thank you for responses. if we look at this article in wilipedia (https://en.wikipedia.org/wiki/Collatz_conjecture ) about Collatz conjecture. It s cited the Undecidable generalization decision. So I'm seeking for algorithmes to have clear vision about if the conjecture is undecidable.