The problem isn't guessing a random number, it's narrowing down the search range to more probable keys. Thank you for posting the percentages; this way we can see that after 54, none have fewer than 40% of bits as 1 or 0, nor more than 60%. I believe we can accept this as a fact.
This means that perhaps you have gained some benefit from my post, which is why the saying "knowledge increases by sharing" is mentioned. Well, if you hold this viewpoint and it might indeed be true, in a 66-bit key, if we allocate zeros between 40 to 60 % and ones also between 40 to 60 % , we can significantly reduce the boundary of the search space. Yes, this statement holds a considerable amount of credibility 👍 👍