In that case every possible way of collecting randomness is a scheme. What is probing /dev/urandom everytime you need a random number, if not a scheme?
It's very difficult to convince someone that they're not as smart as they think they are.
What it really comes down to is this: if you're not a cryptographer, you shouldn't be coming up with your own cryptography (and that definitely applies to coming up with your own RNG). To do otherwise is simply hubris.
I don't mean any disrespect in what I wrote above, it's just that cryptography (and by extension CSPRNGs) is an extremely complex subject which is studied by experts who devote their entire professional life towards that study (and even
they frequently make mistakes!). I
know I'm not in that league, and I can only presume you're in the same boat, therefore the best advice is to follow the advice of people who know better than either of us.
All of this is why using /dev/urandom is a good idea: it
was developed by experts, and it's also why any scheme developed by you or I is fraught with serious risk. (I did not develop the "feed a shuffled deck into a hash function" scheme by the way, i just wrote that simple python script.)