I think the direction of the storage wheel is inverted and the cards now catch a small lip above the opening and are directed to the output stack. While inserting I assume the wheel is fast enough that the cards use the lip as a ramp to jump over the gap. If you stop at :26 you can see that the first card stays just in front of the gap while the second card is already moving.

@@nikolasengelhard6386 I can summarize your entire paragraph into one word: "Gravity"

@radBRCast It cuts them with a knife

My but ho is in use

It's too slow and boring.

The description links to a more detailed video that links to two files describing the build.

I don't see a link in the description. I'm also interested to learn more about how I can build my own!

The gap is perfectly adjusted by the tape so that only one card gets picked up by the wheel.

@@naotost6383 Thanks!!

The order of the cards are pre-computed using numpy.ranodm.shuffle. Could you elaborate on how this could possibly not span the entire set of 52 factorial?

@@naotost6383 Sure. Firstly, your design is great; it’s the best DIY shuffler I’ve seen on KZfaq. As a matter of fact, the more I thought about it, the more I realized that it is nearly a ‘perfect’ shuffler - and - that pre-computing the shuffle in software could, theoretically, random-walk the entire 52! space. But, as you yourself probably know, your design - with numpy.ranodm.shuffle - is now only as good as NumPy’s RNG and shuffling algorithm. Are they perfect? Probably not. Let’s start with the ‘shuffling algorithm’. To be very clear, there is one - and only one - way to perfectly shuffle a deck of cards: Top To Random Shuffle! Google it; read every paper on it; know it inside and out. The ‘Top to Random Shuffle’ is the only way to ‘perfectly’ randomize a deck of cards. It’s not that hard to code once you understand it. I seriously doubt numpy.ranodm.shuffle is using it because it’s not fast in software & is non-deterministic in its run-time i.e. it could take 1 second to shuffle an array and then take 3 seconds to shuffle the exact same array on the next pass! A good, non-perfect, shuffling algorithm (to help understand how shuffling works) is the RC4 stream cipher. Go to Wikipedia and search for RC4. RC4 is ostensibly a stream cipher, but in actuality, it’s a cryptographically secure byte shuffler - It shuffles a deck of 256 cards! Is it perfect? No, but it is fast & easy to understand. Until very recently, RC4 was used in SSL/TLS certificates and it was the original WiFi encryption protocol (WEP). Apple replaced RC4 with AES in their phones only in 2017. RC4 is a great example of trading ‘perfection’ for speed. It may not be able to randomly-walk the entire 256! space, but it can walk a lot of it. This brings us to the RNG and how it’s being seeded. However, rather than dive into this ridiculously complicated subject in the KZfaq comment section, I'll just point you to the seminal book on the subject: Donald E Knuth’s ‘The art of computer programming, volume 2: Seminumerical Algorithms’. Google it; read it; know it inside and out. There are PDF’s floating around - the book can be expensive. Remember, you can only be as ‘perfect’ as your RNG, and, like RC4 demonstrates, if you can randomly-walk only 30! of the 52! space, that’s still impossible for any computer or ‘rain man’ to count into.

I forgot to mention that RC4 is a 'Fisher-Yates' shuffle.

Tape on the orange wheel.