This topic is indeed very interesting, but I think understanding is much easier than practical application. Any counting system gives you the key: High cards are valuable to the player, while low cards are valuable to the dealer.
A bunch of cards is a mixture of high, low, and neutral cards, in a randomized order. Random means that this is a smooth distribution, but since cards are not smeared over the whole shoe, but on distinct positions, cards in a a shoe exhibits "shot noise".
Imagine you get a cup of (raw) rice, and you throw it on the floor. The rice will distribute on the floor over a (more or less) flat random distribution. But since the rice is granular, the distance between some grains will be a lot closer than the average distance, and on some larger spots there will be substantially less grains.
This is shot noise, it's origin is the pure random and independent distribution of a fixed number of granular particles. The number of grains on any given area follows a Poisson distribution. Like the grains of rice, high, neutral, and low cards distribute in the shoe in a random shuffle. There will be high cards that are closer together than the average distance, and cards will be farer away than average. As each card is independent (unless they are somehow "sticky") this local clustering is completely random. In any segment of the shoe, the number of specific cards will also follow a Poisson distribution.
This is very well understood.
The problem is: as the origin is in complete randomization, there is nothing to exploit from such an Poisson shoe.
The only way to exploit a perfectly shuffled shoe is to implement a non-poisson distribution. One method would be indeed "sticky" rice: If you would wet the grains to make them sticky, then two neighbouring grains will tend to be closes together during the throw than statistically expected (if they were dry). This would be artificially clustering wet grains (again, on the average). If you could do something similar with the cards, which means changing the physical properties of how they "interact", it could be used to exploit the game. High cards would then likely to be followed by other high cards, (and vice versa low cards followed by low cards). Then, even with a perfect shuffle procedure, you would maintain an advantage with an adaptive strategy.
Changing the physical properties is not a fiction, you can make the game of 3-card Monte much much smaller (with electrons or similar quantum objects). You are then entering the quantum-mechanical regime, where particles behave substantial different than expected. A version of something like 2of5-card Monte has been created by physicists, where even a completely random strategy (i.e. picking spots at random) performs substantially worser than naive probabilities would predict.
So, unless you change the physical properties of cards, you will always end up with an poisson distributed shoe, which gains no information for the player, as cards do not interact, and will thus be completely randomized.