A Short One2Six Advantage-Play Project Update

The simulator has moved from reconstructing the machine to finding a player-observable signal

The ShuffleMaster project has moved quickly over the past few weeks, and the articles have struggled to keep up with it.

What began as an attempt to understand the One2Six continuous shuffler has developed into a working open-source simulation project. It now includes a blackjack engine, several comparison card sources, stable physical-card identity, realistic discard timing, a configurable One2Six-style carousel model, baseline experiments and an expanding set of source-level and game-level diagnostics.

More importantly, the simulator has started producing results worth following.

Under the current model, very short returns of the same physical card are strongly suppressed even though the average return time remains close to the physical-IID expectation. The next stage showed that some of this mechanical memory becomes visible at the table. Discard history predicted the composition of the next deal on held-out simulation seeds, and the strongest high-card-rich score band moved the estimated game from an unconditional loss of approximately 0.462% to almost exactly break-even at -0.002% by point estimate.

The uncertainty around that strongest state still crosses zero, so I have not established a validated positive-EV game. Even with that qualification, the result is substantial. The project has moved beyond showing that a plausible machine model contains hidden mechanical structure. It has identified information visible to the player, shown that the information predicts future cards, and followed the signal into blackjack expected value.

The site has grown alongside that work, and the series has become harder to follow than it needs to be. I am therefore tidying the articles, improving the navigation, replacing vague chapter titles with titles that explain what each stage actually did, and making the path from mechanism to model to signal easier to see.

I am not rewriting the project into a cleaner history than the one that happened. The earlier assumptions, dead ends and changes in direction remain part of the record. The aim is to make it easier to tell what was known at each stage, what was being assumed, what had actually been built and what the experiments had shown.

Open source and project scope

The simulation code is publicly available through the mathematical-ev/shufflemaster-simulation repository and is licensed under GPL-3.0-or-later.

The code, assumptions, tests and experiment runners are open for inspection. People may run, study, modify and redistribute the project under the terms of that licence, while distributed modified versions remain subject to the GPL's source-availability and copyleft requirements. I chose that licence because I want the work to remain open rather than disappear into a closed derivative.

This is an independent personal project built from public sources and explicit modelling assumptions. It is not affiliated with or endorsed by the manufacturer, any casino or any regulator. The simulator has not been tested against a physical One2Six machine, and I do not claim that it reproduces the exact mechanism or behaviour of any production machine.

Nothing published here is a guarantee of profit or a recommendation that anyone gamble. A result produced by a configurable simulation should not be treated as a description of a real machine without physical validation. I do not recommend using electronic devices to count or record cards in a casino, interfering with gaming equipment or procedures, breaching casino or game rules, or engaging in any form of cheating or dishonest conduct.

Anyone who chooses to gamble does so at their own risk. I accept no responsibility for losses incurred by treating an experimental model as a proven real-world advantage.

This remains a personal project. I am doing it because the question caught hold of me and I wanted to see how far I could take it.

The recent results have made the answer considerably more interesting than I expected.