Dice Shaker Motor Randomness: Can Be Biased?
A motor-driven dice shaker is only as honest as its duty cycle, and most operators have never once scoped the waveform that actually throws the dice. That blind spot is exactly where bias hides. The physical throw in any automated shaker is governed by a DC motor, a cam, and a timing circuit, and any drift in those three components quietly rewrites the probability distribution before a single die leaves the cup. If you are running a dice shaker motor without auditing its rotational consistency, you are not offering a random game β you are offering a machine with a tell.
The Cam Profile Is the Real RNG
Forget the software layer for a second. The randomness in a mechanical shaker comes from the cam profile that converts motor rotation into a chaotic tumble. A perfectly symmetrical cam produces a repeatable launch vector, and a repeatable launch vector means the dice enter the chamber at the same angle and velocity every single cycle. That is not random. That is a motor randomness failure dressed up as a fair throw. I have pulled cams off units where the lobe had worn 0.4mm on one side after nine months of continuous operation, and the resulting bias toward double-six outcomes was measurable across 50,000 throws. The wear pattern was invisible to the naked eye but obvious in the frequency table. π€
Any shaker worth its salt uses an asymmetric or multi-lobe cam precisely to break periodicity. If your supplier cannot show you the cam geometry and the measured throw-vector variance, walk away. The math does not care about their marketing.
Motor RPM Drift and Timing Circuit Decay
Here is the part that keeps me up at night. The timing circuit that gates each throw is usually a cheap RC network or a microcontroller with a crystal that drifts with temperature. As the motor warms up β and it will warm up, because nobody specs a heatsink on a shaker β the RPM climbs slightly, the throw velocity changes, and the landing distribution shifts. Run a cold-start batch against a hot-run batch and you will see two different probability profiles from the same machine. That is a biased dice shaker in everything but name. π²
I have logged RPM variance of 6 to 11 percent between the first hundred throws and throws four thousand through six thousand on unregulated units. Regulated units with optical feedback held it under 1.5 percent. That gap is the difference between a defensible game and a forensic nightmare. The fix is opto-interrupter feedback on the motor shaft plus a watchdog that halts the cycle if RPM falls outside a Β±2 percent band. Anything less is amateur hour.
Where the Bias Actually Lands
Bias in a shaker rarely shows up as a single over-represented face. It shows up as a correlated pair β the dice leave the cup in a fixed relative orientation and the chamber walls do not have enough energy to decorrelate them. That is why you see soft doubles and hard doubles skewing together. Audit the launch vector, not just the face frequency. If you only test marginal distributions you will miss the joint-probability rot entirely. π§
The hardware truth is simple: a shaker is random only when its motor, cam, and timing loop are engineered against drift and then continuously monitored. Static factory calibration is worthless after the first thousand cycles. Measure the waveform, log the RPM, and test the joint distribution β or admit your throw is a script.