All write-ups

Physics9 min read

Heads, tails or edge: a deterministic coin

A coin flip feels random. In classical physics it isn’t: one exact starting state gives one exact landing. This is the model behind the Deterministic Coin Flip Lab, and the reason you still can’t call the toss.

Fig. 1 A tumbling coin along its flight. The face that ends up on top is decided by the orientation at first contact.

Two ideas that get confused

Classical determinism: within the model, one exact state produces one exact trajectory and one first-contact result. Practical predictability: a real measurement only pins the state down to a small region, and that region can straddle the line between heads and tails, even though every single state inside it has one answer.

The state

Everything the coin will do is fixed by four things at release: its position r₀, velocity v₀, orientation q₀ (a quaternion) and body-frame angular velocity ω₀.

X₀ = (r₀, v₀, q₀, ω₀)

The equations

Gravity and air drag move the centre of mass; Euler’s equations turn the coin; the quaternion keeps track of where it is facing.

m r̈ = mg + Fdrag(v − vwind, q)I ω̇ + ω × (Iω) = τdampq̇ = ½ q ⊗ (0, ω)

Translational motion, rotational motion and orientation, integrated by a deterministic solver. No random-number generator anywhere.

Calling the result

At the first geometric contact the solver rotates the coin’s normal into the world frame and reads its vertical component.

n(tc) = R(q(tc)) ezheads if ny ≥ +0.035, tails if ny ≤ −0.035, edge otherwise

Why you still can’t predict it

The lab lets you grow a measurement-error radius around one state and solves 81 deterministic neighbours. Near a boundary they split between heads and tails. Nothing random happened: the uncertainty was in what we knew, not in the coin.