The Probability of Coin Flips Why 50/50 Is Not Exactly Fair — and What That Reveals About Randomness

A coin flip is the universal symbol of fairness. Two possible outcomes. Equal probability. The fairest decision mechanism ever invented. Except it is not exactly 50/50. In 2023, researchers led by František Bartoš at the University of Amsterdam conducted a massive study: 48 people flipped coins 350,757 times. The result: coins land on the same side they started on about 50.8% of the time. The bias is tiny — less than 1% — but it is real, it is statistically significant, and it has been hiding in plain sight for centuries.

Here is the physics behind the bias, what it means for the concept of randomness, and why a coin flip simulator is actually fairer than a real coin.

The Physics: Why the Same-Side Bias Exists

The bias comes from precession — the wobble of the coin's rotation axis as it flips. When a coin is flipped, it does not rotate perfectly around a single axis. The axis wobbles slightly, which means the coin spends slightly more time with the initial side facing up during its trajectory. The effect was predicted by a 2007 physics model by Diaconis, Holmes, and Montgomery at Stanford, but the 2023 study was the first to confirm it experimentally at massive scale.

The bias is about 0.8% in favor of the same side — meaning if you start with heads up, the coin lands heads about 50.8% of the time. This is not enough to exploit in a single bet. But over 1,000 coin flips, the same-side bias would produce about 508 same-side results instead of the expected 500 — a swing of 8 extra wins. Over a lifetime of coin flips, the bias is real.

The practical implication: if someone offers to flip a coin for a decision, and you can see which side is facing up before the flip, call the same side. You have a 50.8% chance of winning — a tiny edge, but an edge. If you cannot see the starting position, the coin flip is fair from your perspective.

What This Reveals About Randomness

The coin flip bias is not a flaw in coin flips. It is a flaw in our model of coin flips. We model a coin flip as "two equally likely outcomes." But a real coin flip is a physical process governed by Newtonian mechanics — initial position, force, angular momentum, air resistance, landing surface. If you could measure all the initial conditions precisely enough, a coin flip would be deterministic, not random.

The randomness we perceive is sensitivity to initial conditions — a tiny difference in the flip force produces a completely different outcome. This is chaos theory, not probability theory. The coin does not "choose" heads or tails. It follows the laws of physics from the moment it leaves your thumb. We call it random because we cannot predict it, not because it is fundamentally unpredictable.

This is true of most "random" processes in everyday life. Dice rolls. Card shuffles. Lottery balls. They are all deterministic physical processes that we treat as random because measuring the initial conditions is impractical. The randomness is in our ignorance, not in the physics.

When You Actually Need Fair Randomness

For a decision between two restaurants, a coin flip is fine — the 0.8% bias is meaningless. For a cryptographic key, a coin flip is terrible — the bias, however small, is a vulnerability. For anything in between, use a digital coin flip powered by a cryptographically secure random number generator. The digital coin is not subject to precession, wobble, or the bias of physics. It is the truly fair 50/50 that real coins only approximate.

Flip a truly fair coin at free coin flip — no physics bias, no precession, just mathematics.