Is a coin biased toward heads or tails?

Neither, in the usual sense. The measured bias favours whichever side started facing up, not heads in particular. If you do not look at the starting position, or you alternate it, the heads-versus-tails rate comes back to essentially 50/50. This is why calling heads or tails is fair even though the toss itself is slightly biased.

Does it matter whether you catch the coin or let it land?

It can. The same-side effect was measured on caught flips. Letting the coin bounce on the floor or a table adds another chaotic stage that scrambles the outcome further, which is generally expected to reduce a toss bias rather than add to it. It also introduces the possibility of the coin landing against an obstacle or on its edge.

Is spinning a coin on a table the same as flipping it?

No, and it is much less fair. A spun coin falls onto whichever face its centre of mass favours, and many coins have an asymmetric rim or relief design. Spinning can produce a large bias — a real skew rather than a fraction of a percent — so it is the wrong method whenever the result matters.

Can a physical coin toss be rigged?

A practised flipper can influence a caught toss, because the outcome is determined by the initial conditions rather than by anything genuinely random. Persi Diaconis has demonstrated flips with a strongly repeatable result. This is the strongest argument for a digital flip when both parties need to trust the outcome: neither side can control the input.

Is an online coin flip more fair than a real one?

On the two measurable dimensions, yes: it has no same-side bias and no way for either party to steer it. What it cannot offer is an independently auditable record. A digital flip is the fairer process; a physical flip is the more publicly verifiable ceremony, which is why televised sports tosses still use a real coin.

Is a Coin Flip Really 50/50? What the Measurements Show

The short answer

For a coin flipped and caught by hand, no: it is about 50.8/49.2 in favour of the side that was facing up before the toss. For a coin flip generated by a correctly built program, yes: exactly 50/50, because there is no coin, no wobble, and no hand.

What was actually measured

In the largest experiment on the question, 47 people flipped coins 350,757 times and recorded both the starting orientation and the result. The coin landed on the side that started facing up 50.8% of the time. That is a small effect, but the sample is large enough that it is not noise: it works out to roughly one extra same-side landing in every 125 tosses.

The finding was not a surprise. A 2007 physical analysis by Persi Diaconis and colleagues had already predicted a same-side bias of about 51% from the mechanics of a real toss, and the measurement landed just under that prediction.

Why the starting side matters

A hand-tossed coin does not spin cleanly about its diameter. It precesses — the spin axis itself moves — so the coin spends slightly more of its flight time with the initial face pointing upward. Averaged over a whole flight, that tilt gives the starting side a marginally better chance of being on top when the coin is caught. The size of the effect depends on how a particular person releases the coin, which is why it is a property of the toss rather than of the metal.

Why this does not make “heads” a good bet

The bias favours the starting side, not heads. Unless you know which face was up before the toss and the flipper is consistent about it, there is nothing to exploit: over many tosses with varied starting positions the heads rate returns to about 50%. Calling heads or tails is still fair. What is not fair is calling it after seeing the starting face against a consistent flipper.

The methods that are genuinely unfair

Spinning a coin on a table

This is a different physical process, and a much worse one. A spinning coin decelerates and topples toward the face favoured by its mass distribution, and many coins carry more metal or more relief on one side. The resulting skew can be large rather than fractional. If a decision matters, do not spin the coin.

A practised flipper

A coin toss is deterministic, not random: it is chaotic and sensitive to initial conditions, which is a different thing. Given enough control over the release, a flipper can make the outcome repeatable, and Diaconis has demonstrated exactly that. A physical toss is fair only because ordinary people cannot control it that precisely.

Letting it land versus catching it

The 50.8% figure comes from caught flips. Letting the coin bounce adds a second chaotic stage that scrambles the outcome further, which should wash out some of the toss bias — at the cost of the coin rolling away, landing against something, or, very rarely, standing on its edge.

Where a digital flip is better, and where it is not

A browser flip removes both problems at once. There is no precession to favour a starting face, and no participant can influence the input. On this site each flip reads one byte fromcrypto.getRandomValues() and uses its lowest bit; a byte holds exactly 128 even and 128 odd values, so the two outcomes are equally likely by construction rather than by approximation. The published method shows the mapping and the complete source.

What a digital flip cannot do is prove itself to a sceptic. Nobody watching can inspect the random source, and there is no signed record of what was drawn. A physical coin is worse randomness but better theatre, which is exactly why a televised kickoff still uses one. For a decision between two people who already trust the tool, the browser flip is the fairer of the two. For a regulated draw, neither is adequate without independent auditing.

Does 50/50 mean the results will look even?

No, and this is where most “the coin is broken” complaints come from. A perfectly even coin produces streaks: across 100 flips there is about a 97% chance of at least five identical results in a row somewhere. The proportion of heads converges toward 50% as flips accumulate, while the raw difference between the two counts tends to grow. Both facts are normal, and both are visible in the probability lab and quantified in the coin flip probability tables.

Common questions

Is a coin flip really 50/50?: Almost, but not exactly, for a physical coin. In the largest measurement to date, 47 people flipped coins 350,757 times and the coin landed on the side that started facing up 50.8% of the time. The bias is in the toss, not the coin: a spinning coin wobbles about an axis that keeps the starting face up slightly longer. A digital flip has no such physics, so a correctly built one is exactly 50/50.
Is a coin biased toward heads or tails?: Neither, in the usual sense. The measured bias favours whichever side started facing up, not heads in particular. If you do not look at the starting position, or you alternate it, the heads-versus-tails rate comes back to essentially 50/50. This is why calling heads or tails is fair even though the toss itself is slightly biased.
Does it matter whether you catch the coin or let it land?: It can. The same-side effect was measured on caught flips. Letting the coin bounce on the floor or a table adds another chaotic stage that scrambles the outcome further, which is generally expected to reduce a toss bias rather than add to it. It also introduces the possibility of the coin landing against an obstacle or on its edge.
Is spinning a coin on a table the same as flipping it?: No, and it is much less fair. A spun coin falls onto whichever face its centre of mass favours, and many coins have an asymmetric rim or relief design. Spinning can produce a large bias — a real skew rather than a fraction of a percent — so it is the wrong method whenever the result matters.
Can a physical coin toss be rigged?: A practised flipper can influence a caught toss, because the outcome is determined by the initial conditions rather than by anything genuinely random. Persi Diaconis has demonstrated flips with a strongly repeatable result. This is the strongest argument for a digital flip when both parties need to trust the outcome: neither side can control the input.
Is an online coin flip more fair than a real one?: On the two measurable dimensions, yes: it has no same-side bias and no way for either party to steer it. What it cannot offer is an independently auditable record. A digital flip is the fairer process; a physical flip is the more publicly verifiable ceremony, which is why televised sports tosses still use a real coin.

Sources

Bartoš, F. et al. — Fair coins tend to land on the same side they started: 350,757 flips by 47 people, same-side rate 50.8%.
Diaconis, P., Holmes, S. and Montgomery, R. (2007) — Dynamical bias in the coin toss: the precession model that predicted a same-side bias near 51%.

Want a coin that is deliberately uneven instead? Configure any split from 1% to 99% with the weighted coin flip.