🪙Heads or tails, settled in a click.

✓ Tested & verified Updated: Aug 6, 2026 How the randomness works

Coin flip

Chance of heads

50 / 50

Press flip — the digital coin has no same-side bias

This session: 0 heads · 0 tails

Coins to flip at once

Try also

Is a coin flip really 50/50?

A digital flip is exactly 50/50. A physical one is not quite. Diaconis, Holmes and Montgomery showed that a tossed coin caught in the hand carries a small dynamical bias towards the face it started on — around 51% — because a real toss precesses rather than tumbling cleanly. Bartoš and colleagues tested it the hard way in 2023 with 350,757 recorded flips and found the same-side rate at 50.8%.

That bias is a property of hands and wrists, not of coins: it has nothing to do with the metal being heavier on one face, and it disappears if the coin bounces. It is also far too small to notice — about one extra same-side result every 125 tosses.

This page has no such quirk. Each flip draws a fresh value from your browser’s cryptographic random number generator, so heads and tails are exactly equally likely every time, with no memory of the last result. Flip up to twenty coins at once; the session tally counts them all.

Runs are the part everyone misreads. These are the odds of a specific sequence of consecutive identical results on a fair coin — and they are much less rare than they feel.

Result	Probability	Roughly
2 in a row	25%	1 in 4
3 in a row	12.5%	1 in 8
5 in a row	3.125%	1 in 32
8 in a row	0.39%	1 in 256
10 heads from 10 coins	0.098%	1 in 1,024

Flip a hundred times and a run of five is close to expected, a run of seven is unremarkable, and the absence of any long run would be the genuinely suspicious outcome. Whatever has just happened, the next flip is 50/50.

The full story of the physical bias — why a caught coin favours its starting face and why a bouncing one does not — is in is a coin flip really 50/50. For more than two outcomes, use the dice roller or put the options on the wheel.

Frequently asked questions

Is a real coin actually biased?

Slightly, and not in the way people assume. The bias is not in the coin’s weight distribution but in the motion: a tossed coin wobbles about an axis, so it spends marginally more time showing the face it started on. Caught in the hand, that produces roughly 51% same-side results — confirmed at 50.8% across 350,757 flips in the 2023 Bartoš study. Letting it land and bounce removes the effect.

How is the digital flip made fair?

Each flip draws random bytes from crypto.getRandomValues, the generator browsers use for encryption keys, and maps them to heads or tails with rejection sampling so there is no rounding bias. There is no seed you could predict, no pattern across flips, and no code path that looks at the running tally to even things out.

Can I flip more than one coin?

Up to twenty at once, which is the fast way to answer probability questions rather than decisions. Ten coins land all-heads about one time in 1,024. Flipping twenty and counting the heads is also the clearest demonstration of why results cluster near half without ever settling exactly on it.

I got six heads in a row. Is something wrong?

No — six in a row happens about one time in 64, so in a hundred flips you should expect a run like it. The next flip is still 50/50, because coins have no memory and neither does this page. Believing an even-up correction is due is the gambler’s fallacy, and it is the single most common misreading of randomness there is.

Is this good enough to settle a bet?

It is arguably better than a real coin, since it has no same-side bias and neither party controls the toss. Flip once, with both people looking at the screen and the call made in advance. What it cannot do is prove afterwards which way it landed, so for anything with money on it, film the screen or use the random name picker with both names on it.

Does the tally save between visits?

No. The heads and tails counters run for the current session and reset when you reload or close the tab. Nothing is stored on a device or sent to a server. If you are collecting data for a class experiment, write the numbers down as you go — or flip twenty at a time and record each batch.

How the randomness works

Every flip draws bytes from crypto.getRandomValues, the browser’s cryptographically secure generator, and converts them to heads or tails using rejection sampling so that neither face is even fractionally favoured. Flips are independent: no seed carries over, no counter is consulted, and the running tally is a display only — it never feeds back into the next draw. Multi-coin flips draw one independent value per coin rather than deriving them from a single number. Everything happens locally in the page, and nothing is stored or transmitted.

Worked example

With Coins to flip at once 1, this page works out 50 / 50. Press flip — the digital coin has no same-side bias

This session0 heads · 0 tails

Sources:

What the flip is really deciding

The interesting thing about coin flips is not the odds, it is what people do with the result. A coin is the standard instrument for decisions that are genuinely balanced, and most decisions people flip for are not — they are decisions where one option is quietly preferred and the flip is being used as permission.

Which is why the most useful thing a coin does happens after it lands: check whether you are pleased. Disappointment at tails is data about what you actually wanted, and it is available nowhere else in the deliberation.

For the mechanical uses — kick-offs, who goes first, splitting a bill — the flip is doing exactly what it says, and the digital version has the edge over metal because nobody can practise the throw. Just call it before it lands, as with any coin, and accept the first result. Best of three is not a coin toss, it is a rematch.