Skip to content

Flip a coin online

Flip a fair coin with an animated toss, or run up to 10,000 flips at once and watch the heads share converge on 50% in a live fairness chart.

  • Runs in your browser
  • Nothing you enter is sent
  • Free, no account
Flip a coin
One flip, or up to 10,000 at once. Every flip comes from the browser's cryptographic random number generator.

No flips yet.

Between 1 and 10,000 flips per batch.

Tally
Flip the coin to start counting.

How to use the coin flipper

  1. Press Flip once for a single animated toss.
  2. Set a batch size and press the batch button to run up to 10,000 flips at once. Progress is shown while it works, and the page stays responsive because the flips are processed in chunks.
  3. Watch the fairness chart: it plots the running share of heads from the first flip to the most recent, against the 50% line.
  4. Press Reset to clear the tally and start a fresh experiment.

Is this coin actually fair?

Yes. Each flip reads a fresh 32-bit value from crypto.getRandomValues — the browser's cryptographically secure random number generator, seeded by the operating system — and maps it to heads or tails using rejection sampling, so no residual bias creeps in from the mapping. There is no seed you can predict, no dependence on the clock, and no memory of previous flips.

That last point matters more than people expect. The coin has no idea what it did last time. After nine heads, the tenth flip is still exactly 50/50.

Streaks are normal — the gambler's fallacy is not

The belief that a run of heads makes tails "due" is the gambler's fallacy, and it is the single most common misreading of randomness. Long runs are not just possible, they are expected:

Number of flips Expected longest run Chance of a run of 5+ Chance of a run of 10+
10~3~22%~0.2%
100~6–7~97%~5%
1,000~9–10>99.9%~38%
10,000~13>99.9%~99%

Run a batch of ten thousand here and check the longest-run figure against the table. A sequence of results with no long runs is the suspicious one — it is the signature of a human trying to fake randomness.

How the proportion converges

The law of large numbers says the proportion of heads approaches 50%. It does not say the counts get closer together — in fact the absolute gap tends to grow, roughly in proportion to the square root of the number of flips:

Flips Typical gap from an even split Typical gap as a percentage
100±5 flips±5.0%
1,000±16 flips±1.6%
10,000±50 flips±0.5%
100,000±158 flips±0.16%

So 5,050 heads out of 10,000 is completely ordinary, while 5,050 out of 10,000 expressed as 50.5% looks impressively close to fair. Both statements describe the same result — that is the whole idea behind the fairness chart on this page.

Real coins are slightly unfair

A physical coin toss is close to fair but not perfectly so. Careful experimental work — including a study that recorded hundreds of thousands of tosses — found a small bias, around 51%, towards the side that was facing up when the coin was flipped. It comes from precession: the coin wobbles about an axis rather than spinning cleanly, so it spends slightly more time with the starting face up. Spinning a coin on a table is far more biased still, because the rim is not symmetrical.

A digital flip has none of that physical structure to exploit.

Good uses for a coin flip

  • Settling a two-way choice quickly and visibly fairly
  • Deciding who goes first in a game
  • Teaching probability — the batch mode and the chart do the demonstration for you
  • Breaking a genuine tie in a decision: the useful trick is to notice how you feel about the result, which often tells you what you wanted all along

Privacy

Flips are generated and counted in your browser. Nothing is sent to a server, and the tally disappears when you close the tab.

Questions

Is this coin flip really 50/50?

Yes. Each flip reads a fresh value from crypto.getRandomValues and uses rejection sampling, so heads and tails are exactly equally likely and no result depends on the previous one.

Why did I get eight heads in a row?

Because runs are normal. In 1,000 flips a run of ten identical results is more likely than not. A streak is evidence of randomness, not of a broken coin — the fairness chart shows the share settling towards 50% as the count rises.

How many flips can I run at once?

Up to 10,000 per batch, and they accumulate across batches. The work is chunked with a progress bar so the page never freezes, even on a phone.

Is a real coin toss actually fair?

Almost, but not quite. Careful experiments find a slight bias — roughly 51% — towards the side that was facing up when the coin was flipped, because of how the coin precesses in the air. A digital flip has no such bias.