A coin flip is the most trusted tie-breaker on Earth. We use it to start Super Bowls, settle arguments, and outsource the decisions we can't face. The whole ritual rests on one assumption: that heads and tails are equally likely. But is that assumption actually true?

The short answer is "almost." A perfectly fair 50/50 is easier to get from a computer than from your thumb. This page digs into the real probability of a coin toss — the odds of a heads streak, why "tails is due" is a myth, and the physics study that caught real coins cheating by a fraction of a percent.

The short answer

  1. A digital flip is exactly 50/50. There's no physics to bias it.
  2. A real coin is close, but not perfect. It lands the side it started roughly 51% of the time.
  3. Streaks are rare, but each flip isn't. Ten heads in a row is 1 in 1,024 — yet the tenth flip alone is still 50/50.
  4. The coin has no memory. Past results never change the next flip. That's the gambler's fallacy.

Want the result, not the theory? Use the tool below and skip to the math after.

Just need a heads or tails, right now?

Our Online Coin Flipper gives a perfectly fair 50/50 result in one tap — no real coin, no download, no thumb bias. Flip it now and read on to see exactly why it's fairer than the real thing.

Is a Real Coin Flip Actually 50/50?

Here's the surprise: a real, physical coin toss is not perfectly fair. It's close enough for a pickup game, but the math and the physics both say there's a small, real bias.

In 2007, Stanford statistician Persi Diaconis and colleagues built a model of a spinning, tumbling coin and predicted it should land the same side it started about 51% of the time. The reason is precession: a flipped coin doesn't spin around a perfectly flat axis, so it spends a hair more time with its starting face pointing up.

For years that was just a prediction. Then in 2023, a team led by František Bartoš put it to the test the hard way — they flipped coins 350,757 times by hand and counted. The result: coins landed the same side they started 50.8% of the time, almost exactly matching the prediction. Over a few flips you'd never notice. Over thousands, the bias is unmistakable.

And that's flipping. Spinning a coin on a table is far worse — the slightly heavier engraved face makes some coins land one way 80% of the time. So if someone offers to spin for it, be suspicious.

None of this makes the coin useless. It makes the point that "fair" is a property you have to engineer, and a real coin only gets most of the way there.

Coin Flip Probability: The Odds of a Streak

Assume a perfectly fair coin for the rest of the math — the kind a computer gives you. The chance of a single specified result (say, heads) is 1 in 2. The chance of getting that same result several times in a row is where it gets interesting.

Each flip is independent, so you multiply. Two heads in a row is ½ × ½ = ¼. The general rule: the odds of getting n of the same result in a row is 1 in 2n.

Heads in a rowProbabilityOdds
11/21 in 2
21/41 in 4
31/81 in 8
41/161 in 16
51/321 in 32
61/641 in 64
71/1281 in 128
81/2561 in 256
91/5121 in 512
101/10241 in 1,024

So a run of ten heads is genuinely rare — roughly the same odds as rolling a specific number on a 1,024-sided die. Keep going and the numbers explode: twenty heads in a row is 1 in 1,048,576. This is exactly what makes streaks feel spooky, and it's exactly what fuels the myth in the next section.

The Gambler's Fallacy, Explained Properly

You've just flipped five heads in a row. Surely tails is "due" now? This feels obviously true, and it is completely wrong. This is the gambler's fallacy, and understanding why it's wrong is the single most useful thing on this page.

The table above gives the odds of a streak before you start flipping. Five heads from a standing start is 1 in 32. But once five heads have already landed, that's history. The coin is not keeping score. It has no mechanism to "balance out." The sixth flip is a brand-new event with the same 50/50 odds as the first.

The trap is confusing two different questions:

  • "What are the odds of six heads in a row?" Answer: 1 in 64 — asked before any flips.
  • "Given five heads already landed, what are the odds the next flip is heads?" Answer: 1 in 2 — still a coin toss.

Both the fantastically unlikely streak and its ordinary continuation cost exactly one more 50/50 flip. The rare part already happened; the next flip doesn't know or care. Casinos have made fortunes on people betting that black is "due" after a run of red. Don't hand them yours.

Why a Virtual Coin Is a True 50/50

If a real coin carries a small physical bias, a good digital one carries none. When you tap our coin flipper, it asks your browser's crypto module for a cryptographically secure random value and maps it cleanly onto heads or tails — no thumb force, no precession, no engraved-side weighting. The result is the exact 50/50 the physical world only approximates. (Our dice roller uses the same source; if you want the full breakdown of how that randomness is generated and why it beats physical dice, we cover it in the dice notation and probability guide.)

The practical upside beyond fairness: a digital flip can run a hundred tosses in an instant. That's genuinely useful in a classroom — have students flip 100 times and watch the ratio wander toward, but rarely land exactly on, 50/50. It's the streak math above, live.

A Short History of Settling It With a Coin

Flipping for it is ancient. The Romans played navia aut caput — "ship or head" — after the ship's prow and emperor's head stamped on their coins, using it to settle disputes and small bets. Medieval England knew it as "cross and pile." The modern ceremonial version reached its peak with the Super Bowl, which has opened with a coin toss since 1967.

Coins have even named cities: Portland, Oregon exists under that name because its two founders flipped for it — the man from Portland, Maine won a best-of-three over the one who wanted "Boston." A small piece of metal, a 50/50, and a city gets its identity. That's the enduring appeal of the toss: it's the universally accepted way of saying "we can't agree, so let chance decide."

How to Actually Decide With a Coin Flip

The coin's real power isn't the physics — it's the psychology. The trick: assign your two options to heads and tails, flip, and watch your own reaction to the result. If the coin says "pizza" and your heart sinks, you wanted tacos all along. The flip didn't decide for you; it forced your gut to show its hand.

A coin only handles two options, though. When the decision is bigger, hand it to the right tool:

  • More than two choices? The Decision Maker spins between as many options as you can list.
  • Want a verdict with a bit more nuance? The Yes / No / Maybe picker adds the honest third answer a coin can't give.
  • Feeling mystical about it? The Magic 8-Ball delivers your fate with appropriate drama.
  • Need a number, not a side? Reach for the Dice Roller instead.

Whichever you use, the principle is the same as the coin's: the randomizer isn't really making the choice. It's giving you a fair, unarguable nudge — and a mirror for what you were quietly hoping for.

Frequently Asked Questions

Is heads or tails more likely?

On a digital coin flip, neither — it's an exact 50/50. On a real coin it's very close but not perfect: the 2023 study of 350,757 flips found a coin lands the same side it started about 50.8% of the time. For a guaranteed-fair result, a virtual flip removes the bias entirely.

How do I flip a coin without a real coin?

Use an online coin flipper. It runs in your browser, needs no download, and returns an unbiased heads-or-tails instantly — ideal when you don't have change on you or want a result nobody can accuse of thumb-tampering.

If I've flipped five heads in a row, is tails "due"?

No. The coin has no memory, so the next flip is still exactly 50/50. Believing tails is "due" is the gambler's fallacy — the long odds of a streak apply before you start, never to the single flip in front of you.

Can I flip a coin 100 times at once?

Yes. A digital flipper can generate a long run instantly, which makes it a great way to show a class how a real sample drifts toward, but rarely lands exactly on, a perfect 50/50 split.