Flip a coin online for free. Get a fair, random heads or tails instantly, flip many at once, and keep a running tally. Great for quick, unbiased decisions.
Press Flip the coin. The coin spins for about two-thirds of a second and lands on Heads or Tails, and that is the decision made. No coin in your pocket, no arguing about whether the flip was caught or slapped, no one calling it late. The result is displayed in large type under the coin, and a running tally keeps count of every flip you have made since the page loaded.
This is the tool for a single binary decision — who goes first, which restaurant, who takes the bins out, which of two equally reasonable options you have been stuck between for twenty minutes. It runs entirely in your browser, so nothing you do here is sent anywhere, logged, or associated with you. Reload the page and the tally starts again from zero.
Say heads and tails out loud before you press the button. Assigning meaning after seeing the result is where coin-flip decisions go wrong, and it is the reason people re-flip: they discover, on seeing tails, that they wanted heads. That is not a failure of the coin. If you feel disappointed by the outcome, you have learned what you actually wanted, and you should take that answer rather than flipping again.
For a group, put the screen where everyone can see it and let the person who is not choosing press the button. The flip is instant and unrepeatable, which removes the usual accusations about a badly caught coin. If a decision has three or more options, a coin is the wrong instrument — use the random number generator with a range of 1 to the number of options instead.
The batch control exists because a single flip tells you almost nothing about the coin, while a few thousand tell you a great deal. Type 1,000 and press it: the tally jumps, the heads-share bar moves close to the 50% marker, and the recent-flips strip shows the tail end of the run. This is the fastest way to demonstrate, in a classroom or an argument, that a fair process converges without ever being tidy along the way.
Some worked numbers to try:
| Batch size | What you will typically see | The point |
|---|---|---|
| 10 | Splits like 7–3 or 8–2 are common; 5–5 is not the usual outcome | Small samples look badly unbalanced even when the process is fair |
| 100 | Usually somewhere in the mid-40s to mid-50s for heads | The percentage tightens quickly at first |
| 1,000 | Heads share sits close to 50%, but the raw gap between heads and tails is often larger than it was at 100 | Proportions converge while absolute differences grow — both at once |
| 10,000 | Heads share close to 50%; long runs of the same face appear somewhere in the sequence | Streaks are a feature of randomness, not evidence against it |
This is the complaint every coin flipper receives, and the answer is worth spelling out. A fair coin has no memory. After five heads, the chance of heads on the next flip is still one half, exactly as it was on the first flip. The coin does not know what it has already done and has no mechanism by which it could compensate. The belief that a run of heads makes tails "due" is the gambler's fallacy, and it is the single most expensive misunderstanding in probability.
Runs are also far more common than intuition suggests. Any specific six-flip sequence is equally unlikely — HHHHHH is exactly as probable as HTHTTH — but only one of those two looks like a pattern to a human. In a long batch, runs of six or more of the same face are close to inevitable simply because there are so many places for one to start. Flip a few thousand times with the batch control and scan the recent-flips strip: you will find clumps. Their absence would be the anomaly.
What people are really asking is whether the process is biased, and the honest way to check is aggregate share over a large number of flips, not the look of any short sequence. That is what the heads-share bar and its 50% marker are there for.
The coin is decided by the browser's built-in Math.random(): a value below one half is Heads, otherwise Tails. That is a pseudorandom generator — a deterministic algorithm seeded by the browser, producing a sequence that is uniform, well distributed, and unpredictable for any practical human purpose. Heads and tails each get exactly half the interval, so there is no thumb on the scale in either direction.
It is not cryptographically secure, and this page does not claim otherwise. Browsers expose a separate function, crypto.getRandomValues(), that draws on the operating system's entropy sources and is built so that seeing past output tells you nothing about future output. This tool does not use it. Math.random() carries no such guarantee: it runs from internal state that is, in principle, inferable from enough observed values.
In practice that distinction does not matter for deciding who buys coffee, and it matters enormously for anything where a person could profit by predicting the next result. Do not use this coin to settle a wager of consequence with someone who controls the screen, to seed anything security-related, or as the mechanism in a promotion that needs to be provably fair to a regulator. For a real-world coin, the fairest fix is the classic one: one person flips, the other calls it in the air.
Because everything happens in your browser, there is also no record. Nothing is uploaded, nothing is stored, and no one — including us — can retrieve or audit your flips afterwards. That is good for privacy and bad for disputes, so if the outcome needs to be witnessed, have it witnessed live.
Not perfectly, and that is a nice thing to know while you are using a digital one. A physical coin flip is a mechanical process, and its outcome depends on how the coin is held, how hard it is struck, how many times it turns and whether it is caught or allowed to bounce. Coins caught in the hand tend to favour the face they started on, because the coin precesses about an axis rather than tumbling perfectly end over end. Spinning a coin on a table is worse still, since a coin with an asymmetric rim will tend to fall toward one side. And a coin landing on a hard floor introduces bouncing, which is effectively unmodellable.
A simulated flip removes all of that. Heads and tails split the interval exactly in half, with no edge geometry, no starting face and no wrist. Whatever else Math.random() is, it is more evenly balanced between two outcomes than the coin in your pocket.
The tool has a second tab, Run an Experiment, which is the same underlying flip wired into a full probability simulator: it runs a set number of trials with an adjustable speed, plots observed frequency against the theoretical 0.5, tracks the absolute deviation as trials accumulate, and records the current and longest run of the tracked outcome. That mode is what the coin flip simulator page is built around, and it is the right destination if you are teaching or checking the law of large numbers rather than making a decision. If you just want a yes or a no right now, stay on this tab.
Yes. Each flip is drawn independently from your browser's random source and mapped to heads or tails with an exact 50% chance each. Unlike a thrown physical coin, there is no spin, catching motion, or starting-side bias, so the long-run split is as even as it gets.
For a single flip it is 1 in 2, or 50%. Past results never change the odds — even after five heads in a row, the next flip is still a clean 50/50. Believing otherwise is the gambler's fallacy.
Yes. Enter a number and use the Flip × button to resolve up to 10,000 flips instantly. The heads, tails, and total counters update so you can see how the run compares to the expected 50/50.
Assign one choice to heads and the other to tails, then flip. For a genuine dilemma, pay attention to your gut reaction the moment it lands — relief or disappointment usually reveals the option you actually wanted.
In practice, yes. Real hand-flipped coins land on their starting side slightly more than half the time because of how they wobble. A digital flip has no physical bias, so it is a cleaner 50/50.