🪙 Coin Flip

Each flip draws a floating-point number from the browser's built-in pseudorandom generator, which produces values uniformly distributed between zero and one. Anything below 0.5 is recorded as heads, and anything from 0.5 upward is recorded as tails. That gives each outcome an equal probability on every flip, and no memory of previous results, which is exactly the behaviour of a fair physical coin.

When you select more than one coin, the same draw is repeated independently for each coin in the set, so a ten-coin toss is ten separate fair flips rather than one result duplicated. The heads and tails counts for that throw are tallied and displayed alongside the individual outcomes.

The statistics panel accumulates across the session. Total flips increments by the number of coins thrown. Heads and tails counts accumulate the same way, and the heads percentage is the heads count divided by the total, expressed out of 100, with the tails percentage being the remainder. The current streak counts consecutive identical results and resets to one whenever the outcome changes; the longest streak records the highest value the current streak has reached. All of this is held in the page and clears when you reload.

Suppose a ten-coin toss returns heads, heads, tails, heads, heads, heads, tails, tails, heads, tails. That is six heads and four tails, so the session reads 60 percent heads and 40 percent tails after ten flips, and the longest run is the three consecutive heads at positions four, five and six.

Many people find that suspicious, but it is entirely ordinary. Across ten fair flips there are 1,024 equally likely sequences, and only 252 of them contain exactly five heads, so an exact even split happens less than a quarter of the time. Getting six heads or more occurs in about 38 percent of ten-flip runs.

The probability of any one specific sequence, including all ten coming up heads, is one in 1,024. That is why long streaks appear more often than intuition expects once you keep flipping: over a few hundred flips, a run of six or seven identical results becomes likely rather than remarkable.

Keep flipping and the percentage drifts towards 50, but slowly. After a hundred flips a split of 55 to 45 is unremarkable, and it typically takes thousands of flips before the ratio settles close to even.

The streak counters are the most instructive part of this tool, because they make visible something people consistently get wrong. A long run of heads does not make tails more likely on the next flip. The coin has no memory, and every flip remains an independent event with a 50 percent chance either way. Believing otherwise is the gambler's fallacy, and it is the single most common misreading of these statistics.

The heads percentage behaves the same way. It converges towards 50 percent over a large number of trials not because early imbalances are corrected, but because they are progressively diluted by the growing sample. An excess of ten heads after 50 flips is a 20 percentage-point skew; the same absolute excess after 5,000 flips is a skew of 0.2 points.

For decision-making the result is as fair as a physical toss. If you find yourself wanting to flip again after seeing it, that reaction has told you your real preference more usefully than the coin did.

This uses a pseudorandom generator, which is ideal for games, demonstrations and casual decisions but is not cryptographically secure, so do not use it for anything requiring unpredictable secrets.