Coin Flip

The physical coin isn't 50/50 — this one is

If you're using this to settle something, know what a real coin actually does. Persi Diaconis, Susan Holmes, and Richard Montgomery modeled coin-flip dynamics in 2007 and predicted a same-side bias near 51% — the coin tends to land on whatever face was up when it left your thumb, because it spins about an axis that isn't quite perpendicular to the face (Diaconis et al., 2007). A 2023 pre-registered replication with 350,757 flips by Bartoš et al. found 50.8% same-side — close to the theoretical prediction, and well outside chance at that sample size.

This digital coin doesn't have an axis or a starting face. It calls Math.random() < 0.5 and maps the boolean to heads/tails, so the distribution is uniform to the precision of V8's xorshift128+ PRNG — not cryptographic, but statistically fine for a binary pick. The edge case: if you flip fewer than ~30 times, short runs of 6+ in a row will look "broken" even though the generator is working correctly. For 10 flips, there's a 1 in 1,024 chance of all-heads, which means in 1,000 visitors a few will see it and assume the tool is rigged. It isn't — small samples just look streaky.