How fair the simulator is
Each flip uses one bit drawn from crypto.getRandomValues. There's no bias toward heads or tails — over many flips the ratio converges to exactly 50/50 (within statistical fluctuation). Real physical coins have a slight bias (~51% same-side per the famous Stanford study), so the simulator is actually more fair than a real flip.
Why streaks aren't surprising
People intuitively expect "fairness" within short sequences — if you've just seen 5 tails, it feels like heads is "due." This is the gambler's fallacy: each flip is independent, so the next flip is exactly 50/50 regardless of history. Try flipping 1,000 coins and inspect the streak length — you'll often find runs of 8 or 9 of one face. Probability:
| Streak length | Probability per flip | Expected count in 1,000 flips |
|---|---|---|
| 3 in a row | 1/8 | ~125 |
| 5 in a row | 1/32 | ~31 |
| 7 in a row | 1/128 | ~7.8 |
| 10 in a row | 1/1024 | ~1.0 |
| 15 in a row | 1/32,768 | ~0.03 |
So a 10-in-a-row streak is actually expected roughly once per 1,000-flip session. Streaks aren't proof of bias.
FAQ
What's a "lucky number" of flips?
None — every flip is independent, so the prior history has zero predictive power. The "luckiest" flip is the one whose outcome you guess right by chance.
Can I use this for serious decisions?
Sure. The output is a fair coin flip, mathematically equivalent to (and easier than) flipping a real coin.
Why does the bar at the bottom not always sit at 50/50?
Because real samples fluctuate. After 100 flips you might see 47% / 53%; after 10,000 you'll typically be within 1% of 50/50. As the number of flips grows, the proportion converges to 50/50 — the law of large numbers in action.