Overview
Coin flip: an overview of the format and its maths
This is the starting point for every coin flip page on the site. It explains the format once, sets out the model the other pages use, and points to the guide that covers each question in depth.
The format
A coin flip game asks one question: heads or tails? The player names a side and a stake. The result is drawn, and a correct call is paid at a fixed multiplier, while a wrong call loses the stake. There are no partial outcomes, no ties, and no decisions after the call. That simplicity makes the coin flip the clearest place to study how a house edge works, because nothing else gets in the way.
Online services have no physical coin. They turn a number into a side. Many use a hash-based method: a secret server seed, committed to in advance by publishing its SHA-256 hash, is combined with a client seed and a bet counter called a nonce. The output is mapped to a number between 0 and 1, and values below one half count as heads.
The model used on this site
Every coin flip page here uses the same generic teaching model. It is not any operator's rule book:
- The coin is fair: heads and tails each have a 50% chance on every toss, independent of earlier tosses.
- A correct call returns 2 x (1 - e) in total, where e is the house edge. With a 1% edge, that is 1.98x.
- A wrong call returns nothing.
- The expected net result per unit staked is therefore -e.
The table shows how the payout and the expected cost move as the edge changes. Each row is computed in code from the formula above.
| House edge | Payout on a win | Expected net per unit | Expected net per 1,000 units staked | SD per flip |
|---|---|---|---|---|
| 0.0% | 2.000x | 0.00% | 0.0 | 1.0000 |
| 0.5% | 1.990x | -0.50% | -5.0 | 0.9950 |
| 1.0% | 1.980x | -1.00% | -10.0 | 0.9900 |
| 2.0% | 1.960x | -2.00% | -20.0 | 0.9800 |
| 3.0% | 1.940x | -3.00% | -30.0 | 0.9700 |
| 5.0% | 1.900x | -5.00% | -50.0 | 0.9500 |
An edge of 0% is the fair-coin benchmark. Real services choose their own edge and should state it.
Two things stand out. First, the standard deviation barely changes with the edge; it stays close to one unit. The edge is tiny next to the swing of a single flip. Second, the expected cost scales directly with what you stake. Over 1,000 units of turnover a 1% edge costs about 10 units, and a 5% edge about 50.
What the detailed guides cover
Each guide page takes one question and works through it with its own numbers. The theme word in each address is only a label; every page describes the same game.
- Why 1.98x and not 2x: the payout derivation and where the missing 0.02 goes.
- Variance and the square-root rule: how far results stray, and when the edge overtakes luck.
- Streaks in 10, 50 and 200 flips: exact run probabilities.
- The expected longest run: a benchmark for judging whether a sequence looks random.
- Letting a win ride: how the edge compounds over a chain of flips.
- Myths and the gambler's fallacy: due results, hot sides and evening out.
- The martingale: why doubling fails, and why a 1.98x payout breaks it sooner.
- Bankroll and risk of ruin: exact chances of losing a whole bankroll.
- Hash verification: checking eight tosses from their seeds.
- A seeded simulation: what sessions look like next to theory.
Before you read on
Nothing in these guides offers a way to beat the game. Every result above follows from one fact: with any edge above zero, each flip has a negative expected value. The guides explain the shape of that loss, how it varies, and how to check that the results are honest. For the idea of an edge across all game types, see the house edge guide, and for the hashing method in general, the hash fairness guide. If play stops feeling like a choice, the responsible gambling page is a good next step.