Coin flip vs dice vs random number: which is actually fair?
"Fair" means every allowed outcome has the same probability. A fair coin is 1/2 heads, 1/2 tails. A fair d6 is 1/6 per face. A fair integer from 1 to 100 is 1/100 per value.
A coin is not fairer than a die. A die is not fairer than a random number generator. They are different outcome counts. People argue about "which feels more random" when they should ask "how many outcomes do I need, and is each one equally likely?"
What fairness means (and what it does not)
Fairness (uniformity): for a set of size N, each element has probability 1/N.
Independence: this flip/roll does not remember the last one. Streaks of heads are normal. They are not "due" to end.
Cryptographic strength: whether an attacker can predict future outputs. That matters for passwords and secrets. It barely matters for "who picks the movie" or a classroom cold-call.
Most decision and game uses need uniformity and independence. They do not need a hardware entropy lecture.
Coin flip: two outcomes
Ideal math: P(heads) = P(tails) = 1/2.
Real coins can be slightly biased by weight, wear, and spin technique. For settling a binary choice among friends, the bias is usually smaller than the social cost of arguing. For a contest with money on the line, publish the method and prefer a visible digital flip or a numbered draw everyone can audit.
Use a coin when you truly need yes/no, A/B, or two people / two options.
Try: coin flip or random 1 to 2 (same two-outcome idea as a number).
Dice: N faces
Ideal math: P(face) = 1/N for faces 1…N.
Physical dice vary. Casino dice are manufactured and inspected to tight tolerances. Cheap opaque dice can be off enough that serious games care. For board game night, agree on the die (or the app) and move on.
Dice shine when the game’s rules already speak die language: d20 checks, 2d6+3 damage. That is notation plus fairness. See dice notation explained.
Try: dice roller, d6, d20.
Random number generator: any range
Ideal math: pick integers from min…max inclusive, each equally likely (when that is what you asked for).
A good implementation maps an underlying random source onto that range without favoring low numbers. A bad implementation uses skewed remainder tricks or reseeds poorly. For casual tools, browser Math.random()-style sampling is typically fine for games and decisions. It is not what you should use to generate passwords or security tokens. Use a purpose-built password generator that documents a CSPRNG when secrets matter.
Use a number generator when:
- N is not 2 or a standard die size
- You need unique draws without replacement (raffle tickets 1…N)
- You need exclusions ("skip seat 7")
Try: homepage generator, 1 to 100, or any preset range.
Side-by-side
| Tool | Natural N | Best for | Weak when |
|---|---|---|---|
| Coin | 2 | Binary decisions | You quietly needed 3+ options |
| Die | 4–20 (typical) | Games with die rules | You force a d6 to pick among 10 people |
| RNG / list picker | Any | Raffles, rosters, custom ranges | You wanted the ritual of a physical flip |
Wheel spinners and name pickers are RNGs with a friendlier UI: they still pick uniformly from the options you entered (when unique mode is off). See wheel spinner and name picker.
"But I flipped five heads in a row"
That sequence has probability (1/2)^5 = 1/32 going in. After you already saw it, it is just history. The next flip is still 1/2 if the process is fair. Hot-hand stories feel meaningful because brains are pattern engines. Lottery versions of the same mistake are covered in lottery odds explained.
Which should you use?
- Two options → coin (or 1–2).
- Tabletop rule says NdX → dice (match the notation).
- People on a list / tickets 1…N / custom range → number generator, name picker, or wheel.
- Public giveaway → published method + frozen list + screenshot, not vibes. Fair raffle guide.
- Passwords / secrets → CSPRNG password tool, not a d20.
Fairness is matching the tool to the outcome set, then sampling uniformly. Ritual is optional. Math is not.
FAQ
Is a coin flip really 50/50?
An ideal fair coin is. Physical coins can be slightly biased. For casual decisions the difference is usually tiny. For high stakes, use a transparent digital method or a numbered draw.
Is an online dice roller fair?
It is fair if each face 1…N is equally likely and rolls are independent. That is a software property, not a guarantee from the word "dice" on the button. Agree on the tool at the table.
Is Math.random good enough for decisions?
For movies, chores, classroom cold-calls, and board games: usually yes. For cryptography, session tokens, or anything an attacker should not predict: no. Use a CSPRNG-backed tool for secrets.
Coin vs wheel vs name picker for a group decision?
Coin if there are exactly two choices. Wheel if you want a visible spin over a handful of options. Name picker if the options are people on a roster and you may need unique draws.
Disclaimer: Educational comparison only. Physical manufacturing tolerances and browser RNG implementations vary. Tools on this site are for personal decisions, games, and practice. They are not certified gambling devices and are not substitutes for regulated drawing equipment where the law requires it.