Wheel, Dice, Hat or Shuffle: Picking the Right Randomiser

Any of these will make a random choice: a coin, a dice, a name in a hat, a spreadsheet formula, a spinning wheel. Mathematically there is nothing to choose between a fair coin and a fair wheel with two names on it. The real differences are practical ones. How many options each handles, whether unequal odds are possible, how long it takes, and above all what an audience can verify while it is happening.

That last property is the one people underrate, and it is usually the reason a method feels wrong even when the maths is perfect.

The coin flip

Unbeatable for two equally weighted options when you and one other person are in the room. It is instant, it needs no equipment anybody has to trust, and both people can watch the whole event from start to finish.

Its limits are hard ones. Two outcomes only, and “flip twice” to get four works but confuses people fast. No way to give one side better odds without inventing a convention nobody remembers. And it does not scale past the people physically watching: a coin flip reported over chat is just an assertion.

Use it when: two options, two people, decision needed in three seconds, nothing much at stake.

The dice roll

A dice has a genuinely useful property that most digital tools lack. It is self-evidently fair to everyone looking at it. Nobody needs to trust the manufacturer, read the source code, or believe a claim about entropy. The mechanism is visible, and that is worth a lot in a room of sceptical people.

The cost is rigidity. Six options, or twenty if you own the right dice, and never the number you actually need. Mapping eleven people onto a twelve-sided dice means agreeing to reroll on a 12, which works but needs explaining every single time. And as with a coin, it convinces only the people in the room.

Use it when: the count happens to match the dice, everyone is physically present, and you want a mechanism nobody can question.

Names in a hat

The oldest method and still one of the most persuasive, for exactly the same reason as the dice: the audience can inspect it. They can see the slips go in, see the hat shaken, and see a hand pull one out. Weighting is trivial and intuitive, since you just put somebody’s name in three times.

The weaknesses are physical. Folded slips are not actually well mixed, larger slips are easier to grab, and a practised hand can find a particular one; stage magicians have made a living from this. It takes real time to write out forty names. And when it is over you have nothing but a memory, with no record of what was drawn, in what order, or when. If somebody disputes the result a week later, there is nothing to point at.

Use it when: everyone is in the room, the list is short, and the occasion suits something tactile.

A spreadsheet or a shuffled list

Sorting a column by RAND() is the most efficient method on this page for large lists. Ten thousand entries, one formula, done. And you get a complete ranked order rather than a single winner, which is exactly what you want for assigning a running order or a waiting list.

It is also the least convincing to anybody who was not driving the keyboard. A sorted column shows a result, not a process. Nothing about the screenshot distinguishes a genuine shuffle from a list somebody arranged by hand, and nothing stops a re-sort until the top row looks agreeable. For your own internal use that is fine. For a public draw it asks the audience for trust it cannot earn.

Use it when: the list is long, you need a full ordering, and nobody needs convincing. Internal scheduling, sampling, test data.

The spinning wheel

A wheel is a worse random number generator than a spreadsheet and a slower one than a coin. What it does better than anything else on this page is make a random choice in public, and that is a distinct job.

The odds are visible in the picture. Every name is a slice, and slice size is probability. A weighted entry drawn three times bigger genuinely is three times more likely, and nobody has to be told, because they can see it. No other method on this page displays its own odds.
It scales past the room. A shared link means an audience of any size watches the same spin resolve at the same instant, which a coin, a dice and a hat cannot do at all.
The build-up is the point. Several seconds of deceleration gives an audience time to react. That is a liability when you are picking a student every ninety seconds and an asset when a prize is on the line.
It leaves a record. Only if the tool is built to keep one, and many are not. A draw decided and logged on a server before it reaches any screen is something you can point to a week later.
Any number of options, with unequal odds. Eleven names is as easy as ten, and giving one of them double weight takes a second.

The honest drawbacks: it is slow if you are drawing repeatedly, the labels become unreadable somewhere past 150 names even though the draw itself is unaffected, and the audience is trusting software rather than physics. The last one is real. It is why where the pick happens matters. See how the live spin works.

Use it when: more than two options, an audience that wants to see it happen, odds that may not be equal, or a result somebody might question later.

Choosing in one line each

Two options, two people, right now: coin.
Everyone in the room, sceptical, count fits the dice: dice.
Short list, in person, want something tangible: hat.
Long list, need a full running order, nobody to convince: shuffle a spreadsheet.
An audience, unequal odds, or a prize involved: wheel.
Nobody can attend and you need it above suspicion: a scheduled wheel that spins itself at a published time, with nobody at the keyboard.

The question underneath all of this

Before picking a method, work out whether you need a random result or a demonstrably random result. They are not the same requirement and they have different answers. Deciding who reviews a pull request needs the first, and a spreadsheet formula is the efficient choice. Drawing a winner for something people entered needs the second, and no amount of cryptographic rigour substitutes for the audience having watched it happen.

Most disagreements about fairness are not really about randomness at all. They are about evidence. Pick the method that produces the evidence your situation actually calls for.

How random number generators actually work, covering pseudorandom versus cryptographic, and modulo bias.
How to run a raffle people trust, on the process around the draw.
Using a wheel as a decision maker.