It sounds like an abdication. It is not, for a specific and narrow class of decisions: the ones where the options are close in value and the deliberation costs more than the difference between them.
If two restaurants are roughly equally good, twenty minutes of discussion is a pure loss. If two features are both defensible, a week of debate costs more than picking the slightly worse one and shipping. The philosophical version is Buridan ass — an animal placed exactly between two identical bales of hay starves while deciding. The practical version is a coin.
There is also a genuinely interesting empirical finding here. An economist ran a large experiment in which people facing real dilemmas — quit the job, end the relationship, move city — were assigned an outcome by coin flip. Among those who followed it, the ones who made the change reported higher happiness six months later than those who stayed. The interpretation the author offered is that people are systematically biased toward the status quo, so when a decision is genuinely marginal, the change is more often the better bet than we feel it is. That is one study on a self-selected sample, so hold it loosely — but it suggests that if you are agonising, inertia is not the neutral option it appears to be.
Anything irreversible, anything high-stakes, anything where you have not actually gathered the available information, and anything where the options are not close. Randomising a medical decision, a large financial commitment or a hire is not decisiveness. And critically: a random pick does not legitimise a bad list. If every option is poor, the tool returns a poor option with a clean conscience attached.
A useful test before using it: if the coin gives you the answer you did not want and you feel relief rather than resistance, the decision was genuinely marginal. If you immediately want to flip again, it was not marginal at all — you already knew, and what you wanted was permission.
Two technical issues determine whether a picker is genuinely unbiased.
The source. The standard Math.random in browsers is a pseudo-random generator: fast, statistically fine for choosing lunch, but deterministic and not suitable where an adversary might predict it. For anything where fairness must be defensible — a prize draw, a raffle — a cryptographic source such as crypto.getRandomValues is the correct choice.
Modulo bias. This is the subtle one. To pick from 6 options, code often takes a large random integer and applies the remainder operator. But if the generator range is not an exact multiple of 6, the low-numbered outcomes occur slightly more often than the high ones, because the leftover partial block at the top of the range only covers some of the options.
With a 32-bit range and 6 options the bias is around one part in 700 million — utterly irrelevant for a dice roll. With a small range or many options it becomes measurable. The proper fix is rejection sampling: discard any value that falls in the incomplete final block and draw again. It costs an occasional extra draw and produces a perfectly uniform result.
Sometimes you want unequal odds — a 3 in 10 chance for one option. The standard technique sums the weights, draws a number in that total range, then walks the list subtracting weights until the remainder goes negative:
| Option | Weight | Cumulative | Range |
|---|---|---|---|
| A | 5 | 5 | 0 – 4.99 |
| B | 3 | 8 | 5 – 7.99 |
| C | 2 | 10 | 8 – 9.99 |
A draw of 6.4 lands in B range. Weights do not need to sum to any particular number, since they are normalised by the total.
Randomness has no memory. Five heads in a row does not make tails more likely on the sixth — that is the gambler fallacy, and it is the single most persistent misunderstanding about random processes. A fair picker will produce runs and clusters that feel wrong; genuinely random sequences look far lumpier than people expect, which is why deliberately spread-out sequences feel more random and are not.
Also, small samples say nothing. Picking the same name three times from a list of five is entirely unremarkable — the probability is 1 in 25 for any specific name, and something surprising happens most of the time.
You can, but then you are not using a random picker — you are using it as a prompt to notice your preference, which is a legitimate but different tool. Decide in advance whether the result binds you.
Not meaningfully, and physical coins have small measurable biases from their weight distribution and the spin imparted. A cryptographic generator is more uniform than any coin.
Shuffle the whole list with a Fisher-Yates shuffle and take the first n. Repeatedly drawing and rejecting duplicates works but gets slow as the list empties.
No. The list you enter is used only in your browser and nothing is transmitted or saved.
Every tool comes with a written guide, and every category is one click away.