The Hidden Cost of the Re-Roll

What a re-roll actually does

A draw produces an outcome. A re-roll produces a second outcome and a choice between them, and somebody has to make that choice — which is precisely the thing the draw was supposed to remove.

If re-rolls happen whenever anybody objects, then the process is: draw repeatedly until nobody objects. That is not a random selection. It is a veto system with a random proposal generator attached, and the person with the strongest preference wins.

It is almost always the same person. Groups have a member who is most willing to say they are unhappy with an outcome, and an unlimited re-roll rule hands them the decision permanently.

Why it feels harmless

Because each individual re-roll is trivially defensible. The wheel landed on something we did earlier; two people were away when we drew; the list had a typo. Each of those is a reason, and each of them is offered in good faith.

The problem is that the reasons only ever arrive when somebody dislikes the outcome. Nobody has ever asked to re-roll a result they were happy with on the grounds that the list had a typo.

So the reasons are real and their arrival is not random, and a rule that accepts good reasons accepts them asymmetrically.

The rule that works

State before the draw what would justify a re-draw, and accept nothing else afterwards. Almost always the answer is: a genuine error in the inputs, and nothing about the output.

"We will re-draw if a name was missing from the list" is a rule. "We will re-draw if it seems wrong" is not a rule, it is a description of who has the most social capital.

For anything public the same logic applies with more force. A giveaway re-drawn because the winner looked suspicious is indistinguishable from a giveaway drawn until the organiser liked the result, and no explanation afterwards fixes that.

The exception worth allowing

There is one legitimate category and it is worth naming so it does not get swept up: a re-draw for a constraint that was always true and was not encoded.

If a drawn restaurant turns out to be closed, or a drawn meeting slot turns out to be a public holiday, that is a genuine failure of the input rather than a dislike of the output. Re-drawing is correct.

The test is whether the objection would have applied regardless of who or what was drawn. "It's closed" passes. "I don't fancy it" does not, and the difference is usually obvious to everyone in the room including the person raising it.

Where the same failure appears with better clothes

The re-roll is easy to spot because it is explicit. The same failure appears in less obvious forms and is much harder to object to.

Choosing which tool to use after seeing what one of them produced. Adding an option to the list because the drawn answer was unappealing. Deciding that this particular draw was informal and the real one is next week. Each of these is a re-roll with a story attached, and each hands the decision to whoever proposed it.

The test that catches all of them is whether the change would have been made if the result had been different. If a rule, a tool or an entry list is being adjusted in a way that would not have happened had the draw gone another way, the draw has already stopped being a draw.

The one place re-rolling is the correct design

There is a technical use of re-rolling that is not this, and conflating them causes confusion, so it is worth separating out.

Rejection sampling — the technique that removes bias from an integer draw — works by discarding results and drawing again. So does the search behind a constrained gift exchange, which shuffles repeatedly until it finds an arrangement satisfying every rule. Both re-roll, both are correct, and neither is what this page is about.

The difference is who decides and on what basis. Rejection sampling discards on a rule fixed before the draw and applied without reference to the outcome's desirability: the byte was in the leftover range, so it goes, regardless of what number it would have produced. A group re-rolling because they dislike the answer is discarding on the outcome itself.

That is the whole distinction and it generalises cleanly. A rule that was written down before the draw and applies to results you would have been happy with is a legitimate rejection criterion. A rule invoked only when the result is unwelcome is not a rule, whatever it is called at the time.

Frequently asked questions