Only the previous output is excluded
The first choice is uniform among all N entered names. For every later position, only the person just selected is temporarily removed. With four names, each later choice is uniform among three names. A person can return after one different result, so Avery, Casey, Avery is valid; Avery, Avery is not. This is a transition constraint, not a no-repeat draw.
Four names, six positions
With Avery, Blake, Casey, and Dev, one valid output is Avery → Casey → Dev → Casey → Blake → Avery. After Avery, the next candidates are Blake, Casey, and Dev. Casey occurs twice because the tool does not remove a name forever. With only two names, the sequence necessarily alternates after its first selection. One name cannot produce a second valid position, hence the two-name minimum.
What probability changes
An independent picker gives any particular person chance 1/N on every trial and permits a repeat pair with chance 1/N. Here a person's next chance is zero when that person was just picked, and 1/(N−1) otherwise. The people remain symmetric over many runs, but a short sequence need not give equal totals. Five appearances for one person in 16 selections can be entirely valid.
Choose the rule that matches the event
Use this for turns, speakers, prompts, or games where an immediate repeat feels unhelpful. State the desired sequence length before generating it. If every person must receive exactly four turns, use a shuffled round or duty rota instead. If certain pairs cannot follow each other, that is a larger rule than this page's one-step exclusion.
Roster errors are probability errors
Duplicates would give extra chance to a name, so the page rejects them. Give identical first names distinct labels. Blank lines do not count. Adding or removing someone after observing output creates a new population and therefore a new experiment. Do not rerun merely because totals look uneven: equal totals were never the stated constraint.
Scope of the result
A normal picker makes independent choices; a unique-number tool removes every selection permanently; this page removes only the immediately previous selection. That narrow difference is the reason to use it. It runs in the current browser tab and does not transmit the list or result. It is for transparent low-stakes sequence-making, not consequential allocation.
For a quick audit, copy the sequence and mark every neighboring pair. The fact to test is local: each position after the first must differ from the one immediately before it. This check remains valid regardless of which participant appears most often, so it separates the stated constraint from an unstated desire for perfectly even turns.
Inspect transitions, not just totals
A useful check reads adjacent pairs. For the output Avery → Casey → Dev → Casey → Blake → Avery, compare positions 1–2, 2–3, and so on: no pair contains the same name twice. That is the complete promise. Counting how often each name appears answers a different question and should not be used to reject an otherwise valid result.
The restriction also changes how a sequence feels. When the list has three names, a selected person has two possible successors, so patterns such as A → B → A arise naturally. They are not evidence that the tool is trying to alternate or favour a participant. With a large roster, the exclusion has a smaller visible effect because N−1 remains close to N. This makes the setting useful when the social concern is specifically the immediate repeat.
For an event that spans several sessions, record the last selected name only if the no-repeat rule is meant to cross the session boundary. Starting a new sequence resets the rule because its first selection has the full roster available. Decide that boundary in advance. A copied output can be checked later without needing the browser that generated it, because the constraint is visible in the sequence itself.