Sampling without replacement means that once you select an item, you remove it from the available choices for the rest of that sample. Draw three numbered cards from a bag without putting any back, and every selected card must be different.
The same idea helps you choose classroom questions, review records or select several participants from a numbered list. This guide explains five practical examples and shows how to use the Our Daily Tools Random Number Generator without accidentally allowing repeat selections.
What is sampling without replacement?
Start with a defined collection, called the population. A sample is the smaller collection you select from it. Without replacement, an item can appear at most once in that sample. With replacement, you return each selected item to the available pool, so it can appear again.
For a simple random sample without replacement, each possible collection of the requested size has the same chance of selection. Penn State’s introduction to simple random sampling explains this definition and the use of numbered records for selecting a sample.
The important distinction is between an item and its label. If one person appears twice in your source list under different numbers, choosing unique numbers will not prevent that person from being selected twice. Prepare the list before you generate the numbers.
1. Choose three participants from 30 names
Suppose you need three different volunteers from a list of 30 people. Give each person one number from 1 to 30, keeping the mapping fixed.
- Open the Random Number Generator.
- Set Minimum to 1 and Maximum to 30.
- Set How many numbers to 3.
- Turn on No repeats (each number once).
- Select Generate, then use Copy result to keep the numbers.
An illustrative result might be 8, 21, 4. Match these numbers to the original list. Each person has a 3 ÷ 30 = 10% chance of inclusion in a properly uniform sample of three. The example numbers illustrate the format; they are not a saved draw.
If the three roles differ, decide beforehand whether the generated order assigns the roles or whether you will allocate them separately. Do not choose the rule after seeing the names.
2. Select eight practice questions from 40
A teacher has 40 numbered questions and wants eight distinct questions for a practice session. Use a minimum of 1, a maximum of 40 and a count of 8, with no repeats enabled.
Both endpoints are included, so the available pool contains 40 − 1 + 1 = 40 numbers. Each question’s inclusion probability is 8 ÷ 40 = 20% under uniform selection.
Random selection does not guarantee balanced topic coverage. If questions 1–20 cover addition and 21–40 cover subtraction, one combined sample might favour either topic. To guarantee four questions from each topic, generate four unique numbers from each non-overlapping range. That is a different design, with a fixed allocation to each section.
For newly generated arithmetic exercises rather than selections from an existing bank, use the Random Math Question Generator.
3. Review ten records from a numbered list
Imagine a team wants to review ten completed requests from a list of 120. Number the eligible requests from 1 to 120 and generate ten unique numbers.
Use row positions from the prepared list rather than raw record IDs if those IDs contain gaps. A range from ID 100 to ID 250 would include every integer between those endpoints, including IDs that may not exist.
Write down the list version, selection date, minimum, maximum, count and returned numbers. The tool supplies numbers, but your retained list explains which records those numbers identify. This also keeps a later sort or newly inserted row from changing the meaning of the result.
You can enter only the numerical range into the tool; there is no need to paste names or record contents. A random sample is a selection method, not a guarantee that every issue or category will appear in the review.
4. Understand how probabilities change after a draw
Consider ten cards numbered 1–10. Four cards are blue and six are red. If you draw two cards without replacement, what is the probability that both are blue?
The first card has a 4 out of 10 chance of being blue. If it is blue, only three blue cards remain among nine cards. Therefore:
Probability of two blue cards = 4/10 × 3/9 = 2/15 ≈ 13.33%
With replacement, the first card goes back before the second draw. The corresponding probability is 4/10 × 4/10 = 16%.
These answers differ because removing a card changes the pool. Without replacement, consecutive draws are dependent. Do not automatically reuse the first probability for every later draw. Use the Percentage Calculator to express a resulting fraction as a percentage. Our fraction-to-percentage guide explains that conversion step by step.
5. Avoid duplicates across separate batches
Suppose you choose five numbers from 1–50 today, then generate five more tomorrow. The tool’s no-repeat setting prevents duplicates within each result. It does not retain an exclusion list between separate generations.
If you need ten distinct selections overall, the simplest approach is to generate all ten together with no repeats, save that result and split it into two batches of five. If you must select later, prepare a new numbered list containing only the remaining eligible items, then select from that list.
For example, after choosing 5 of 50 items, 45 remain. Renumber those remaining items 1–45 before using that new range. Merely lowering the maximum from 50 to 45 while keeping the old mapping would exclude some eligible items and could leave earlier selections in the pool.
Checks before generating unique numbers
- Count the available integers: maximum − minimum + 1. The tool requires the minimum to be lower than the maximum.
- Keep the requested count within the pool: you cannot choose 12 different numbers from 1–10. This tool supports up to 100 numbers in one result.
- Use a complete, deduplicated list: missing or repeated entries change who can be selected.
- Keep the result: the tool does not provide a persistent selection history or a seed field for replaying the same sequence.
The implementation uses the browser’s crypto.getRandomValues() interface. MDN documents this as a source of cryptographically strong pseudorandom values. That technical property does not replace a clear eligibility list and a consistent selection procedure.
Frequently asked questions
Is sampling without replacement the same as no repeats?
For selecting unique numbered items within one sample, yes. Each selected item is removed from that sample’s available pool. Separate samples can still contain some of the same items.
Can a random number generator with no repeats produce the same result again?
Yes. A fresh generation starts another selection. No repeats within a batch does not mean that numbers or entire results can never recur in future batches.
Does random sampling without replacement guarantee a representative sample?
No. A particular random sample can still be uneven. Define the population carefully, choose an appropriate sample size and decide whether you need fixed selections from particular categories.