Paste your class list once, then re-group in one tap for the rest of the term. The demo roster below is a 25-name class — replace it with yours, and use the Pairs and Groups of 4 preset chips instead of fiddling with settings mid-lesson.
25 names entered · lines without a weight count as weight 3
Pure shuffle — every split equally likely.
Everything runs in your browser — the names you type are never sent to a server. The share link encodes the roster in the URL itself, so only share it with people who may see the list.
The awkward part of grouping a class is never the shuffle — it is the remainder. This table shows the exact breakdown the generator produces for typical class sizes, computed by the same code that runs the tool. Sizes never differ by more than one, so no student lands in a conspicuously small group.
| Class size | "Pairs" preset produces | "Groups of 4" preset produces |
|---|---|---|
| 20 students | 10 × 2 | 5 × 4 |
| 22 students | 11 × 2 | 4 × 4 + 2 × 3 |
| 24 students | 12 × 2 | 6 × 4 |
| 25 students | 12 × 2 + 1 × 1 | 4 × 4 + 3 × 3 |
| 27 students | 13 × 2 + 1 × 1 | 6 × 4 + 1 × 3 |
| 28 students | 14 × 2 | 7 × 4 |
| 30 students | 15 × 2 | 6 × 4 + 2 × 3 |
| 32 students | 16 × 2 | 8 × 4 |
Read "6 × 4 + 1 × 3" as six groups of four plus one group of three. When pairs leave one student alone (25 students → 12 pairs + 1 single), most teachers merge the single into a pair to form a three — one manual move, sizes stay sensible.
A science class of 27 needs benches of 4. Set team size to 4: the generator computes ⌈27 ÷ 4⌉ = 7 groups and deals sizes 4, 4, 4, 4, 4, 4, 3 — six full benches and one bench of three. The three-bench is the natural place for the equipment-monitor role, which is a nicer framing than "the leftover group". Re-generate next week and the trio will be three different students — the remainder rotates with the shuffle, so nobody is repeatedly stuck in the small group by design.
Grouping by seating chart or register order feels neutral but is not: friends sit together, so adjacency grouping recreates the same social clusters every time. A fresh shuffle each lesson is the cheapest classroom-management win available — it normalizes working with everyone and removes any suspicion that groups were engineered. When you do want engineered groups (mixed ability), the balanced mode's snake draft does it quietly: the 5-a-side picker shows the same mechanism in a sports context, and it works identically for reading levels.
Running a class gift exchange? The Secret Santa generator guarantees no student draws their own name. For adult offsites and hackathons, see work project groups, or go back to the full random team generator for every option in one place.
Just re-generate — every click is an independent shuffle, so cliques get broken up by default rather than by your intervention. If two students genuinely must be separated, the quickest workaround is to run the draw, then swap one of them manually with any member of another group; one swap preserves all group sizes.
Group sizes never differ by more than one. 25 students into pairs gives 12 pairs plus one group of 1 — the tool shows that final singleton clearly, and most teachers fold that student into an existing pair to make a three. 25 into groups of 4 gives 7 groups: four groups of 4 and three groups of 3, because the tool evens out the remainder instead of leaving one tiny group. The table on this page shows the exact breakdown for common class sizes.
Yes. Add a discreet 1–5 number after each name ("Ben, 2"), pick Balanced, and delete the weights before projecting the result — the group cards show names only if you copy them as text and remove the numbers. The weights never leave your browser, but do treat them like any other sensitive note about a student: the share link would include them.
Use "Copy share link" once and bookmark the result. The link stores the whole roster after the # symbol, which browsers never send to any server — opening the bookmark restores your class list instantly, ready to re-shuffle.
Once the page has loaded, all shuffling happens locally in the browser, so a dropped connection mid-lesson does not matter. You only need connectivity to load the page in the first place.