Directed feedback with equal review counts

Peer Review AssignmentGenerator

Give each person the same number of reviews to write and receive. Exclude self review, optionally prevent mutual pairs, and avoid supplied past assignments.

Assign peer reviewers Read the guide

Free • Local processing • No account

Five clay essay pages with different portrait seals linked by one-way review arrows and annotation notes

Plan who reviews whose work

Every listed person is both a reviewer and one author with one submission.

Authors and reviewers

One person and one submission per line. 2–40 people, 120 characters per name, 10,000 input characters. Matching names keep separate IDs.

UTF-8 TXT up to 40,000 bytes per file within text limits. Roster edits or imports clear all past links.

Person ID reference
  • Person 1: Alex
  • Person 2: Blair
  • Person 3: Casey
  • Person 4: Drew
  • Person 5: Emery
  • Person 6: Finley

Reviewer → author is directional. Everyone writes and receives the same review count. No self review is ever allowed.

6
People
—
Directed reviews
—
Mutual pairs

1–10, fewer than people. Everyone writes this same count. Without mutual pairs, 2 × count must be at most people − 1.

One reviewer ID,author ID per line, e.g. 1,2. Comma or Chinese comma. Directed hard exclusions; reverse links remain possible. Duplicates ignored. Up to 10,000 characters.

Only complete balanced plans are accepted. History plus nonreciprocal rules uses exact branching up to 2,000 nodes or about 3 seconds; a limit is not proof of impossibility.

Six people are ready for two reviews each, without mutual pairs.

Who each person reviews

Generate to see outgoing review work and incoming coverage.

Names, rules and results stay in this tab and local worker without uploads or persistent tool storage. This organizer view is not anonymous. Export before refreshing.

A review relationship has a direction

Check what each reviewer writes and what each author receives.

Reviewing a classmate is different from pairing up

A peer review assignment generator builds directed links from reviewers to authors. Alex reviewing Blair does not automatically mean Blair reviews Alex. Every person on this page represents one author with one submission and also a reviewer. The requested count applies in both directions: each writes that many reviews and each submission receives that many different reviewers. Self review is always excluded. This creates a complete workload and coverage plan rather than a list of undirected partners or a random sequence of names.

Try the review sample
Students in a library writing comments on different classmates’ essays while a teacher checks a reviewer list

Mutual pairs and prior links are separate rules

Prevent mutual reviews forbids a two-person exchange within the new plan, such as Alex reviewing Blair while Blair reviews Alex. Past assignments are directional exclusions supplied in the history field. Excluding an old Alex-to-Blair link does not automatically exclude Blair-to-Alex in a later plan. Both rules are enforced when enabled or supplied, never quietly relaxed. A three-person or longer directed cycle is allowed: preventing mutual reviews does not forbid all feedback cycles or claim that nobody knows their reviewers.

Inspect workload and coverage
Teacher’s routing clipboard connecting reviewer portrait markers to essay icons with one-way links and matching coverage checkmarks

How to assign peer reviewers

Finish the roster before entering history by numeric ID.

  1. 01

    List the authors who also review

    Paste one person per line or import a UTF-8 TXT roster. The page supports two to forty people, names up to one hundred twenty characters and ten thousand input characters. Blank lines are ignored, while matching names remain separate numbered people. Use the ID reference to distinguish them. Everyone must have one submission available and be able to complete reviews; this tool does not maintain a separate reviewer-only pool, handle several papers per person or check who actually submitted work.

  2. 02

    Choose a count and whether mutual reviews are forbidden

    Set one to ten reviewers per submission, always fewer than the number of people. Each person will also write that many reviews. With mutual reviews forbidden, twice the requested count must be at most the number of other people, so a count of two needs at least five people. This is a necessary size rule, not proof that a history-constrained plan exists. Turning the option off permits reciprocal pairs while preserving no self review, complete coverage and equal outgoing counts.

  3. 03

    Supply past links and generate a complete plan

    Enter one reviewer ID,author ID pair per history line, using a comma or Chinese comma. Repeated history rows are ignored; unknown IDs and self-review rows are rejected. These supplied links are hard exclusions. Generate and inspect both directions before exporting. Editing or importing the roster clears all history and the old result so numeric IDs cannot shift silently. Other input changes also clear the plan and cancel its worker. No partial assignment is accepted when the full set of requirements cannot be satisfied.

Balanced workload and clear solver outcomes

A successful result passes explicit structural checks.

Exactly the requested counts in both directions

Every successful plan contains the roster size multiplied by the requested count in directed review assignments. Each reviewer has exactly that many different authors, and each author has exactly that many different reviewers. The tool checks the completed edge list for self links, duplicate links, past links, count mismatches and forbidden reciprocal pairs before accepting it. Equal counts describe the number of planned reviews, not equal reading time, essay length, expertise, feedback quality or eventual completion. Those aspects still need the teacher’s judgment and a suitable rubric.

Different methods for unconstrained and historical plans

Without history, a shuffled cyclic construction produces a complete regular assignment and avoids opposite offsets when mutual reviews are forbidden. It samples a useful family of valid plans rather than uniformly sampling every possible directed graph. With past links, capacity matching checks exact incoming and outgoing counts. If mutual pairs are also forbidden, bounded branching removes one direction of each conflicting pair and repeats the matching check. The method and searched node count are displayed, without an unsupported promise that every valid outcome has equal probability.

Impossible and search limit are different results

A person with fewer allowed outgoing or incoming options than requested receives a specific ID warning. A capacity bottleneck or an exhausted exact search can prove that no complete plan fits the entered rules. Separately, search stops at two thousand nodes or approximately three seconds and reports an unconfirmed outcome, not mathematical impossibility. Cancel also discards the calculation. Reduce the review count, review the supplied history or change the mutual-pair rule as appropriate; a slow or failed attempt never produces a misleading almost-complete assignment.

Example: six authors, two reviewers each

Twelve directed reviews can be checked from either side.

  1. 01

    Read a balanced directed cycle construction

    Use Alex, Blair, Casey, Drew, Emery and Finley, request two reviews each, and leave history blank with mutual pairs forbidden. One valid plan sends each person’s work to the reviewers two and one positions behind them in a six-person cycle. Equivalently, each person reviews the next two authors. All six write two reviews and receive two, for twelve links in total. There are no self links or mutual pairs in that example. The shuffled order and chosen offsets may produce a different valid result.

  2. 02

    Understand a shared bottleneck despite enough individual options

    Use three people, a count of one and history exclusions 1,2 and 2,1. Person one can only review person three, and person two can also only review person three. Both have at least one allowed target, but person three needs exactly one incoming review, so no balanced complete assignment exists. This is more than a shortage in a single person’s option count. The tool stops with a no-solution message rather than giving person three two reviews or leaving another submission uncovered.

Use the allocation for useful feedback

The plan allocates work; it does not assess the work.

  1. 01

    Keep a usable rubric and a realistic review count

    Decide what students should comment on and how much reading time is available before setting the count. Several independent perspectives can be useful, but this tool does not recommend a universally correct number of reviewers or measure grading accuracy. Authors and reviewers are the same complete roster here. If someone is absent or lacks a submission, revise the actual list and rebuild the assignment instead of deleting a few links from the result, which could break workload balance and coverage.

  2. 02

    Carry the right historical links forward

    Download current links as a plain reviewer-ID,author-ID TXT file or choose Use this plan as past links. The latter replaces the history field with the current plan and clears the result so you can generate the next one. It does not silently accumulate all previous rounds. You may paste several rounds’ links together if you want to avoid all of them; duplicate rows are ignored. Keep the same roster order and identities. Merging more past links can eventually make a new balanced round impossible.

  3. 03

    Export a complete organizer view without claiming anonymity

    Copy and print include the reviewer-to-author list and the author-to-reviewer coverage list. CSV contains every directed edge with IDs and names; its four columns can be read in either direction. Formula-leading names are protected for spreadsheet use. The page does not upload essays, send review requests, collect feedback, grade work or hide names from participants. All inputs and results remain in this tab and local worker without persistent tool storage. Export before leaving. For actual undirected partners, use Random Pair Generator instead.

Peer review assignment questions

Answers about direction, workload, history and feasibility.

How do I randomly assign peer reviews without self review?

Enter the full list of people who will both submit and review, set the reviewers-per-submission count, and generate. The page always excludes self review. A successful plan gives every person exactly the same number of outgoing reviews and incoming reviewers, with each directed link appearing once.

What does preventing mutual or reciprocal reviews mean?

It forbids A reviewing B and B reviewing A within the same new plan. Longer cycles are allowed. The count must satisfy twice the reviewers per person being no greater than the number of other people. History can still make a size-valid configuration impossible; the tool never relaxes the rule silently.

How can I avoid reviewers repeating the previous round?

Paste or import the previous directed ID pairs in the history field. Every supplied reviewer-to-author link is excluded from the next plan. Its reverse is not automatically excluded. Use the current plan as past links to prepare another draw, or combine several rounds manually; the tool does not automatically save or accumulate history.

Will review workload and submission coverage be balanced?

Yes, in every accepted plan both counts equal your requested number for every person. Equal review counts do not guarantee equal reading effort or feedback quality. This tool treats each roster entry as one author, one submission and one reviewer, without separate pools or per-person workload settings.

Why can no solution exist when everyone has some available reviewers?

Several people may depend on the same limited targets or incoming slots. Individual option counts are necessary but not sufficient. Past links and the nonreciprocal rule can also conflict. A proven failure is labeled impossible; reaching the bounded search budget is labeled unconfirmed, and no partial result is accepted.

Can an odd number of students use the tool?

Yes. Directed reviews do not require the roster to split into pairs. Three people can form a one-review cycle without mutual pairs; five can each write and receive two reviews. Two people need reciprocal permission for one review each. The size and history restrictions still apply to every request.

Can I export both who reviews and who is reviewed by whom?

Copy and print include both complete views. CSV exports all directed reviewer-author rows with IDs and names, and the history TXT contains the numeric pairs for reuse. Scrolling the on-screen panels does not truncate exports. There are no private reviewer links, email delivery or learning-platform import guarantees.

Are duplicate names allowed, and is the plan saved?

Matching names remain separate IDs, so use labels that identify the actual people. Changing or importing the roster clears history to prevent ID mismatches. UTF-8 TXT files can be up to 40,000 bytes within the text limits. Inputs and results are not uploaded or persistently saved by this tool; export before a refresh.

Peer Review Assignment

Give every author a complete feedback plan

Choose a realistic review count and clear rules.

Check both directions before sharing the assignment.

Assign reviewers Explore all tools
  • Free to use
  • No signup
  • Runs in your browser

A simple path to your next activity

  1. List the authors

    One available submission and reviewer per person.

  2. Set the rules

    Review count, mutual pairs and supplied past links.

  3. Check and keep

    Copy, print or export both review directions.

Equal counts, no self review and explicit history exclusions; no partial plan is accepted.