A Renban Sudoku book punishes a bad file in a way a sample puzzle never shows. Your buyer judges puzzle forty-one, when a purple line of five cells in the key holds 2, 3, 5, 6 and 7 β a set with a hole at 4, which the rules forbid outright β or a grey group of four holds the same digit twice because two of its cells sit in different rows, columns and boxes, where ordinary Sudoku would never have caught the repeat, or a group the listing calls length six turns out to print with only five cells on it, so the forced middle digits the buyer was counting on were never there. Renban is a Sudoku variant, published in two honest forms that are one rule β the modern line form, where a coloured line runs through a set of cells, and the older group form, where a connected set of shaded cells is the group β stated in the same standard form by variant guides, contest booklets and solving sites: place 1 to 9 once in every row, column and box, and let every marked group hold a set of distinct consecutive digits in any order. Underneath the friendly coloured line is a contract in five parts, each part exact. First, the classic rule: every row, every column and every three-by-three box contains each digit from 1 to 9 exactly once. Second, the set rule: the digits inside any one marked group are consecutive β they form an unbroken interval of the number line, like 3, 4, 5, 6 β with no gaps inside the set. Third, the no-repeat rule inside the group: no digit appears twice in one group, even where the group's cells share no row, column or box and classic Sudoku alone would have allowed the repeat. Fourth, the any-order rule: the interval is a set, not a sequence β the digits may sit along the line in any arrangement, and no buyer is ever promised that neighbouring cells on a line differ by one. Fifth, closing the contract, given integrity and uniqueness: the printed given digits and the printed groups are fixed, and exactly one completed grid satisfies everything at once, reachable by deduction alone, with no guessing. A puzzle that breaks any part is not slightly harder. It is false as stated. A good Renban book is a verified set of sets: every group in every puzzle re-checked as a set against its own key, puzzle by puzzle.
Three warnings belong at the front, because they are the mistakes this genre makes most often, and all three start with variants Renban is not. Renban is not a thermometer, even though both draw a line through cells and talk about the digits along it. Digits on a thermometer must strictly increase from the bulb end, in that direction, in that order β gaps are allowed, so 1, 3, 4 is a legal thermometer reading, but 4, 3, 1 along the line is illegal however consecutive the set is. A Renban group has no bulb, no direction and no order: 4, 3, 1 is not a legal Renban set at all because of the gap at 2, while 6, 4, 5, 3 is perfectly legal because the set is whole. A setter who proofs a Renban key with thermometer habits will reject good puzzles and accept broken ones in the same afternoon. Renban is also not a killer cage, even though both mark a set of cells and constrain the digits inside. A killer cage is defined by a stated sum printed in its corner; its digits are not required to be consecutive, and a cage holding 1, 4 and 9 is ordinary. A Renban group prints no sum anywhere β the constraint is the shape of the set itself, an unbroken interval β and a group holding 1, 4 and 9 is simply broken. And Renban is not a consecutive-dot puzzle, and not a whisper line, even though all three use the word consecutive or live near it. A white dot or a consecutive marking is a promise about one pair of neighbouring cells: those two digits differ by exactly one, and it says nothing about any larger set. A German Whisper line is the opposite promise about pairs: neighbouring digits on it must differ by at least five. A Renban line makes no pairwise promise at all β two neighbours on the line may differ by one, by three, or by the full width of the interval β because its promise is about the whole set, all at once. A book must state the standard form plainly, name its form honestly β lines and shaded groups are one rule, not two product lines β and never let a puzzle generated under an in-order, a sum, or a pairwise reading slip in unnamed.
The book is the whole set of groups, not the sample puzzle
A sample Renban puzzle almost always behaves. It is usually built around one or two short groups, and a short group is a gentle thing: a two-cell group is just a consecutive pair, a three-cell group has only seven possible sets in the whole of Sudoku, and the givens usually pin the interval down before the buyer has done any real work. None of that tests the things a full book can get wrong: a group transcribed one cell short in export, so a length-six group prints as five cells and the digits the key spreads across six cells are now demanded of five; a repeat hiding inside a long group whose cells sprawl across three boxes, invisible to every classic check the setter ran; an interval quietly shifted in the key, so the group that the givens force to be 2, 3, 4, 5 is keyed as 3, 4, 5, 6 and the puzzle now has no solution at all, only an argument; or a late puzzle with a second completed grid that honours every group as a set, so the buyer who finds it has followed every printed rule and arrived somewhere your key calls wrong. The buyer does not judge your book by the puzzle on the listing. They judge it on the quiet Sunday evening when puzzle forty-one's long line will not resolve and one stubborn group insists on containing both a 2 and another 2.
That is why the unit of quality in this genre is not the puzzle, it is the verified book. Every group in the file has to carry the same five promises at the same standard: classic Sudoku satisfied in every row, column and box; every group a whole interval with no gap; no repeat inside any group; no order promised or implied anywhere along a line; the printed groups and givens sitting exactly where the key thinks they sit; and one solution only. A book that keeps those promises on ninety-nine puzzles and breaks one has not scored ninety-nine percent. It has taught the buyer that your groups cannot be trusted, and every later stall β including the fair stalls you designed on purpose, where an interval could still be one of two and the buyer must earn the difference elsewhere β will be read as another error.
Sell the book as that promise, plainly. State the grid size and the form you use β lines or shaded groups β and say in plain words that a group is a set: the digits in it are consecutive, they do not repeat, and they may appear in any order along the group, in words a first-time buyer can test on the worked example. State how long your groups run and how difficulty is graded, and show how the key talks one solved group aloud, set by set β here is the group, here are its digits sorted, see that they form an unbroken run with nothing missing and nothing doubled β so the buyer can watch a set being earned. A buyer who can re-check one solved example before page one will not write the review that says the lines cheat. A buyer who discovers on page thirty that the digits were supposed to run in order along the line, or that a group may repeat a digit its cells' boxes never forbade, has been ambushed by a convention you owed them on page one.
The contract, part by part
Start with the classic rule, because it fixes the board everything else stands on. Every row, every column and every three-by-three box of the finished grid contains each digit from 1 to 9 exactly once. This is ordinary Sudoku, unchanged and unrelaxed: the Renban condition is an addition to classic Sudoku, never a replacement for it, and a grid that satisfies every group but repeats a digit in a column is not a Renban solution with a blemish β it is not a solution. State it first on the rules page, because every later check, including the group checks, assumes it.
The set rule then gives the genre its name and its engine. Take any marked group β a line, or a connected patch of shaded cells in the group form β and collect the digits in its cells. Sorted, they must form an unbroken interval: a run of consecutive digits with nothing missing. A group of three cells may hold 1, 2, 3 or 4, 5, 6 or 7, 8, 9, and nothing else; a group of four may hold 2, 3, 4, 5 but never 2, 3, 4, 6; a group of nine cells, if you print one, must hold 1 through 9 exactly. The rule has a consequence worth printing, because it is the buyer's strongest tool and your audit's first arithmetic: the longer a group is, the more of its digits are forced before a single cell is placed. Every five-cell group contains a 5. Every six-cell group contains 4, 5 and 6. Every seven-cell group contains 3, 4, 5, 6 and 7. Every eight-cell group contains every digit except possibly one end β it holds 2 through 8 always, and then either 1 or 9. A setter who keys a six-cell group without a 4 in it has not made a hard puzzle. They have made an impossible one, and the interval arithmetic proves it without solving anything.
The no-repeat rule inside the group is the part setters most often understate, because classic Sudoku hides it. Within any one house β a row, a column, a box β a digit cannot repeat anyway, so a short group lying inside one box gets its distinctness for free. But a Renban group is under no obligation to stay inside a house. A line can wander across three boxes and two bands, and two of its cells may share no row, column or box at all. In those cells, classic Sudoku permits a repeat and Renban forbids it: the set is a set, each digit at most once, wherever the cells happen to sit. Print the rule in those words β digits do not repeat within a group β because a buyer who is never told it will accept keys a careful buyer would reject, and a setter who forgets it will generate them.
The any-order rule is the genre's most misunderstood sentence, and it must be printed as a freedom, not left as a silence. Consecutive describes the set, not the path. Along a line, the digits may rise, fall and jump: an interval of 3, 4, 5, 6 may sit on its four cells as 5, 3, 6, 4 and be entirely correct. Neighbouring cells on a group carry no relation at all β they may differ by one or by five β and the absence of a pairwise promise is itself information the buyer is owed, because buyers arriving from consecutive-dot and whisper variants will invent pairwise rules you never printed and then blame the puzzle when their invention fails. Say it plainly: order along the group means nothing; only the completed set is judged.
Given integrity and uniqueness close the contract as they close every honest logic genre. The printed givens are fixed digits the solver never moves, and the printed groups are fixed markings: a line's cells, a shaded patch's cells, exactly as printed, are the group β never one cell more or fewer in the key's private copy. Every given must also be legal where it stands: a given sitting inside a group is a member of that group's set, so two identical givens inside one group, or two givens in one group that cannot belong to any single interval of the group's length β a 2 and an 8 inside a four-cell group, which no run of four consecutive digits contains both of β end the puzzle before it starts. Exactly one completed grid then satisfies the classic rule, every group's interval, every no-repeat, and every given at once, and it must be reachable by deduction alone β from the forced middle digits of long groups first, then from intervals narrowed by givens at their edges, then from the steady refusal of any placement that would tear a gap in a set or double a digit inside one. A second solution is not added value and not a harder puzzle; it is a broken promise, because the buyer who finds the other grid has followed every printed rule and arrived somewhere your key calls wrong. The setter's obligation is therefore not to have a key β every grid a generator finishes has a key, the arrangement it happened to print β but to have proven the key is the only grid the groups allow.
The verification audit: eight checks before any file ships
One, board and group integrity. Re-read the printed puzzle against the source: the grid is the stated size, every given digit sits in the cell the source says it sits in, and every group β every line's exact run of cells, every shaded patch's exact set of cells β matches the source cell for cell, with none dropped, none added and none shifted by one cell in any direction. Then re-read the same puzzle on the exported file, not the working file, because a line that loses its last cell in export is exactly the kind of error a working-file proof will never see.
Two, given legality inside groups. Walk every given that sits inside a group and confirm two things: no group carries the same given digit twice, and the givens inside each group can all belong to at least one interval of that group's length β a four-cell group holding givens 2 and 5 can still be 2, 3, 4, 5, while a four-cell group holding 2 and 8 can be nothing at all. Record any failure as a puzzle defect, never as a difficulty feature: a puzzle whose givens already tear its own sets cannot be solved, only abandoned.
Three, the classic recount in the key. Walk the finished key row by row, column by column and box by box, and confirm each house holds 1 through 9 exactly once. Do not proof from memory of the generator's output, and do not let the group checks substitute for this one: a key can satisfy every interval and still repeat a digit down a column, and that key is false.
Four, the set recount as its own pass. For every group in every puzzle, collect the key's digits in that group, sort them, and check the sorted list three ways: it has exactly as many entries as the group has cells, no digit appears twice, and the largest minus the smallest equals the cell count minus one β the arithmetic fingerprint of an unbroken interval. A group of six cells whose sorted digits run 3, 4, 5, 7, 8, 9 fails on the fingerprint alone, 9 minus 3 equalling 6 where 5 was required, before anyone argues about which digit is wrong. Record each group by its length and its interval, puzzle by puzzle.
Five, the forced-digit sweep. For every group of five cells or more, check the digits the interval arithmetic forces: a five-cell group contains a 5; a six-cell group contains 4, 5 and 6; a seven-cell group contains 3, 4, 5, 6 and 7; an eight-cell group contains 2 through 8 and exactly one of 1 or 9. This pass is separate from the set recount on purpose β it is the check a setter can run on a puzzle before it has a key at all, and it catches the generator that quietly builds groups around intervals that never existed.
Six, decision completeness. Confirm that the printed rules decide every question a buyer will ask of the file: that order along a group is free, that repeats inside a group are banned even across houses, that groups are judged as sets and never by a sum, and that nothing on the page β no colour, no line style, no worked example β silently teaches a pairwise or in-order reading the rules page disowns. Walk the rules page, the worked example and the listing copy as one document, because a contradiction between them is a defect in the book, not a nuance.
Seven, the independent solve at the advertised level. Solve each puzzle fresh, from the printed file, using only the printed rules: classic logic, the interval a group's length and givens allow, the forced middle digits, and the exclusions a settled set casts into its houses. Note where the solve stalls. A stall that breaks only when you peek at the key is a uniqueness or difficulty defect β the buyer has no key to peek at mid-puzzle, and a puzzle that needs one is not finished.
Eight, the uniqueness probe. Take the finished key and try, honestly, to build a second grid: swap digits inside a group while keeping its set whole, exchange two digits that share every house the swap touches, and re-check every group as a set afterwards. If any second grid survives the classic recount and the set recount, the puzzle is not unique and must not ship as if it were β tighten it with a given or a group change and re-run the whole audit, because a change that rescues uniqueness can tear a set two groups away.
Difficulty is the width of the window, not the size of the grid
New setters reach for the obvious dials first β a bigger grid, longer groups, fewer givens β and all three are weaker than the dial that actually governs a Renban solve: how wide the window of possible intervals stays open, and for how long. A six-cell group whose interval could still be 1β6, 2β7, 3β8 or 4β9 is a wide window: the buyer knows the middle digits 4, 5 and 6 are in there somewhere and almost nothing else. The same group with one given 8 at any of its cells collapses the window to 3β8 or 4β9 at a stroke. Difficulty lives in that collapse β how many placements, in the group and in the houses crossing it, must be earned before the window narrows to one interval β far more than in the raw count of cells.
Placement is the second dial. A group lying mostly inside one box hands the buyer the box's own exclusions for free, so its interval settles early; the same length of group strung across three boxes and two bands keeps its repeats ban doing real work for much longer, because no single house sees enough of the set to corner it. Given placement matters the same way: a given at the edge of an interval's possible range cuts the window from one side only, while two givens near opposite ends of the range can fix the interval outright β which is why counting givens is a poor difficulty label and describing a typical puzzle's longest stretch of honest work is a better one.
Decision distance is the third dial, and the one honest grading is built on. An easy Renban puzzle pays out early and often: short groups resolve, their sets cast exclusions into their houses, and those exclusions narrow the next window in a steady chain. A hard one asks the buyer to hold two or three wide windows open at once, tracking which intervals remain possible for each, until a placement far from any group forces one window shut and the rest cascade. Grade your book by that β how many windows a puzzle holds open at its hardest point, and how far a buyer must travel from the forcing fact to the group it settles β and state the grade in those terms on the listing. A buyer who was told a book's hard section holds three windows open will trust the stall when it comes. A buyer who was told only that the grids are the same size as the easy book's will call the same stall a misprint.
One honest limit belongs here too: length is not virtue. A nine-cell group is just a row's worth of digits restated β it adds the no-repeat rule across houses and nothing else β and a book padded with very long groups is often easier, not harder, than a book of well-placed fives and sixes whose windows stay wide. Build difficulty from window width, placement and decision distance, and let group length serve those, never the other way round.
Print honesty: a purple line that photocopies to nothing is a broken clue
The modern form of this genre is a colour, and colour is the first thing a home printer takes away. The standard Renban line is purple. Printed in greyscale on a tired cartridge, a thin purple line can fade to a whisper or vanish entirely β and a vanished line is not a cosmetic flaw, it is a missing rule: the buyer is now solving a classic Sudoku that the key insists is something else, and every set check they run against cells they cannot see marked will fail. Print your lines thick enough and dark enough to survive a greyscale photocopy as a visible grey track, and consider a line style β a band, a double line, a shaded corridor β that carries the grouping in shape as well as hue, so the marking survives in black and white on its own terms.
The older group form has the mirror-image problem. Its shaded patches must print as a gentle, even grey that a pencilled digit stays legible on top of β too dark and the buyer's own writing drowns; too light and the patch boundary blurs, and a boundary misread by one cell is a group misread by one cell, with the whole set arithmetic shifted. In both forms, test on the exported file at final size: every group on the page accounted for, every cell of every group visibly inside its marking and no cell of any other group, every given digit crisp, and the worked example's markings identical in style to the puzzles', so the buyer learns one visual language and is never asked to relearn it mid-book.
State the print facts on the listing as plainly as the rules: the grid size as printed, whether the lines are designed to survive greyscale printing, and the paper size the file is laid out for. A buyer printing on a smaller sheet deserves to know the lines were proofed at that size too β or that they were not. Print honesty is not a disclaimer. It is the difference between a book whose markings a buyer can trust at arm's length, in pencil, on a Sunday evening, and one whose groups dissolve exactly when the solving gets serious.
Licensing, provenance, and whose variant this is
Renban belongs to no single seller, and your book should say what it honestly is. The word means a consecutive run of numbers, and the variant grew up in the shared Sudoku culture of contest setters and variant collectors β grey Renban groups in contest booklets, purple Renban lines in the modern solving apps β refined in public by many hands rather than owned by one. That history is a selling point when you state it straight: buyers are purchasing a recognised variant with a real solving culture behind it, not a private invention whose rules exist only in your listing.
What you may not do is take the puzzles themselves from that culture. A puzzle set by a named setter β in a contest booklet, on a solving portal, in a video solve β is that setter's work, and reprinting a book of harvested puzzles is theft however public the source was. Generate your own grids, or license them from a setter who can grant the right, and keep your generation records: the source grid, the groups as generated, and the audit trail that shows each set re-checked. Own generation only is not a slogan; it is the difference between selling a book in this genre and selling someone else's book with your cover on it.
Delivery is the last promise, and it is a file promise. When a buyer writes to say a group in puzzle forty-one cannot be satisfied β and one day a buyer will be right, because no audit is perfect β the fix is a corrected file sent to every buyer of that edition, not a quiet patch for the one who complained. That is the whole job of a delivery tool: keep the list of who bought which edition, and push the corrected file to all of them. It is the job FileFreight was built for β flat-rate delivery for digital sellers, with corrected files reaching past buyers instead of stranding them β and any tool that cannot re-deliver a fixed file to every past buyer turns a one-cell erratum into a permanent defect in every copy already sold. Publish an errata note, date the corrected edition, and keep the old edition's buyers whole. A seller who corrects loudly keeps the trust the error spent.
The launch checklist
One, every puzzle in the file has passed all eight audit checks on the exported file β board and group integrity, given legality, the classic recount, the set recount with its interval fingerprint, the forced-digit sweep, decision completeness, the independent solve, and the uniqueness probe β and the results are recorded puzzle by puzzle, group by group.
Two, the rules page states the standard form in plain words: classic Sudoku in every row, column and box; every marked group holding distinct consecutive digits; no repeats inside a group, even across houses; order along a group free and meaningless; no sums, no pairwise promises; givens and groups fixed; exactly one solution, reachable without guessing.
Three, the worked example is solved aloud set by set β each group's digits collected, sorted and shown to form an unbroken run with nothing missing and nothing doubled β so a first-time buyer can re-walk the checking method before puzzle one.
Four, every confusion a buyer brings from a neighbouring variant is disowned on the rules page by name: not a thermometer (no order, no bulb, no direction), not a killer cage (no sum is printed or implied), not a consecutive-dot or whisper rule (no pairwise promise between neighbouring cells, in either direction).
Five, difficulty is graded and described by window width, group placement and decision distance β how many intervals stay open at the hardest point and how far the forcing facts sit from the groups they settle β and the listing says so in those terms, not in grid sizes or given counts alone.
Six, the print proof was done on the exported file at final size, in colour and in greyscale: every line visible as a track without its colour, every shaded patch's boundary exact, every given legible, and no group a cell larger or smaller than its key.
Seven, provenance is clean and the delivery path is tested: every puzzle your own generation or properly licensed, the generation and audit records kept, the listing honest about the variant's shared history, and a test purchase has delivered the file β and re-delivered a corrected copy β exactly as a buyer's will.
