Two symbols are enough to create a lot of logic
Equinox does not need large numbers, arithmetic, or a paragraph of clues sitting beside the grid. You see suns, moons, a handful of fixed cells, and small signs between neighbouring spaces. That economy makes the puzzle look unusually calm: a clean grid, two possible values, and rules that fit into a short introduction. The depth comes from scale. Balance is a whole-row or whole-column constraint, the no-three rule inspects windows of three cells, and = or × describes an exact relationship between a specific pair of neighbours.
Blupoli's version starts from the public mechanics described on Puzzleship's Equinox page, while the browser engine, generator, boards, copy, tests and visuals are our own implementation. The reference identifies the three constraints that matter here: an even split of the two symbols in every line, no run of three matching symbols, and local markers requiring two neighbours either to match or differ. That places Equinox near the broader binary-puzzle family, but the relationship clues change how information travels through a solve.
Rule one: every line ends perfectly balanced
On a 6×6 grid every row and column finishes with three suns and three moons. An 8×8 uses four of each, a 10×10 uses five, and a 12×12 uses six. Order is irrelevant. The requirement is purely about quantity: once a line is complete, half its cells hold one symbol and half hold the other. It sounds almost too simple, yet its power grows as a line fills. The moment one symbol reaches its quota, every remaining blank in that line becomes the other symbol without any further assumptions.
Counting is therefore worth doing before chasing a clever pattern. If an 8×8 row already contains four moons and still has three blanks, all three blanks are suns even if they are scattered across the row. If a 6×6 column has three suns, one moon and two empty cells, those empties must both be moons. A line-wide deduction can immediately trigger something local: a newly placed moon may satisfy a relationship, create a matching pair, or bring the crossing row to its own quota.
Rule two: never allow three suns or three moons in a row
The no-three rule is the pattern that turns into instinct fastest. Two adjacent suns mean an empty cell directly before or after that pair cannot be another sun; it must be a moon. The same logic works for moons. The separated form is just as useful: sun, blank, sun forces a moon into the gap, while moon, blank, moon forces a sun. These are compact, visible and certain deductions, which makes them ideal opening moves when you want progress without guessing.
There is a practical reason to train your eye on groups of three. A 10×10 or 12×12 board may feel visually busy, but any potential triple depends on only three consecutive positions. Scan a row for XX_, _XX and X_X, then repeat vertically. You never have to hold the whole board in working memory at once. Equinox rewards moving between two scales: look very close when checking triples, then zoom mentally back out when counting the total balance of a row or column.
Rule three: = keeps two cells the same, × keeps them different
The small markers between cells are the feature that most clearly separates this experience from ordinary Takuzu. An = says the two neighbouring cells must contain the same symbol. If one is a sun, the other is a sun; if one is a moon, the other is a moon. A × says the opposite: the neighbours must differ, so knowing either side fixes the other. These are not hints with varying strength. They are exact constraints that every valid intermediate and final state must respect once both cells are known.
Relationships become especially powerful in chains. Imagine cells A, B and C with A=B and B×C. Discover that A is a moon and B immediately becomes a moon while C becomes a sun. One fact has settled three cells. If C then sits beside another sun, the no-three rule may force a fourth position. That is why = and × should not be treated as decoration squeezed into the gaps. They behave like tiny equations, carrying information from one part of the grid into another and often turning a local fact into a multi-step cascade.
Strong moves usually wake up a different rule
Solving becomes smoother when you stop asking “which rule should I use?” and start asking “what changed because of that placement?” Completing the sun quota in a row may turn several blanks into moons. Those moons can create a pair. The pair can force a sun. That sun can cross an = and create another sun elsewhere. The new cell may complete a column. A useful chain rarely belongs to one named technique; it moves from balance to triples to relationships and back again until the board reaches a temporarily stable state.
A simple routine helps: after entering a symbol, check the immediate neighbourhood for pairs or sandwich patterns, follow any = or × touching the cell, and then recount its row and column. The order is not mandatory, but it limits unnecessary eye movement and gives each move a short follow-up checklist. If all three checks produce nothing, move to another region instead of staring at the same blanks. Another line may create the information that eventually returns to the area that currently looks stuck.
Equinox is not merely Takuzu wearing prettier icons
Blupoli already includes Takuzu, so the first implementation question was whether Equinox could simply replace zeroes and ones with suns and moons. It could not. Our Takuzu also uses the familiar rule that two completed rows may not be identical and two completed columns may not be identical. That rule creates valuable deductions in Takuzu, but it is not one of the three constraints described for this Equinox variant. Quietly carrying it over would change the puzzle even if the screen looked convincing.
Equinox therefore has its own solver and deliberately does not reject duplicate lines. Two completed rows may match as long as both are balanced, contain no triple, and satisfy every relationship clue that touches them. The distinction changes player reasoning. In Takuzu, a nearly complete row can be compared with a finished row to prevent duplication. Here that argument is invalid; the missing symbol must be derived from balance, a triple pattern, or =/×. Keeping that boundary explicit prevents the engine from enforcing a hidden rule the interface never taught.
How to begin without trying to read everything at once
When a new grid opens, do not attempt to follow every relationship simultaneously. Start from the fixed symbols and collect the zero-cost deductions: adjacent matching symbols, matching symbols with one blank between them, or lines that have already reached the quota of suns or moons. Each move increases the information available without requiring a long chain in your head. Then inspect relationships touching the cells you just resolved and recount the affected row and column before moving on.
Easy boards often turn that first sweep into an immediate cascade. Hard and Expert may yield only two or three cells before the obvious work runs out. That is the moment to look for relationship chains. A pattern such as =, ×, = can encode several dependencies even when none of its cells is close to completing a line. You do not need to know whether the first position is a sun or moon yet. Knowing which cells will match and which will oppose one another is already useful structure waiting for one anchor value.
Count possibilities, not just the symbols you can already see
Balance becomes much stronger when you stop using it only for “I already have four moons.” Consider an 8×8 row with three suns, two moons and three blanks. No quota has been reached, but you know the final three cells must contain exactly one sun and two moons. If two of those blanks are joined by =, they cannot both be suns because the line has room for only one additional sun. Therefore that equal pair must be moon-moon and the remaining blank is a sun. You have combined counting and a relationship without either blank being directly clued.
A × pair provides a complementary kind of information. Two blank cells linked by × always contribute one sun and one moon in some order. Inside a row or column, you can mentally reserve that one-of-each contribution even before deciding which side gets which value. Sometimes that reservation settles all other blanks in the line. This is an important step toward harder boards: partial information has value. You do not need to know every cell individually if a pair or group already tells you enough about the total balance.
Read long = / × chains as parity
As a relationship chain grows, it helps to imagine two groups: cells that will match the chain's starting cell and cells that will contain its opposite. Crossing = keeps you in the same group; crossing × flips you to the other group. For A=B, B×C, C=D, A and B belong together, C and D belong together, and the two groups are opposite. This description remains true even if all four cells are still blank, so the chain already contains logic before any actual sun or moon is known.
You do not need formal algebra or annotations to benefit from this. Think “same” and “opposite.” If one end eventually meets a known clue, the whole connected chain may resolve. If two members of the same group sit in a nearly balanced row, the row count can decide which group must be suns and which moons. Harder difficulties lean more heavily on this kind of reasoning because they provide fewer fixed symbols. The relationships are not extra decoration compensating for missing clues; they are a major source of structure in their own right.
Board size and logical difficulty are separate choices
Blupoli offers 6×6, 8×8, 10×10 and 12×12. Size changes how much information is visible, how many rows and columns interact, and how long a session may last. Difficulty instead changes how many fixed symbols you receive and how much deduction is left for you. We do not want “large” to be a synonym for “hard.” A 12×12 Easy grid can provide generous footholds, while a compact 6×6 Expert board can hide a tighter, more demanding dependency among just a few cells.
Keeping those dimensions independent is useful for learning. You can stay on 6×6 while moving from Easy to Expert and study how the clue density changes without adding visual load. Or keep Normal difficulty while moving up through the four sizes to practise attention management with roughly similar logical expectations. Results record the actual size-and-difficulty combination, so a strong 6×6 Expert performance does not get treated as the same event as a much longer 12×12 Normal solve.
What actually changes from Easy to Expert
The four difficulty levels are derived from the same solution family for a given seed. The generator first builds a complete valid solution and a compatible set of relationship clues. It then searches for a sufficiently small set of fixed cells that still leaves exactly one solution according to the solver. That sparse core becomes Expert. Hard, Normal and Easy progressively add more fixed symbols from the same solved board. Difficulty therefore does not secretly switch rule sets or simply enlarge the grid.
This construction gives us something measurable to test. The number of fixed cells decreases as difficulty rises, while uniqueness must survive at every level. Easy shortens the uncertain opening and lets you practise the basic patterns. Normal makes relationships more important. Hard asks you to connect rules more often. Expert gives the relationship graph and partial counts much more responsibility. Different players will still perceive difficulty differently, but the generator changes the information available in a deliberate, reproducible direction rather than relying on labels alone.
A generated puzzle is not accepted just because it looks plausible
Random generation can easily produce an attractive board with two valid endings, or a position where the visible clues cannot distinguish between symmetrical alternatives. In a logic puzzle that is a defect, not charming ambiguity. Equinox includes a solver that applies the same three rules the player sees. For every generated grid it searches for solutions up to a limit of two. If a second solution exists, uniqueness has not been established and the generator must add more information before the puzzle can be considered publishable.
The solver is also a guardrail for implementation mistakes. It verifies that every fixed clue agrees with the intended solution, every relationship joins compatible neighbours, and the finished board is balanced and contains no triples. Automated tests exercise all four sizes and all four difficulty levels with deterministic seeds. That cannot replace human judgement about whether a board feels good to play, but it removes a more basic failure mode: depending on a handful of manual games and hoping the unseen generator output behaves the same way.
Seeds make a puzzle family reproducible
Given the same seed, size and difficulty, the generator must recreate the same puzzle. Reproducibility matters for debugging, tests and any future feature where people might share the same challenge. If a bug occurs on a particular board, we can reconstruct it instead of trying to catch a case that changes on every reload. Browser tests can solve a known generated puzzle without storing copied puzzle data, because the expected solution comes from the exact generator used by the product.
For normal play, New game simply chooses a fresh seed and leaves that machinery invisible. Saved state keeps the seed, size, difficulty, current cells and history together, so leaving the page does not replace a half-finished puzzle with another board. That continuity becomes increasingly valuable on 10×10 and 12×12. Restart intentionally keeps the seed so you can replay the same challenge from its initial clues; New game intentionally changes it. Small distinctions like that make the controls predictable rather than surprising.
The interface makes the rules visible without solving for you
Suns use Blupoli Coral and moons use Violet, while = and × sit in compact markers between cells. Fixed clues receive a stronger border so you can distinguish what belonged to the original puzzle from what you entered. When a rule is already broken, the affected cells are highlighted as conflicts. The aim is not to turn the board into a traffic-light system that evaluates every possible move in advance. It is to tell you when your current state contradicts a condition that is objectively checkable from the published rules.
The board keeps a square geometry across phones, tablets and desktops. On a 390-pixel screen a 12×12 grid leaves little room, so relationship markers have to remain legible without becoming extra grid columns or rows. They sit across the gap between their two cells instead. That preserves equal cell sizing and keeps the touch target where it belongs: the whole cell. The small marker communicates a relationship, but the player never has to hit that tiny sign with a finger to interact with the puzzle.
Touch, mouse and keyboard all use the same board
On a phone, tapping an editable cell cycles empty → sun → moon. A desktop click does the same, while the secondary mouse button cycles in reverse. Keyboard players can move with the arrow keys, enter S or 0 for a sun, M or 1 for a moon, and clear with Delete. Space and Enter cycle the current cell as well. Fixed clues stay focusable so their location and relationships remain available to assistive technology even though the cells cannot be edited.
Each cell exposes its row, column, current state, whether it is fixed, and the relationships connected to it. That matters because a visual = hovering between two squares is not sufficient information for someone who cannot see it. Accessibility has to describe the logical structure, not merely give the grid a name. Visible focus, contrast, and reduced-motion behaviour complete the same interaction model. The goal is not a separate “accessible version” with fewer features, but one board whose essential reasoning survives different input and perception methods.
Undo, Redo and Hint are analysis tools
A reversible history lets you test a deduction you believed was safe, inspect what follows, and retreat if a contradiction proves the assumption wrong. That does not turn Equinox into a guessing game; it preserves context and prevents one accidental tap from destroying a long chain of work. Redo restores the branch you just backed out of, Clear removes your editable entries while preserving fixed clues, Restart returns to the beginning of the same seed, and New game generates another seed. Each action has a distinct purpose.
Hint takes a deliberately small role. It chooses an editable cell whose current value does not match the verified solution and places the correct symbol there. It does not reveal a sequence of deductions or fill an entire line. The best way to use it is to ask why that one cell can now be known: perhaps a relationship, a completed quota or a nearby pair explains the value. On Expert, a single revealed cell can act as an entry point to a chain you had failed to notice without turning help into an automatic solver.
Conflicts check rules rather than spying on the hidden answer
Equinox can identify three kinds of contradiction: a line already containing too many suns or moons to reach balance, a completed run of three matching symbols, or a =/× relationship violated by two known cells. The affected cells are highlighted. A 6×6 row with four suns is already impossible even if blanks remain, because no later move can reduce four to the required three. Likewise sun-sun-sun is invalid before you have filled the rest of that row or column.
What the game does not do is mark a cell wrong merely because it differs from the stored final solution. If a partial state still satisfies every visible rule, it remains a logically possible hypothesis from the player's perspective. That boundary is important. Constraint checking should help you notice demonstrable contradictions, not compare every move with an answer key behind your back. If the interface reveals the secret solution too eagerly, the puzzle stops teaching you to read consequences and becomes an exercise in following corrective colours.
An example chain: from one relationship to a completed column
Imagine positions four and five of an 8×8 row are connected by =. Position four is a sun, so position five must also be a sun. If position six is blank, the matching pair at four and five forces position six to be a moon; otherwise three suns would appear consecutively. That new moon can alter the count of its column even though the original relationship ran horizontally. A local equality has produced a triple-pattern deduction and then changed a line in the perpendicular direction.
Suppose the column now reaches its quota of four moons. Every remaining blank in that column becomes a sun. One of those suns may be linked by × to another cell, turning that neighbour into a moon. If the moon forms moon-blank-moon in its row, the middle blank becomes a sun. In only a few placements you have moved through relationship, no-three, balance, another relationship and another no-three deduction. That cross-rule propagation is the characteristic rhythm of Equinox and explains why a stalled grid can suddenly unravel in bursts.
Easy strategy: turn the rules into reflexes
Easy contains enough fixed symbols that an obvious starting point is usually available. Use short scans: matching pairs, matching symbols around a blank, lines that have filled one quota, and relationships touching known clues. Do not worry about minimising moves at first. The goal is to make the three rules automatic and build the habit of rechecking consequences after every symbol. If one part of the grid stops producing information, switch to another row or column. A deduction elsewhere will often send information back into the quiet region later.
Easy is also the best place to practise not importing rules from another puzzle. If you arrive from Takuzu, two rows that appear capable of ending identically may feel suspicious. Leave them alone: Equinox does not ask you to make lines unique. If you arrive from Sudoku, there are no regions that require unique values. The puzzle speaks only through the rules it publishes. Better solving begins with using those rules fully and resisting the temptation to add a familiar fourth condition that is not actually present.
Expert strategy: work with structure before values
Expert supplies fewer fixed symbols, so scanning only for visible pairs can run dry quickly. Start reading relationship components. Work out which cells are forced to match and which must be opposite even if none has a sun or moon yet. Then intersect those groups with row and column counts. Two =-linked blanks in a line that has room for only one additional sun cannot both be suns. A × pair contributes exactly one sun and one moon, which can be enough to settle the rest of the line even before you know which endpoint is which.
When several chains intersect, avoid holding the entire graph in memory. Work in small components. Choose a nearly constrained row, resolve what you can, propagate only the relationships that leave it, and then recount. A large board becomes a succession of connected local problems. Difficulty should hide the useful starting thread, not force a coin flip. Once you find the right dependency, the rest of the board should become increasingly concrete as the abstract same/opposite structure acquires actual sun and moon values.
Four sizes change the rhythm more than the rules
6×6 is compact and excellent for short sessions. Quotas of three are reached quickly and every deduction affects a large share of the board. 8×8 gives relationship chains more space while remaining easy to scan on a phone. 10×10 begins to demand deliberate attention management: you can resolve one corner and forget to revisit a column changed on the other side. 12×12 amplifies that effect, making the tracking of consequences a skill in its own right alongside the local logic.
No size introduces a new rule, which is essential for transferable learning. Sun-sun-blank means the same thing on 6×6 and 12×12; only the surrounding context grows. A × pair still contributes one of each symbol. = still copies a value. As the board expands, perceptual difficulty increases through density and distance rather than hidden exceptions. You can return to a smaller board to practise one technique, then carry the exact same reasoning into a larger grid without translating it into a new ruleset.
Where Equinox sits in the Blupoli catalogue
If you enjoy pure binary reasoning, Takuzu is the closest relative and makes the extra unique-line rule easy to compare. If you like local clues relating neighbouring cells, Kropki uses dots to impose numerical relationships inside Sudoku. If you prefer a network of constraints that eventually forms a continuous structure, Balance Loop replaces symbols with connections and produces a very different style of propagation.
Comparing these games is a reminder that “logic” is not one skill. Equinox mixes simple counting, pattern recognition and relationship propagation. Takuzu adds global line comparison. Kropki combines local arithmetic with Sudoku uniqueness. Balance Loop reasons about continuity. Moving between them trains the ability to identify which information is legal in a particular system, which matters as much as spotting an individual deduction. A strong solver always asks what the current game actually permits before applying a technique learned somewhere else.
Why Blupoli built its own engine instead of copying the reference
A reference URL is useful for understanding a mechanic; it is not permission to transplant somebody else's product. Blupoli does not import Puzzleship's daily boards, copy, source code or graphical assets. This implementation generates its own solutions and relationships from seeds, draws its own presentation with Blupoli's current visual system, and uses the platform's shared state, results, achievements and accessibility layers. That keeps the game maintainable and lets us verify exactly what the rules mean instead of depending on an external representation that could change.
The distinction also forced us to inspect what we already had. Reusing the Takuzu solver would have been shorter, but wrong because it would silently introduce row and column uniqueness. We reused the architecture that truly is shared — controls, persistence, progress and the native-engine boundary — and rewrote the puzzle domain that needed different semantics. That is the kind of reuse we want: common infrastructure stays common, while two superficially similar games are allowed to remain genuinely different when their rules demand it.
An Equinox solve is still a Blupoli session, not a separate island
When the board is complete, Equinox reports moves, size, difficulty and duration through the same result system used by other Blupoli Puzzles games. Local progress can compare variants, preserve best results and feed achievements without the engine knowing storage details. During play, state is saved through a capability supplied by the host, so the puzzle domain never reaches directly into localStorage or creates a private persistence layer. That boundary is invisible when it works, but it keeps the engine ready for future platform-level synchronisation.
There are also three optional game-specific challenges: finish without ever violating a = or × relationship, without temporarily creating a triple, and without exceeding the allowed symbol quota in any line. They do not change the puzzle rules; they observe how cleanly you travelled through valid states. You can still finish after correcting contradictions, or you can aim for a solve in which every intermediate position remains rule-consistent. Achievements become a way to examine your process rather than a reason to hide Hint, Undo or other learning tools.
What to remember when you open the next grid
Begin with what is certain and let information travel. Count each line. Look for matching neighbours and matching symbols with one gap between them. Read = as “same” and × as “opposite.” After every placement, recheck the row, the column and nearby relationships. When you get stuck, reason about groups that must match or differ even before you know their actual symbol. Most importantly, do not add rules Equinox does not have. The puzzle remains fair only when both player and engine obey the same published contract.
After a few games, the board stops feeling like 36, 64, 100 or 144 unrelated decisions. You begin to see pairs, quotas and chains. A sun is no longer merely a sun: it may complete a row count, force a moon beside it and travel through = into another column. That conversion of simple symbols into a network of consequences is what makes Equinox such a natural fit for Blupoli Puzzles. Play Equinox now, then compare it with our Takuzu guide or our Kropki guide to see how a small change in constraints creates a different style of reasoning.