Board Generation

Board generation is the process of filling a board with objectives from your goal set. It balances randomness with constraints to produce fair, varied, and interesting boards every time.

Overview

At a high level, generation works like this:

1. Prepare the pool: all enabled objectives are gathered and filtered by weighting

2. Place forced objectives: any objectives with designated positions are placed first

3. Fill remaining spots: open positions are filled one at a time, respecting all active constraints

4. Apply modifiers: post-generation effects like shinies and blanks are applied

If placement gets stuck, the system has multiple recovery strategies (from backtracking to constraint relaxation) so a board is almost always produced.

---

Seeds & Determinism

Every board is generated from a seed. The same seed with the same goal set and settings will always produce the exact same board. Generation is fully deterministic, so boards are reproducible and shareable. Two players with the same seed and goal set will always get identical boards.

---

The Objective Pool

Building the Pool

When generation starts, all non-disabled objectives from the goal set are loaded into a single pool. Each objective appears in the pool exactly once, regardless of its individual limit.

Weighting

Each objective can have a weighting value between 1 and 100 (defaulting to 100). This controls how likely the objective is to make it into the generation pool for any given board.

• A weighting of 100 means the objective is always included
• A weighting of 50 means roughly a 50% chance of being included
• A weighting of 1 means roughly a 1% chance, the lowest possible without disabling the objective entirely

Weighting is rolled once at the start of generation. Either an objective makes it into the pool or it doesn’t. Once in the pool, all objectives have an equal chance of being selected.

Note

Objectives with forced positions always bypass the weighting roll and are guaranteed to be in the pool.

If weighting excludes too many objectives and the pool becomes too small for the board size, excluded objectives are pulled back in to fill the gap. A warning is issued when this happens.

---

Objective Properties

Individual Limits

Individual limits control how many times a specific objective can appear on a single board.

Most objectives have a limit of 1, meaning they can only appear once. Some objectives have higher limits (2, 3, etc.), allowing them to appear multiple times. After each placement, the limit decreases by 1. When it hits 0, the objective is removed from the pool entirely.

Caution

An objective’s individual limit cannot exceed the number of range values it has. This is enforced by validation to ensure every copy on the board gets a distinct range value, preventing duplicate identical goals.

Ranges

Ranges represent target amounts within goal text (like “Collect {{X}} gold” where X could be 100, 500, or 1000). They’re always positive quantities that the player is working toward. See Using Ranges for guidance on what makes a good range.

The objective is chosen first, and then a value is picked from its available ranges. If the same objective appears multiple times (via a higher individual limit), a different range value is used each time.

Tags

Tags are unique identifiers that enforce mutual exclusivity. Only one objective with any given tag can appear on a board. This is useful for preventing objectives that would conflict or overlap with each other.

Danger

Tag violations are the one constraint that can never be relaxed. If a tag conflict can’t be resolved, generation will fail rather than place conflicting objectives.

Forced Positions

Objectives can specify a list of board positions where they must be placed. These are 1-indexed positions on the board (position 1 is the first cell).

An objective with forced positions will only ever appear in one of its designated spots. If none of its positions are available or constraints prevent placement, the objective is skipped entirely and won’t appear in a non-designated spot as a fallback.

Forced objectives are always placed before regular objectives in the generation process.

---

Categories

Categories are the primary tool for controlling board composition and balance. There are two types, and they solve different problems.

Board Categories

Board Categories

Scope: The entire board · Purpose: Control overall composition

Board categories set limits on how much of the total board can be occupied by objectives of a given type. They’re defined as percentages and converted to absolute counts based on board size.

For example, a board category limit of 25% on a 5×5 board (25 cells) means a maximum of 7 objectives of that type (rounded up from 6.25).

Tip

A category limit of 0% doesn’t mean zero allowed. It means “at most 1”. This lets you heavily restrict a category without completely blocking it.

This can also be achieved through using the Tag system.

Line Categories

Line Categories

Scope: Individual lines (rows, columns, diagonals) · Purpose: Prevent clustering

Line categories set limits on how many objectives of a given type can appear in any single line. They work the same way as board categories but are checked independently for every row and column (plus diagonals on square grids).

Note

Line categories are only supported on grid boards. Pyramid and triple layouts don’t have traditional lines.

Additionally, the same objective can never appear twice in the same line, even if its individual limit allows multiple copies on the board. This prevents a single row or column from containing the same goal twice with different range values.

Why Both?

Board and line categories solve fundamentally different problems, even when they use the same category names.

Only Board Categories

You could limit the board to 25% combat goals, but without line limits you could end up with all of them clustered in the same row, making one line trivially easy and others devoid of combat.

Only Line Categories

You could limit combat to 2 per line, but without board limits you could end up with 40% of the entire board being combat-related. The overall composition would be unbalanced.

Both Together

Board categories keep the macro balance in check while line categories prevent micro-level clustering. A 25% board limit with a 40% line limit ensures combat goals are both proportional overall and evenly distributed across lines.

---

Progression

Progression controls the difficulty distribution across the board, ensuring that objectives are placed in appropriate positions relative to where a player is likely to encounter them during gameplay.

Note

Progression is only active in modes that use it: Rush, Ascend, and Summit. Standard Bingo does not use progression.

Progression Zones

Every objective can be tagged with one or more progression zones:

Zone	Range	Typical Position
Early	0–40%	Bottom of the board / start of the list
Mid	20–60%	Lower-middle area
Late	40–80%	Upper-middle area
Endgame	60–100%	Top of the board / end of the list

The zones intentionally overlap. The Mid zone overlaps with both Early and Late, and Late overlaps with Endgame. This flexibility prevents zones from being too rigid while still maintaining a general flow from easier objectives at the start to harder ones at the end.

How Positions Map to Zones

Each position on the board gets a progress percentage based on its physical location:

• Grid boards (Ascend): Counted from the bottom row (0%) to the top row (100%)
• Pyramid boards (Summit): Bottom row is 0%, peak is 100%
• Triple/Rush: Sequential from start (0%) to end (100%)

An objective can only be placed at a position if at least one of its progression zones includes that position’s progress percentage.

Exclusive Zones

Some positions are exclusive and can only accept objectives from a single progression zone. These are typically the very bottom rows (early-only) and very top rows (endgame-only). The size of these exclusive zones scales with the board size so that smaller boards aren’t dominated by locked-off areas.

Tip

Objectives tagged with more progression zones are more flexible about where they can be placed. An objective tagged with both Early and Mid can appear anywhere in the bottom 60% of the board.

---

The Placement Process

Phase 1: Forced Placement

Objectives with forced positions are placed first. The system identifies all positions that are targeted by at least one forced objective, shuffles them randomly, and then attempts to place a valid forced candidate at each one.

Even forced placements must respect all active constraints (categories, tags, progression, etc.). If no valid forced candidate can fill a position, it’s left open for the regular placement phase.

Any forced objectives that weren’t placed during this phase are excluded from the regular phase and won’t appear in non-designated spots.

Phase 2: Regular Placement

Remaining empty positions are filled one at a time. For each position, the pool is filtered down to only objectives that satisfy all active constraints:

1. Tag check: remove any objectives whose tag is already on the board

2. Progression check: remove objectives whose zones don’t match the position

3. Board category check: remove objectives whose board categories have no remaining quota

4. Individual limit check: remove objectives that have already hit their placement limit

5. Range check: remove objectives that have no unused range values left

6. Line category check: remove objectives that would exceed any line’s category limit

From the filtered pool, an objective is randomly selected and placed. Its constraints are updated (category counts decremented, individual limit reduced, etc.) and the process moves to the next position.

Position Ordering

Positions are never filled in a fixed order like left-to-right or top-to-bottom. Instead, each position is randomly picked from the remaining unfilled spots. This prevents goals from categories that exhaust quickly from always clustering in the same area of the board.

When progression is active, unfilled positions are split into two groups: exclusive (early-only or endgame-only) and overlap (everything else). Exclusive positions are filled first, with random selection within each group. This ensures the most constrained spots are filled while there are still plenty of suitable candidates in the pool.

---

Multi-Level Generation

Simple random placement works for most goal sets, but some have tighter constraints like strict category limits or lots of cap-1 line restrictions. The generator has several extra tools to handle these, and they kick in automatically when needed. Most boards never use them.

Tip

The easiest way to avoid generation issues is usually to add more objectives to your goal set. A larger pool gives the generator more room to work with, though this depends on the categories and tags those new objectives carry.

Greedy Heuristics

If the first attempt at greedy placement fails, retries use heuristics to pick smarter. The same candidates are available, but they’re scored and the pick is biased toward better-scoring ones:

• Board category pressure: favour goals whose board categories still have plenty of room
• Line category pressure: favour goals that won’t strain already-tight lines
• Progression scarcity: favour goals with fewer valid zones so they get placed before their options dry up

Randomness still comes first. Rather than always picking the single “best” option, it picks randomly from a top tier of well-scoring candidates. Heuristics nudge the generator away from bad picks, but don’t dictate the outcome.

Backtracking

If greedy can’t find a valid objective for a position, the generator switches to backtracking. This is a more thorough solver that tracks which objectives are valid at each position and updates that tracking incrementally as it places and undoes objectives.

The solver starts from the greedy frontier, keeping the objectives that greedy already placed successfully and only filling the remaining empty positions. This preserves the randomness from the greedy phase while giving the solver a smaller problem to work with.

The solver works by:

1. Using MRV (Minimum Remaining Values) to decide which position to fill next, always picking the one with the fewest valid candidates. This tackles the tightest spots first while there are still plenty of options left.
2. Auto-placing singletons: if any position has exactly one valid candidate, it’s placed immediately without spending search budget.
3. Checking feasibility: after each placement, the solver checks the remaining board is still solvable. If the supply of available objectives drops below the number of empty positions, or if any progression zone runs out of candidates for its exclusive positions, the branch is abandoned immediately.
4. Picking a random candidate at the selected position, then moving to the next tightest position.
5. If a pick leads to a dead end, undoing it and trying a different candidate.

If the frontier-seeded attempt fails (the greedy placements may have caused the dead end), the solver automatically retries from a clean slate before counting against restart phases.

Each phase has a budget that scales with the number of open positions. If one phase runs out, a fresh phase starts with a new seed to try a different direction. If a phase exhausts its budget but barely made any progress, remaining restart phases are skipped since a different search direction won’t help.

Constrained Mode

Before generation starts, the system checks your goal set’s constraints. If they’re restrictive enough that random candidate picks would keep running into dead ends, backtracking switches to constrained mode. This typically means lots of cap-1 line categories covering most of your objectives, or many tight board categories.

Constrained mode activates when either of these is true:

• Line-restrictive: most line categories cap at 1 per line, and most objectives carry line categories
• Board-restrictive: most board categories are tight (allowing very few objectives each), and most objectives carry board categories

Instead of picking candidates randomly, constrained mode scores them using heuristics that check how much category budget remains and whether placing a candidate would leave enough supply for other positions that still need it. It then picks randomly from a top tier of the best-scoring candidates, with the tier narrowing across restart phases.

When forced-position objectives exist, constrained mode also checks after each placement that every unplaced forced objective still has at least one viable spot. If a placement would strand a forced objective, it gets skipped immediately.

Normal goal sets are unaffected. Constrained mode only kicks in when the constraint profile calls for it.

Constraint Relaxation

If backtracking runs out of budget without finding a valid board, the generator falls back to constraint relaxation. Rather than giving up, it:

• Finds the candidate that breaks the fewest constraints
• Places it anyway, ignoring those constraints
• Finishes the board and flags which constraints were relaxed

A relaxed board is still a valid output and won’t get thrown away for another attempt. The warnings let you know which constraints were too tight so you can tweak them.

Caution

Tag constraints are never relaxed. If a tag conflict can’t be resolved, generation will fail rather than place conflicting objectives.

Escalation Strategy

Generation has two layers of escalation. Greedy gets smarter across generation attempts, and backtracking gets smarter across restart phases within each attempt. Most boards finish on greedy attempt 1.

Retries only happen on hard failures (tag conflicts or a pool too small to fill the board). If constraints are just tight, constraint relaxation finishes the board without retrying.

Greedy escalates across up to 3 attempts:

Attempt	Candidate Selection	Board Diversity
1st	Random	Maximum, every board is unique
2nd	Random from top ~50% of scored candidates	High, still varied but steered away from poor picks
3rd	Random from top ~10% of scored candidates	Lower, different seeds may give more similar boards

Backtracking runs up to 2 phases per attempt. Each phase uses a different random seed. Normally candidates are random. In constrained mode, they’re scored and the pick is biased toward candidates that preserve future constraint supply:

Phase	Normal Boards	Constrained Boards
1st	Random	Random from top ~50% of scored candidates
2nd (new seed)	Random	Random from top ~25% of scored candidates

Each attempt gets its own backtracking phases, so there’s plenty of room to find valid boards before falling back to constraint relaxation.

---

Modifiers

Several game modifiers can alter generation behaviour by disabling specific constraints, forcing range values, or adding post-generation effects like shinies and blanks.