How We Make Puzzles That Never Need a Guess
Dev notes · June 2026
If there's one promise we make about every logic puzzle on Funduk Games, it's this: you can always reach the answer by reasoning, never by guessing. No "try this and see if it blows up later," no lucky coin-flips. It sounds simple, but guaranteeing it is the hardest — and most important — part of generating a good puzzle. Here's how we do it.
Why "no guessing" matters
A puzzle that forces a guess isn't really a puzzle — it's a maze with a hidden trapdoor. You can do everything right, hit a wall, and only progress by picking a branch at random. That feels terrible, and worse, it means the puzzle had more than one path (or you got lucky). A puzzle with a single solution reachable by logic respects your time: every step is earned, and the satisfying "click" at the end is yours, not the dice's.
The core idea: generate, then prove
Most of our generators follow the same two-beat rhythm. First we build a candidate board — a layout plus a valid solution. Then, crucially, we run a logic solver over it that is only allowed to use real deduction techniques — the same moves a human would make. If the solver can finish the board start-to-end with no guessing, we keep it. If it gets stuck and would have to branch, we throw the board away (or repair it) and try again. The board you play is one that provably survives a no-guess solve.
This is the opposite of just "make sure there's one solution." A board can have exactly one solution and still require guessing to find it. We test for the stronger property: solvable by pure logic.
How each game does it
- CrownGrid grows coloured regions around a hidden solution, then runs a repair loop: it checks that the board is both unique and logically solvable, and reshapes regions until it is — while also avoiding ugly one-cell regions and oversized blobs.
- Sudoku goes a step further and grades difficulty by which human techniques the solver needed: scanning and singles for Easy, naked/hidden pairs for Medium, pointing pairs and box/line reduction for Hard. The difficulty label is earned by the logic, not faked by removing more digits.
- SunMoon only ships boards a deductive solver can crack using the no-three-in-a-row, balance and = / ✕ rules together — so you never have to gamble on a 50/50.
- Trail guarantees a single continuous path that visits every cell, so the "one line" you're hunting for is the only one that exists.
The cost (and why it's worth it)
This approach is expensive. We sometimes generate and discard dozens of boards before one clears the no-guess bar, and the solver runs in full on every attempt. But the payoff is a stack of puzzles you can trust: hit a wall, and the next deduction is always there to be found. That trust is the whole product.
Want to put it to the test? Every board is free and unlimited, with a daily challenge too.