A Good Snowman is Hard to Build
game
7/14/2026

A Good Snowman is Hard to Build

byDraknek
8.7
The Verdict
"A Good Snowman is Hard to Build is a small game that's far smarter than it lets on, and its greatest trick is convincing you it's harmless. The cozy shell is real — the calm, the hugs, the forgiving undo — but it's a delivery mechanism for some of the most refined grid-logic design on mobile. The short campaign is a fair knock, and hiding the Dream World behind the credits is a genuine mistake that will cost casual players the best hours the game has to offer. But what's here is crafted with obvious love and unusual precision. Draknek made a puzzle game brave enough to be gentle and rigorous enough to be respected. Take the hug. Then stay for the part that fights back."

Gallery

Screenshot 1
View
Screenshot 2
View
Screenshot 3
View
Screenshot 4
View

Key Features

Growth-Based Sokoban: Snowballs aren't static blocks. Roll one across fresh snow and it grows — small to medium to large. Every push changes the puzzle's state permanently until you undo, turning simple pathing into spatial calculus.
The Stacking Constraint: A finished snowman needs large-on-bottom, medium-middle, small-on-top. That ordering rule quietly governs everything — where you let each ball grow, and where you dare not.
Instant, Infinite Undo: No fail states. No penalties. A single button rewinds your mistakes, which reframes the entire game from a test of execution into a space for pure thought.
Name and Hug Your Snowmen: Finish one, and you can name it and give it a hug. It's twee. It also works, anchoring the abstraction in something weirdly tender.
The Hidden Dream World: A post-game layer of interconnected metapuzzles that abandons the gentle curve entirely and asks you to think across rooms, not just within them.

The Good

Elegant, deceptively deep growth-Sokoban design
Instant undo removes stress and encourages real experimentation
Genuinely charming, sincere emotional framing
Flawless performance and crystal-clear visual legibility

The Bad

Main campaign is short (~2–3 hours)
Best content (Dream World) is hidden and easily missed
Front half may feel too gentle for hardcore puzzlers
Dream World's difficulty spike is steep and unsignposted

In-Depth Review

Bottom Line: Draknek has smuggled a genuinely fiendish Sokoban into a game that looks like a greeting card, and the disguise is the point. If you finish the main rooms and stop, you've only seen the polite half.

The Gameplay Loop

The genius of Snowman is that it hands you all its rules in about ninety seconds, then spends hours proving you didn't really understand them. You have three snowballs to build into one snowman. They grow as they roll over snow. They stop growing on bare tiles. Stack them big-to-small. That's it. That's the whole ruleset.

The trap is that snow is a finite, consumable resource baked into the board's geometry. Every tile of fresh snow you roll over is snow you've spent. Grow a ball one square too far and it's too big for the middle slot — and there's no shrinking it. So the real puzzle isn't "how do I build a snowman," it's "how do I ration this exact patch of snow across three balls that all need different final sizes, in a room the size of a postage stamp." You're not solving a maze. You're solving a subtraction problem disguised as one.

This is where the tight grids earn their menace. Corners become traps. Push a large ball against a wall and, in classic Sokoban fashion, it may be stuck forever. The monster can only push, never pull, so every move is a commitment you can see coming a beat too late. The difference between Snowman and a punishing traditional Sokoban is the undo button, and I want to be clear about how much design weight that single feature carries. By removing punishment, Draknek removes friction between thinking and testing. You experiment freely. You try the reckless idea. The game becomes a conversation with the board instead of a series of held-breath commitments. It's the same reason a whiteboard beats carving in stone — the cost of being wrong drops to zero, so you're braver, and being braver is how you actually solve hard puzzles.

Difficulty and Pacing

The main campaign is generous. Rooms escalate in complexity at a measured pace, and few will truly wall you. This is also the game's most legitimate criticism: the core runs short, roughly two to three hours, and a chunk of that is pleasantly breezy. If you buy Snowman expecting a marathon of brutal logic gates from minute one, the front half will feel like a warm-up that forgot to end.

It is a warm-up. The Dream World is the real exam.

Unlocked after the main game, this hidden layer discards the room-by-room structure for metapuzzles — solutions that depend on state and understanding carried across multiple screens, demanding you re-read mechanics you thought you'd mastered. This is content built for the people who finish most puzzle games annoyed they weren't harder. It's dense, occasionally opaque, and genuinely brilliant. The design decision to hide it is bold and, I'd argue, slightly self-defeating: plenty of players will roll credits on the cozy campaign and never learn that the game's best material was sitting behind the curtain. The marketing sells you the hug. The masterpiece is the part nobody warned you about.

The Emotional Layer

I came in cynical about the naming-and-hugging. I left convinced. Giving each completed snowman a name and a hug does something quietly smart: it converts an abstract logic puzzle into a small act of care. You're not clearing a level. You're building a little friend and being kind to it. That framing lowers the temperature of failure and gives the whole experience an emotional throughline most puzzle games never even attempt. It's not deep, but it's sincere, and sincerity is rare.

Editorial Disclaimer

The reviews and scores on this site are based on our editorial team's independent analysis and personal opinions. While we strive for objectivity, gaming experiences can be subjective. We are not compensated by developers for these scores.