The Compass that points itself
A design discussion — and a small data mystery that solved itself
There’s a phrase I come back to often when I try to explain how Crowns & Towns came together: the game practically designed itself.
It sounds like a cop-out — designers always say things like that after the fact, when everything has settled into place and the rough edges have been smoothed away. But in this case I mean it almost literally. The scoring system, the map geometry, the balance between Crowns and Towns — each of these decisions came from a single guiding principle: what makes this puzzle feel right to solve? And again and again, the answer arrived on its own.
The daily maps are the most recent version of that story. And they came with a small data mystery that, once solved, confirmed something I had only sensed intuitively.
What Makes a Good Daily Puzzle?
When I started building the daily puzzle rotation, the question was simple: of the hundreds of maps I can generate, which ones belong in the hands of a player who opens the app over their morning coffee?
The answer had to be felt, not calculated. A good daily puzzle should:
- Have one clear best path — or close to it. A map with forty valid solutions isn’t a puzzle, it’s a maze with too many exits. You want the player to feel like they found something.
- Give the player enough to work with. A map with five total builds is solved almost by accident. You want enough build options that the solution requires genuine choices.
- Reward strategic depth. The bonus structures — Knights near mountains, Farms near lakes — should be present and meaningful. They’re the game’s way of saying look closer.
These three things became the curation criteria, in exactly that order:
- Fewest Solutions — uniqueness first
- Most Total Builds — complexity second
- Most Bonus Builds — strategic depth third
I didn’t write these criteria because I understood the theory. I wrote them because they matched what I felt when I played a map and thought: yes, this one.
The Ratio Nobody Was Looking For
Crowns & Towns has a scoring system built around a four-tier dependency chain. Every Crown you place needs 2 Knights to fully score. Every Knight needs 2 Towns. Every Town needs 2 Farms. All four build types contribute points — Crowns the most, Farms the least — but no tier reaches its full value without the support of the tier below it.
Players discover quickly that there’s a natural “best ratio” to aim for. If your score is roughly twice your total builds, you’ve balanced the entire chain efficiently — every tier supporting the one above it without waste. The 2:1 ratio becomes an implicit compass. You’re not told it’s the goal; you infer it as you play.
I’ve always believed that ratio was a design virtue — something worth mentioning to players as a guiding star. But here’s what I never calculated: how many of the daily maps actually land at exactly 2:1?
When I finally ran the numbers across all 368 daily puzzle maps, here’s what came back:
- 73.9% of all daily maps hit the exact 2:1 ratio
- 89.7% fall within 10% of it (the 1.8–2.2 range)
- Zero maps exceed 2.2 — the distribution is bounded above by design
And the maps in the remaining 26%? They cluster between 1.7 and 1.95. Not random outliers — a coherent band of maps that are almost 2:1 but not quite.
What makes this more striking: the larger maps in the printed Crowns & Towns puzzle book — 4×6, 4×8, 5×6, and bigger — carry more pieces, more build combinations, and considerably more room to get lost in. And they land at essentially the same ratio distribution. Roughly 75% hit exact 2:1. The 4×4 daily format isn’t a simplified version of the full game. It’s the full game, concentrated. If you want to live in that math longer, the book is where it opens up.
That 26% number stopped me cold. Because if you had asked me before I ran this analysis — what percentage of your puzzle maps should fall outside the “perfect” ratio? — I would have said: around 25%.
Why 25% Is the Right Number (and I Didn’t Choose It)
Here’s why that proportion matters, stated as directly as I can:
Too few outliers — say, 5% — and the 2:1 ratio becomes a rule, not a compass. Players learn it, apply it, and the puzzle becomes a checklist. Did I hit 2:1? Yes. Done. The game stops being interesting.
Too many outliers — say, 50% — and the 2:1 ratio becomes frustrating. Players spend half their sessions chasing a score that the map was never going to reward. The compass spins. Trust erodes.
But at roughly 1-in-4, the ratio occupies the exact cognitive space a puzzle designer wants it to occupy. You chase it every time, because you’ve hit it enough to believe it’s reachable. And you don’t always hit it, but not so rarely that you’ve given up trying. Every session ends with a version of: was that my best possible score, or was there a 2:1 in there that I missed?
That uncertainty — that beautiful, productive uncertainty — is the point.
And I didn’t engineer it. The curation algorithm selected for it without knowing it was doing so.
Why the Algorithm Produced This Without Trying
This is the part that genuinely surprised me.
The three criteria — low solutions, high builds, high bonus — do not mention the 2:1 ratio. Not once. They’re not optimizing for it. They’re optimizing for puzzle quality in terms of uniqueness, complexity, and strategic depth.
And yet they produce a 74% hit rate on 2:1, almost by necessity. Here’s the mechanism:
A map with many total builds and a strong unique solution is a map where the solver has explored a large space and found a tight answer. That space tends to be populated with Crowns — because Crowns are worth more, players build toward them, and a map that rewards that efficiently produces 2:1 scores.
The maps that don’t hit 2:1 tend to be the ones with the highest raw build counts and the most bonus builds — the most complex puzzles. Look at the top-20 non-2:1 maps in the dataset: most have 14–15 builds and 4 bonus builds. These are the maps where the bonus structures pull the solver toward Knights (bonuses require mixed build types), which drags the ratio below 2.0.
In other words: the hardest puzzles are the outliers. They’re the ones where hitting 2:1 is genuinely difficult because the map rewards a different kind of strategic depth. Bonus builds, by their nature, shift the balance toward Knights.
So the 26% outliers aren’t randomly distributed — they’re the richest maps. The ones most likely to make a player think hardest, question their best answer, and come back again.
What This Means for Players
If you play the daily puzzles regularly, you’ve probably already internalized the 2:1 compass without being told about it. You’ve felt the satisfaction of a clean 2:1 and the nagging doubt of a 1.8 — was there a better layout hiding in there?
What you might not know is that the maps designed to challenge you most are exactly the ones where 2:1 is genuinely out of reach. The game isn’t broken when you can’t hit it. The map is simply built around a richer kind of strategy, where the bonus placements compete with the scoring ratio and you have to decide which one to chase.
That tension — maximize score vs. maximize bonus completeness — is what makes the outlier maps feel different. Harder, yes. But also more layered.
The Game Practically Designed Itself
I said earlier that this phrase sounds like a cop-out. Maybe it is. I made thousands of decisions building this game. None of them were accidental.
But some decisions create structures that produce outcomes the designer never explicitly intended — outcomes that, once you see them in the data, feel more right than anything you would have consciously designed. The 2:1 ratio is a player-discovered compass. The 74/26 split is a balance nobody calculated. And the fact that the hardest maps are the ones that resist the ratio most — that’s not a feature I shipped. It’s an emergent property of the scoring system and the geometry of the maps.
Here’s another one, and I still smile when I think about it. The game evaluates builds by looking at groups of three cells arranged in specific patterns. There are four possible patterns. When I first sketched them out — before naming them anything — I noticed that they look like a Crown, a Knight, a Town, and a Farm. The shapes named themselves. Or rather, the names I had already chosen for the game’s pieces matched exactly the visual impression of the shapes the engine had independently arrived at. I didn’t design that correspondence. I discovered it.
The game designed that part of itself.
And when a designer gets to say that honestly, it means something went right at a very deep level.
This article was written based on an analysis of the full daily puzzle dataset (368 maps, May–July 2026). The ratio distribution, simulation of alternative ranking criteria, and build count comparisons were all run against the live curation data used to build the daily puzzle rotation at crownstowns.com.