← Dev Log

Holding Back the Tiers

In Post 03, we cut room rarity. The problem it was solving didn't disappear — it just went quiet. A German farmer's game just gave it a new voice.

Tiered wooden card display board showing a three-row card market

Post 03 was called "The System Underneath." The headline decision: rarity removed. Every Core Room tile goes into one bag. No tiers, no counting, no random exclusion per run. The bag is the deck.

It was the right call at the time. The tiers were producing complexity without drama — you had to count tiles and track pools to get an effect the player couldn't feel. Cutting it simplified setup, eliminated a bookkeeping step, and let playtesting focus on the actual core loop.

But the problem rarity was trying to solve didn't disappear. It just went underground.

The VG's mansion gets harder as you go north. Early ranks feel accessible; late ranks feel dangerous. The drafting pool in the VG is weighted — common rooms dominate early, rare and unusual rooms surface as you push deeper. That escalating tension is part of what makes each run feel like a three-act structure. In a single flat bag, you don't get that. Every rank draws from the same distribution. The manor doesn't breathe.

I've known this for months. The difficulty_curve_comparison page has been sitting in the repo since April — a side-by-side of Model A (our flat bag) and Model B (rank-gated progressive draw). Open question, pending playtesting. I've been waiting for the right frame to reopen it.

That frame just arrived from a German farmer's game designed in 2007.

Uwe Rosenberg

Career at a glance

269Games credited (as of 2025)
6Games in BGG top 100
1Designer with more top-100 entries

Born 1970, Aurich, Germany. Co-founder of Lookout Games. Studied statistics at Dortmund — his thesis was on probability distributions in Memory. Designs roughly full-time since the late 1990s. Has four children and by his own account has mostly stopped playing other designers' games, preferring to iterate inward on his own mechanics.

The career arc

Rosenberg's career breaks into three distinct phases, each defined by a different relationship between cards and game feel.

Phase 1 — Card games, social mechanics (1992–2006). His first major success was Bohnanza (1997), a bean-trading card game built on one elegant constraint: you cannot reorder cards in your hand. You play from the front. That forced order — combined with a trading system — produces a game about negotiation, desperation, and social dynamics. Rosenberg didn't design a trading game and add hand order as a feature. He designed hand order as the mechanic and let the trading emerge from the pressure it creates.

Bohnanza1997
Fixed hand order forces desperate trades. The constraint is the mechanic. Produces social interaction that no explicit trading rule could engineer.
BGG #415

Phase 2 — Worker placement, agricultural engine building (2007–2015). Agricola (2007) changed the genre. Players are farmers who must feed their families or face penalties — an anxiety engine embedded in a resource game. It held the #1 spot on BoardGameGeek for nearly two years. Caverna (2013) followed: same structure, more freedom, dwarves instead of farmers, underground rooms instead of fields.

Agricola2007
Worker placement + starvation threat. Occupation and Minor Improvement cards create wildly different strategic paths each game. The card hand you draft at game start shapes your entire arc.
BGG #12
Le Havre2008
Port city economics. Variable setups and special buildings ensure each game feels structurally different from the last despite identical rules.
BGG #47
Glass Road2013
Simultaneous role selection. Players choose 5 specialists from a personal deck of 15 — if two players pick the same card, both get a diluted action. Bluffing, deduction, and resource conversion in a compact package.
BGG #290
Caverna2013
Agricola reimagined with more flexibility and less punishment. Cave rooms alongside farm fields. Supports up to 7 players.
BGG #42

Phase 3 — Spatial puzzles, tile-based farms (2014–present). A Feast for Odin (2016) is Rosenberg at his most ambitious: a Viking-age game where players fill a polyomino board with resources and land tiles, scoring based on coverage. Atiwa (2022) and Hallertau (2020) continue experimenting with cards as upgradeable action spaces — flip a card to reveal new capabilities as the game progresses. The spatial and the systemic merged.

A Feast for Odin2016
Polyomino resource placement on personal player boards. Unusual for Rosenberg: mechanisms first, theme discovered afterward. The largest rulebook of his career; also one of the most replayable.
BGG #23

How he works

In a 2016 interview at Spiel, Rosenberg described his process plainly: "The usual approach is to spend more time thinking about mechanisms that encourage play. Once I have a mechanism that gets players going, I quickly start looking for a theme." Mechanism first, then the world that makes the mechanism make sense.

He also described a second approach — playing another designer's game and changing the part he doesn't like until it becomes something new. He noted this is less common now that he has less time to play widely.

His statistics background surfaces in the design. He thinks carefully about probability distributions — what will actually come up, how often, what shape the experience curve takes across a session. This isn't accidental. It's trained.

Known design principles

Constraints produce behavior The fixed hand order in Bohnanza doesn't describe what players should do — it creates a situation where trading becomes the natural response to pressure. The constraint does the design work. This principle runs through everything: the starvation threat in Agricola isn't a punishment system, it's a pressure generator.
Mechanism-first, theme-second Find the thing that "gets players going," then find the world that makes it legible. He explicitly says he doesn't want theme slapped on afterward — but he does look for theme early, as soon as a mechanism is generating play, so the two develop together rather than in sequence.
Replayability through composition, not randomness The same rules can produce wildly different games by changing what cards are available. Agricola's E/I/K deck system (below) is the clearest expression of this. You don't change the rules. You change the pool. The feel shifts completely.
Iteration inward He returns to his own games repeatedly — Agricola begat Caverna, which fed back into Agricola's revised edition. Each iteration explores what the prior version left unresolved. He treats his catalog as a single ongoing project rather than a series of discrete products.

The deck-switching mechanic in Agricola

A fan of Agricola occupation cards spread out, showing illustrated character portraits and card titles including Parish Nurse, Parish Marksman, Parish Drunkard, and Parish Brothers

Agricola ships with three card decks: E (Einsteiger — Beginner), I (Interaktiv — Interactive), and K (Komplex — Complex). Each contains a different set of Occupation and Minor Improvement cards. The rulebook encourages players to mix decks and experiment. Aftermarket decks — Z, L, NL, and regional country decks (France, Netherlands, and others) — continue expanding the pool.

The mechanism is elegant in its simplicity: the rules don't change. The board doesn't change. The worker placement structure is identical. But swap the deck at setup and you get a fundamentally different strategic texture. E-deck produces accessible, friendly games. K-deck produces dense, combo-driven games where experienced players can break rules in layered ways. The difficulty curve doesn't live in the rules — it lives in the card pool.

The difficulty curve doesn't live in the rules. It lives in the card pool.

This is the thing that stopped me when I first encountered it. Because it's exactly the shape of the problem we cut in Post 03.

The ghost of Post 03

In April, we cut room rarity because it was "creating complexity without drama." That diagnosis was accurate. The two-tier room deck was bookkeeping overhead in exchange for an effect players couldn't feel.

But Rosenberg's E/I/K system isn't a two-tier deck. It's a replaceable pool. You don't run two pools simultaneously with weighted draw rules. You choose which pool to use at setup and run one clean bag. The complexity is a setup decision, not a session mechanic.

That's the difference. And it means the problem Post 03 was solving — how to make the drafting pool feel different across experience levels and campaign progression — has a solution we haven't tried.

Proposals

Proposal 1 — Base Bag / Advanced Bag (Rosenberg-style swap)

Two curated room pools. Base Bag: current 57-tile flat Active Bag, unchanged. Advanced Bag: same rules, but seeded with a selection of rooms that have higher gem costs, more conditional mechanics, or greater interdependency (Locker Room, Mechanarium, Pump Room, Showroom-adjacent rooms if reintroduced). Player chooses at setup. One bag, one clean draw, same rules. No mid-session switching.

Preserves everything cut in Post 03. No rank-aware draw rules, no rarity tracking, no parallel piles. Just a different bag for a different experience.

Proposal 2 — Campaign unlocks seeded into bag

Start campaign with Base Bag. As players hit milestones (Studio Additions, Days Played thresholds, Room 46 found), permanently add specific room tiles to the bag. The pool expands over the life of the campaign rather than at setup. Legacy-adjacent but non-destructive — tiles are always returnable.

Mirrors the VG's progressive room unlock system more faithfully. The manor literally grows as the player learns it.

Proposal 3 — Hybrid: Base Bag + rank-gated draw for one slot only

Retain the flat Base Bag for Slots 1 and 2. For Slot 3 only, at Ranks 7+, draw from a small curated "deep manor" pool rather than the main bag. One additional pile, one additional draw rule, applied to one slot at late-game ranks only. The manor gets dangerous at the top without changing anything below.

This is Model B from difficulty_curve_comparison.html, minimized. Least disruptive way to introduce rank-aware escalation.

All three are worth testing. None requires changing the rules the playtester reads on their first run. That's the Rosenberg constraint: the complexity lives in the pool, not the procedure.

Post 03 cut the right thing for the right reasons. The flat bag was the correct prototype decision. But a prototype that survives into a publisher conversation needs to answer a question the flat bag defers: does the manor feel harder as you go deeper? Right now, the answer is not reliably yes.

Rosenberg has been answering that question for thirty years with cards. The method translates.

— QQ · May 31, 2026
← Post 25