How this works

We’re hunting for the next “Dominion → Slay the Spire” pattern — board game mechanics that have a clean translation to a single-player digital game but no one’s shipped it yet. The data flows through four layers, each adding a coarser-to-finer signal. Below: the same game (Heat: Pedal to the Metal) at every layer.

Step 1 · Layer 1

Scraped from BoardGameGeek

The bedrock layer. We pull every game's metadata from BGG's XML API: complexity, ratings, mechanism tags, categories, player counts. ~2,000 games today, refreshed on demand.

Year
2022
BGG rank
#47
Complexity
2.20
0–5
Rating
7.79
bayes
Mechanisms
Catch the LeaderDeck, Bag, and Pool BuildingHand ManagementProgrammed MovementPush Your LuckRaceSimulationSimultaneous Action SelectionSolo / Solitaire GameTrack MovementVariable Set-up
Categories
RacingSports

→ browse all scraped games on /inspect

Step 2 · Layer 2 · LLM-enriched (wide)

Terse design primitives from a v2 prompt

An LLM pass over each game's rules + reviews extracts a small terse-phrase fingerprint: core verb, decision shape, port status, tags. Runs across the whole corpus.

Core verb
play speed cards, manage heat
Decision shape
mixed:combinatorial+probabilistic
Reward schedule
mixed:immediate+delayed
Translation difficulty
Easy
Core loop pitch

Choose a gear, play that many speed cards; corner too fast and you stuff heat cards into your deck.

Primitive tags
gear_shift_hand_sizeheat_card_deck_pollutioncorner_speed_checkslipstream_catch_upchampionship_meta_persistence
MDA aesthetics
ChallengeSensationFellowship
Direct digital port

—

Closest loop translation

none yet

→ ranked translation candidates on /candidates

Step 3 · Layer 3 · LLM-enriched (deep + v4)

Paragraph-length design analysis + archetype fit

On top translation candidates only, a richer v3 pass extracts long-form mechanics analysis, theme/audience, and concrete digital-meta-layer ideas. v4 adds five canonical archetype fits (Balatro, Cozy, Coop, Snap, Wordle).

Core verb (long)

Each round you grip a hand of 7 speed cards and choose a gear (1-4); the gear is the number of cards you must commit face-down behind your screen. Everyone reveals simultaneously, sums their speed values, and moves their car that many spaces. Optional spice: spend Heat to boost (flip the top of your deck for bonus speed) or to shift gears more than one step.

Decision space
Each turn you balance gear-against-hand (which gear can my current cards actually fill), corner-against-tempo (downshift early to handle the apex versus push and pay overage Heat), and engine-against-deck-pollution (every boost or overage seeds a Heat card into your discard, where it will choke a future hand). The choice space at any single decision is moderate — typically 3-7 reasonable gear+hand combinations — but planning depth runs 2-3 turns ahead because gears are sticky (one step free, two steps costs Heat) and corners must be set up in advance. The famous strategic lattice lives in this constraint: you cannot dump from 4th to 1st to dodge a chicane, so cornering becomes a rhythmic deceleration puzzle.
Skill expression
The dominant skill is tempo planning — knowing when to sandbag in 1st gear to flush Heat back to your engine and reset your hand for the next big push. A close second is corner-positioning forecast: judging whether a high-card hand will arrive in time for the next chicane, and whether to pre-emptively burn Heat now to avoid a disastrous overage later. Opponent reading (slipstream timing, gauging when a leader is bleeding Heat) is real but secondary; memory load is near zero and the math is simple addition. The game rewards drivers who think two corners ahead and punishes turn-by-turn play.
Tactile dependency
low
Theme promise

“You are a 1960s F1 driver. Late-brake into the chicane, scream past the leader, feel like a stunt driver — without simulating tire physics.”

Digital meta-layer ideas
  • 5-race roguelite season: procgen tracks, draft 1 of 3 upgrade cards plus 1 of 2 garage relics between races, run ends if you finish last twice
  • Garage / Workshop build-identity: choose engine archetype (more boost vs less heat-payoff) and tire profile (better cornering vs better straights) between runs
  • Daily challenge: fixed track + fixed AI rivals + fixed starting deck + a constraint (e.g. 'start with 4 Heat in deck'), global leaderboard by finish position
Archetype fits (v4 composite, 0–10)
7.6
balatro
5.0
cozy
4.7
snap
4.3
wordle
MDA vector (v4)
Sensation: 1.0Fantasy: 2.0Narrative: 1.0Challenge: 3.0Fellowship: 1.0Discovery: 1.0Expression: 0.0Submission: 2.0

→ explore archetype fits on /archetypes

Step 4 · Layer 4 · Deep dive

A long-form design document

When a candidate looks strong, we generate a Heat-style 7-section deep dive: precise mechanics, what makes it work, existing digital attempts, the translation problem, an iOS prototype spec, risks, references.

Heat: Pedal to the Metal

Deep Dive for an iOS Roguelite Translation

  1. 1. The mechanics, precisely
  2. 2. What makes Heat work as a tabletop game
  3. 3. Existing digital attempts
  4. 4. The translation problem
  5. 5. Concrete iOS prototype spec
  6. 6. Risks and unknowns
  7. 7. References

Project context: Translate the core loop of *Heat: Pedal to the Metal* (Asger Harding Granerud & Daniel Skjold Pedersen, Days of Wonder, 2022) into a single-player iOS roguelite. The "Dominion -> Slay the Spire" pattern, but for a racing deckbuilder. No direct digital competitor …

Read the full Heat deep dive →

→ all deep dives on /deep-dives

Now you’ve seen the data flow

Pick a place to go next

Browse the landscape
Filter all 2,000+ games by year, complexity, mechanic.
See trends over time
Which mechanics emerged when, how complexity has shifted.
Find games for me
Six questions → a Claude Code prompt for personal recs.
Read the deep dives
Long-form design docs on top translation candidates.