Skip to content

Repository files navigation

TIGRIS

A board game factory. Named for Tigris & Euphrates (Knizia, 1997).

Claude Code-driven. Markdown-first. Euro tradition. TIGRIS reviews published games, designs new ones, and — most unusually — has built up an empirical map of what "design space" actually looks like by doing so at scale.

Review roles: This repo uses ROLES, the .roles convention for repository-local review panels.

What TIGRIS has done

Reviewed 150 games. Every review is a Parliament: eight designer personas (Knizia, Rosenberg, Feld, Lacerda, Chvátil, Kramer–Kiesling, Stegmaier, Vaccarino) plant incompatible stakes against the design, then a narrated playthrough attacks, defends, and collides them. 69 games went through the full three-phase pipeline; 81 more went through a faster variant to build out the corpus. TRACKER.md has every row.

Discovered TIGER BEAT — a coordinate system for board games. Once you have 150 games scored across 28 axes, you can ask: what are the real underlying dimensions? The corpus answers back with nine. Five big ones, four smaller ones. Every game now has a 9-number profile. That means you can plot games, find neighbors, find gaps, and design into the gaps on purpose.

TIGER — the five primary dimensions (72% of the variance in the corpus):

Dim What it measures Corpus leader
T Tension Stakes, pressure, fear of losing Space Alert, Bridge (7.5–7.2)
I Interaction Direct conflict, hidden info, targeting Letters from Whitechapel, Bridge (8.7)
G Game-spectrum Deep Architecture ↔ Elegant Minimalism (bipolar) On Mars (7.8) ↔ Crokinole (0.3)
E Experiential Social contract, physical affordance, cognitive load Secret Hitler, Hanabi, Codenames (5.7)
R Range Variability, replayability, strategic breadth Welcome To, Sagrada, Qwixx, Crokinole (8.0)

BEAT — the four secondary dimensions (the remaining 28%):

Dim What it measures Corpus leader
B Breadth Solo mode, player-count robustness, multi-mode flexibility Crokinole (9.0)
EA Emotional Arc Narrative journey vs. pure mechanical replay Pandemic Legacy S1 (9.0)
A Accessibility Onboarding ease, first-turn footprint, language independence Pandemic Legacy S1 (9.0)
TX Texture Scarcity bite, pressure shape, floor mechanics On Mars (7.7)

The biggest single finding: G is bipolar. "Deep engine" and "elegant minimalism" aren't two different things you can score separately — they're opposite ends of the same axis. Lisboa is at 6.5. Crokinole is at 0.3. Nothing sits in the middle; the corpus bifurcates.

Here's what TIGER BEAT looks like for games you know. Every one of these was scored by TIGRIS through a full Parliament review:

Game T I G E R Notable BEAT
Tigris & Euphrates 7.2 5.7 3.5 3.0 5.7
Catan 3.0 3.3 2.2 4.0 6.7
Dominion 4.5 5.0 4.5 2.3 6.7
Agricola 5.8 4.3 5.0 3.0 5.3 TX 6.3
7 Wonders 4.2 4.7 3.7 3.0 6.0
Azul 3.5 4.3 1.3 3.0 7.7
Wingspan 4.2 4.0 3.8 2.7 6.7
Brass: Birmingham 5.2 5.3 6.0 3.3 4.7 TX 6.7 · A 7.0
Gloomhaven 6.0 6.0 4.3 4.3 5.0 EA 8.0
Pandemic Legacy S1 6.2 7.0 3.7 3.7 5.0 EA 9.0 · A 9.0
Codenames 4.8 7.3 1.2 5.7 6.7
Crokinole 3.0 4.3 0.3 4.7 8.0 B 9.0
Through the Ages 6.8 5.3 5.3 3.3 5.3 TX 6.3
Lisboa 6.0 6.7 6.5 3.3 5.0 TX 6.3

Read those rows. Crokinole and Azul both live at low G, high R — elegant and replayable. Pandemic Legacy is a story game (EA 9.0) that nothing else in the corpus comes near. Brass and Lisboa occupy the deep-architecture corner that Azul deliberately avoids. Codenames' E score of 5.7 makes it one of three games in 150 that really do the social-contract thing.

Designed 31 original games. Including Parliament (the factory's self-model — mechanics that are the pipeline), CODEX (a Schools-of-Magic engine-builder, first game designed to attack an empty region of TIGER space), and RITE — the first TIGRIS game to land in Gap 1, a region previously empty across 150 games.

Found 5 design gaps. Empty regions where no game sits in the corpus. Each one is a real unsolved problem — with nearest neighbor, a design concept, and five meetup-crowd inspirations to prime a designer's intuition. Two entered this month (Gap 1 via RITE, Gap 5 via CODEX). Three still open.

Grew the rubric from 24 axes to 32. The rubric is forward-only: axes earn adoption through repeated Parliament evidence, or they get retired when they consistently fail. 29 adopted, 1 retired (C5 Anti-Catch-up Pressure), 3 live and gathering data, 144 consecutive games with zero silent retirements. Live state: personas/axis-pool.md.

How a review reads

Every evaluative claim in TIGRIS names (a) an axis from the Pool, (b) the persona making the claim, (c) a player count or moment anchor. "Fun," "well-designed," and "solid" are forbidden words. The factory's job is well-formed disagreement, not smoothed consensus.

The shape of a Parliament: eight designers draft 3 axes each from the Pool in snake order. Each plants stakes. A narrated playthrough attacks, defends, and collides those stakes on adjacent axes. Then deterministic amendment promotes, retires, or holds axes based on what happened — the rubric evolves from the play record, not by committee.

Open any game's panel/SUMMARY.md for a completed review. games/0001-parliament/tierA-stakes.md is a clean example of stakes-drafting.

Where to go next

Simulator

tools/tigris-sim/ is the RALLY-backed board-game simulation pilot. It starts with Parliament and reports seeded axis-collision, adoption, and chair-activity telemetry.

cd tools\tigris-sim
cargo run --quiet -- --seed parliament-smoke
cargo run --quiet -- --seed parliament-smoke --runs 20 --players 4
cargo run --quiet -- --seed parliament-smoke --compare-variants --runs 20 --players 4
cargo run --quiet -- --game upstage --seed upstage-smoke --players 8 --compare-variants --runs 24
cd tools\tigris-sim
cargo run --quiet --bin tigris-muddle-macroquad

The native MUDDLE/Macroquad launcher opens the Parliament AI table directly with default save/transcript/import/export paths. The visual table now shows persona seating, axis vocabulary, collision pressure, tiger markers, dissent tags, the rubric ledger, amendment closeout, phase readouts, subject-card context, and next-action guidance. The first visible table arc is: choose persona, go board, draft axis, stake claim, reveal collision, place tiger, end turn, challenge ai, go score, inspect ai, score amendment, close parliament.

tools\tigris-sim also exposes Parliament as a COURT adoption fixture through parliament_ai_court_snapshot() and parliament_ai_court_validation_packet(). This proves COURT/RACKET can describe a product-owned tabletop slice beyond escape rooms without moving TIGRIS rules or RALLY simulation evidence into COURT.

Status

Rubric v2.24.85 · 153 games · 32-axis Pool · latest: RITE #153 (Gap 1 entered) + UPSTAGE (PARLOR party-game pipeline launched). Resume point: docs/handoff/2026-04-23-post-rite-upstage.md. Full architecture spec: docs/specs/2026-04-19-tigris-v2.0-design.md.

Research

Six papers document the TIGER BEAT framework and its corpus. LaTeX sources in research/publications/; build all PDFs with make -C research.

Games infrastructure

TIGRIS shares its simulation and UX layer with the rest of the Games Design series:

  • MUDDLE — shared room-command UX engine. TIGRIS hosts the Parliament AI table and UPSTAGE sessions through MUDDLE's adapter contracts, with CLI, browser-window, and native Macroquad clients.
  • RALLY — shared simulation and validation substrate. tools/tigris-sim/ uses RALLY for seeded Parliament axis-collision, adoption, and chair-activity telemetry.
  • COURT / RACKET — scalable experience framework and engine adapter. tools/tigris-sim/ exposes Parliament as a COURT adoption fixture that RACKET consumes for compatibility diagnostics beyond escape rooms.

License

MIT — © 2026 Gio Della-Libera.

About

Board Game Factory

Resources

Stars

0 stars

Watchers

0 watching

Forks

Releases

Packages

Contributors

Languages