Would You Fund It? (repo: pitch-judge-investor-game)

Play: https://theguysaccount.github.io/pitch-judge-investor-game/ · Every pitch (spoilers): /deck/ · Screens: docs/screens

A daily startup-judgment game. You read real pitches as investors saw them, with the company name hidden: the year, the stage, the traction and the ask. You decide invest or pass, then see what actually happened and its source. Each outcome is one of three:

• 🦄 Breakout: $1B+ via acquisition, IPO or private valuation;

• ✅ Success: acquired or a lasting business below $1B;

• 💥 Bust: shut down, bankrupt or sold off for a fraction.

• Daily: the same five pitches for everyone, by UTC date, with a Wordle-style share grid (🦄 ✅ ❌).

• Practice: ten at random, no limit.

• Case files: one page per pitch (76), titled for the searches people make ("Would you have funded Doorbot?").

• TV pitches: you are scored against what the show's investors actually did.

• Share cards for every score; an embeddable "today's pitch" widget.

No accounts, no cookies, nothing uploaded: results stay in the browser.

The deck

data/pitches.json: 76 pitches (1997-2019), with 24 breakouts, 18 successes and 34 busts. 11 come from a televised pitch show. Every pitch:

• is written in our own words from facts that were public at the time;
• has an outcome sourced to a public page (mostly Wikipedia, which cites the reporting);
• passes an automatic check that the visible pitch does not contain the company's distinctive name.

Busts make up 45% of the deck, so "always invest" scores badly.

How it runs (no human in the loop)

Job	Workflow	When	Fails safe because
Validate the deck, test, build, render share images, deploy to Pages, ping IndexNow	deploy.yml	on content or code pushes; weekly Monday 08:11 UTC; by hand	an invalid deck, a failed test or a failed build stops before deploy, and the last good deploy stays live
Live pages, deck, embed, and every cited source link	selftest.yml	weekly Thursday 06:29 UTC	read-only
Keep schedules alive	keepalive.yml	called by every scheduled workflow	runs even if the caller fails
Tests (Python 3.9/3.12 + Node)	ci.yml	every push	n/a

New pitches and newly resolved outcomes come from a quarterly Claude routine (see OPERATIONS.md). The daily game needs no rebuild: the browser picks today's five from the date.

Layout

judge/data.py     load + validate the deck (sources, tiers, anonymisation)
judge/site.py     static site generator (game pages, case files, deck, score pages, embed, sitemap)
site/assets/      core.js (daily selection, scoring, share text), game.js, embed.js, app.js, style.css
data/             pitches.json (the deck), backlog.json (candidates awaiting outcomes)
scripts/          og.py (share images), indexnow.py, health.py (live + link check), screens.mjs
tests/            unittest suites + Node tests for core.js

Run, test, deploy

python3 -m judge.site                    # build dist/
python3 scripts/og.py dist               # optional: share images with local Chrome
python3 -m http.server 4751 --directory dist
python3 -m unittest discover -s tests -t .
node --test tests/js.test.mjs

Deploys run from deploy.yml on push to main, or with gh workflow run deploy.yml.

Licence

Code MIT; deck text CC BY 4.0. Company names identify the companies described. A game, not investment advice.