Sriganesh Srinivasan
← All work

$20 Blind Draft.

A two-player bidding party game you can play on one phone or two

2026–now · Solo — design, frontend, backend · Active

Live link ↗Code ↗
$20 Blind Draft
Problem
The "$20 draft" party game only works if nobody knows what's coming next. Played remotely, whoever holds the shuffled deck could peek at it.
Built
A SvelteKit game whose rules are one pure reducer. Online play runs that same reducer in Supabase Edge Functions behind row-level security, so clients only ever see a redacted copy of the game, and simultaneous moves are settled with optimistic versioning.
Stack
SvelteKit, TypeScript, Supabase, Postgres, Deno, Vitest, Playwright
Outcome
Live, with no accounts needed, and found on Google by about 200 visitors a day. 190+ tests across unit, real-browser component and end-to-end suites, including a seeded simulation of 840 full drafts per run. CI runs 4 parallel jobs and gates on axe-core accessibility checks.

What it does

Two players each get a fake $20 and a few empty roster slots. Items from a category (NBA players, foods, movies, or a list you type yourself) are revealed one at a time, so nobody knows whether something better is coming. Either player opens the bidding, they raise each other, and when one concedes, the other buys it. Every item ends up on someone’s roster.

Real players

Around Oct 10, Google search traffic jumped from about 40 visitors a day to about 200, from the US, the UK, Australia and elsewhere. In 30 days the site had about 3,900 visitors and 11,000 page views.

Vercel Analytics chart over 30 days: about 40 visitors a day until Oct 9, then 175 to 300 a day. Totals: 3,879 visitors, 11,352 page views.
Vercel Analytics, last 30 days. Vercel counts a visitor per day, so returning players count again.

Pageviews don’t prove anyone plays, and a local game never leaves the home page, so Vercel’s bounce rate counts a finished game as a bounce. I wanted real numbers, so:

How it works

$20 Blind Draft database schemaSix Postgres tables. Only the server can read rooms, room_players, games and request_log; rooms has up to two room_players and at most one games row, which holds the full game state including the shuffled deck. Browsers can read only game_public, a redacted projection of games, and public_room_listings, which a trigger keeps in sync while a room is public and open. Player tokens and the full game state never reach a browser.SERVER ONLY · NO READ POLICIESBROWSERS CAN READ (RLS)10..2 seats10..1REDACTEDCOPYTRIGGER · WHILE PUBLIC + OPENroomsiduuid PKcodetextvisibilityenumstatusenumcategory_idtextvariant_idtextcategory_labeltextvariant_labeltext?budgetintslotsintcreated_attimestamptzlast_active_attimestamptzroom_playersiduuid PKroom_iduuid FKseat0 | 1nametexttokentextjoined_attimestamptzlast_seen_attimestamptzgamesroom_iduuid PK FKstatejsonbversionintupdated_attimestamptzrequest_logidbigserial PKactortextactiontextcreated_attimestamptzgame_publicroom_iduuid PK FKpayloadjsonb, redactedversionintabandoned_bysmallint?updated_attimestamptzpublic_room_listingsroom_iduuid PK FKcategory_labeltextvariant_labeltext?budgetintslotsintcreated_attimestamptz
The Postgres schema, from the migrations. Orange dots: secrets that never reach a browser. Browsers can only read the two tables on the right.
  1. Jul 2026Pass-and-play on one phone, with the rules engine and the full content pools.
  2. Aug 2026Online play: schema, row-level security, Edge Functions and realtime updates. CI added.
  3. Aug 2026Fixed quitting, the self-award exploit, and invite links that seated nameless guests.
  4. Sep 2026Keepalive fix, then a cleanup that cut 1,383 lines of stale comments and dead code.
  5. Oct 2026Google traffic jumped from about 40 visitors a day to about 200. Added PostHog for game events and errors.

What broke

Testing

I tested at three levels, each for a different reason:

CI runs these as 4 parallel jobs, so a red X says what broke. The end-to-end job boots a local Supabase stack and fails, rather than skips, if the online tests can’t run.

Next projectLegal AI →