💩 SHIT DROP

You'll never win.™

A parody casino. Four rigged games, a 6.57% payout rate printed honestly on every screen, and a jackpot siren that goes off when you lose money. You start with 5,000 credits.

This is satire. No real money, payments, deposits, or accounts exist here. Your "balance" is a number in your own browser's localStorage, the cashier accepts nothing, and there is no server. The point is to take the psychological machinery of real gambling sites and turn the dial up until it's visible.

---

The joke

Real gambling sites are built on a small set of tricks. This one uses all of them, and labels each one as it happens:

The trick	What it looks like here
A net loss presented as a win	Any payout above zero fires the full jackpot sequence — sirens, confetti, screen shake, a gold counter rolling up. Bet 50, get 0.01 back, and the screen screams JACKPOT. Your actual net loss is printed underneath in 8px grey.
Engineered near-misses	82% of total losses show you an almost-win first. The plinko turd lands in the jackpot and is then slid sideways into NOTHING. The slots stop with 🖕🖕 and roll the third one past the line, slowly. The wheel's arrow enters the JACKPOT slice, sits there, then rocks back out. None of it is bad luck; all of it is functions.
The tiny real win	About 0.5% of rounds genuinely pay out. A machine that never pays is abandoned in a minute, so it pays — just enough to keep you in the chair.
Six named rules	The odds change based on what you're doing: beginner's luck (your first three goes are generous), comeback tax (win twice and it's taken back), big bet penalty (betting 100+ halves your payout rate), session decay (worse after 60 goes), plus crumbs thrown when you're on a losing streak or nearly broke. Whichever rule is currently bending your odds is named on screen, in the corner of the game, before you press anything.
Gamification of losses	Fourteen achievements. Every single one is for losing. The VIP ladder fills up based on money destroyed, and it tells you the reward is nothing.
Dark patterns	A welcome bonus that expires in 4:59 and resets forever. A cookie banner whose "Reject all" button runs away from your cursor and then accepts everything. A cash-out button with a 1,000,000-credit minimum. A cashier that pays 2,000 credits for a kidney and 1 for a promise.
Fake social proof	A live-winners ticker of amounts nobody can win, with the truth spliced in between the entries.
Abuse	Full profanity. It insults you after every round, harder as your losses grow. It's more honest than "so close!"

The games

• 💩 Shit Drop — plinko. Twelve peg rows, thirteen buckets, JACKPOT at both far edges and NOTHING filling the middle. The fall is honest: a real bounce sequence that hugs the wall and genuinely lands in the jackpot. Then the turd just slides sideways into whatever the rig decided — no bounce, no arc, no sound, no pause, no explanation. Nobody misses it.
• 🎰 Shit Slots — three reels, one payline. Reels one and two land 🖕🖕 on losing spins and the third crawls the last one out of the window over two full seconds.
• 🚽 Flushpoint — a crash game. Your 💩 climbs out of a toilet along a line; the higher it gets the more it's worth, and the more water is waiting for it. Grab it before the flush. If the round was already a loss, the flush is retro-fitted to land exactly 0.01 under the multiplier you clicked at, and the poo slides all the way back down its own line into the bowl and gets flushed. If it was a win, you grab it fine and then get charged a ~97% "handling fee", itemised.
• 🎡 Wheel of Misfortune — twelve big slices: one JACKPOT, one LOSE ALL, ten insults and rounding errors. The overshoot into the jackpot is exactly one slice wide, by construction, and the wheel rocks back out of it.

How the rigging actually works

Everything settles through one function: settle() in js/core/rig.js. It runs before any animation starts. The reels, pegs, wheel and toilet are theatre replaying a result that already exists.

~64%  →  payout 0
~32%  →  payout 0.1%–35% of your stake   (a loss, presented as a win)
~4%   →  payout ~1x                      (break even, presented as legendary)
~0.5% →  payout 2x–3x                    (a genuine win)
                                          ────────
                          real RTP ≈ 6.6%,  house edge ≈ 93.4%

and betting more makes it worse, on purpose:
   bet  10  →  you get back 6.5%
   bet 500  →  you get back 3.4%

The second half of the design is that presentation is deliberately decoupled from the result:

// rig.js — the smaller the crumb, the bigger the party
if (ratio < 0.02) return weighted([['JACKPOT', 6], ['MEGA WIN', 3], ['BIG WIN', 1]]);

Read the file. The six rules are forty lines and they are the whole business model of an industry.

Sound

Every sound is synthesised at runtime with WebAudio — there are no audio files in this repo. Farts are a sawtooth dragged through a resonant lowpass while its pitch wobbles down a random staircase, with a wet noise burst on the tail; there are four variants, from "short" to "catastrophic". There's also a jackpot siren, an air horn, a sad trombone, coin cascades, wet squelches, ratchet clicks, a toilet gurgle and a 1.7-second flush.

Every play logs [sfx] name#n to the console, which is also how the audio layer is tested.

Running it

Zero build. No dependencies, no node_modules, no backend.

Double-click serve.command. It picks a free port, starts a static server and opens your browser. Or do it yourself:

python3 -m http.server 8080
# → http://localhost:8080

Or drop it on GitHub Pages as-is.

"Why can't I just open index.html?"

You'll get Cross-Origin Request Blocked … CORS request not http. Browsers treat every file:// document as its own opaque origin, so a <script type="module"> can't import its siblings — the imports are blocked before they load. Nothing to fix: it needs to come off an HTTP server. Any static server will do.

Debug flags

Query param	Effect
?rig=nothing	force a total loss
?rig=win / ?rig=jackpot	force the signature gag: a payout that's still a net loss, at maximum volume
?rig=breakeven	force a ~1x return
?rig=real	force a genuine win
?seed=123	seed the RNG for reproducible rounds

e.g. http://localhost:8080/games/slots.html?rig=jackpot

Tests

There are three, and they exist to check the claims above rather than to protect the code:

python3 -m http.server 8080
#   http://localhost:8080/_test/core-gag.html      the core gag, in all four games
#   http://localhost:8080/_test/persistence.html   state survives rounds and page loads
node _test/rtp.mjs                                 # measured RTP over 200,000 rounds

core-gag.html drives each game in an iframe and asserts, per game, that a forced "win" pays out less than the stake, that the balance goes down, and that the full jackpot fanfare fires anyway. rtp.mjs simulates 4,000 whole sessions through bank.js — it has to, because the rules react to your session — and produces every number quoted above. If you change the rig, re-run it and update them. See _test/README.md.

Layout

index.html                 lobby
games/*.html               one page per game
js/core/rig.js             the rigging engine — the only file that decides anything
js/core/round.js           round lifecycle: stake → rig → animate → pay → insult
js/core/bank.js            credits, bet sizing, streaks, lifetime stats
js/core/store.js           the only thing that touches localStorage
js/core/sfx.js             WebAudio synth (farts included)
js/core/fanfare.js         the win theatre: confetti, shake, rolling counter
js/core/taunt.js           the insult dictionaries
js/core/achievements.js    trophies, all of them for losing
js/core/chrome.js          shared casino furniture
js/core/gameui.js          the shared control rail
js/games/*.js              the four games — animation only
css/*.css                  casino chrome

Seriously, though

This exists because the tricks above are effective, and they're much easier to notice at 100x scale with a fart sound attached. If real gambling has stopped feeling like a choice: BeGambleAware.org, or GamCare on 0808 8020 133 in the UK.

Licence

MIT. See LICENSE.