Next Year, No Fish

One shared lake: what you take today, someone else can't catch tomorrow.

Start here

No sign-up, no install. One of you opens a room and sends the link; the other opens it in a second Chrome profile and joins. Pick a game and press Start — two people are enough for three of them, and Who's Poaching seats four, so leave "bots fill empty seats" ticked. Both screens move together. Reopen the room link later: the lake is where you left it.

What good means here

A commons is a dilemma only when others fish too, so playing together should make it better. Good means:

1. Every game's dilemma is felt within a minute: one more cast or home (Boat Rush); grab now or leave fish for the double round (Fishing Party); whom to believe (Who's Poaching); haul or let it breed (Lake Holdings). Greed costs you through others, not luck.
2. No one can cheat or peek. The server decides every catch, coin and vote, and sends each seat only what it may see.
3. No single move wins, or nobody need think about others.
4. It reads at a glance, in English or Chinese: one goal, one choice per game.
5. The lake outlives the match, restarts and a day away.

Where this comes from

Hardin (1968) called commons doomed; Ostrom (1990) saw them last where users policed each other. Fehr and Gächter (2000): people pay to punish free riders (the buoy and the vote). Milinski et al. (2008): groups fail to avert a shared loss. Salen and Zimmerman (2004): outcomes must be discernible. The title is the Lüshi Chunqiu's: drain the lake, and next year there are no fish.

Not built

Chat (talk aloud), accounts, cross-room leaderboards, restocking a dead lake, run-time AI content.

Enforced and judged

Enforced, by checks that fail if broken:

• A seat's frames and events hold only what it may see; matches replay from inputs (spec/unit/game-platform.test.ts).
• No fisher gets another's role before its reveal (spec/unit/poach.test.ts).
• A lake carries over and dies below 40 (spec/unit/game-platform.test.ts).
• Two cookie-separate sessions join a room over HTTP; the first sees the second act within a second; history survives a restart; each can use its own language (spec/marker-path.test.ts).
• Both dictionaries share keys (spec/unit/i18n.test.ts); every current page renders in both languages (spec/unit/site-pages.test.ts).

Measured, not enforced (300 bot matches per table, lake sized to boats; Boat Rush from docs/checks/balance-unload-countdown-2026-11-07.md, Party and Poaching from docs/checks/balance-party-poach-2026-11-08.md). About bots and seeds, not people:

• Boat Rush, four boats: no bot kind won over 41% of random-mix tables (keeper 26.3%); all-greedy ended Lean or No fish in 65.0%; at 80% of their catch, Fair or better in 72.3%.
• Fishing Party, four boats: no shop habit won over 40% (the best, Fast Boat always, 27.7%); last net held 23.3%.
• Who's Poaching, six boats: fishers won 69.3%; 47.7% always skipping votes.
• Lake Holdings 3.0 (docs/checks/balance-hold3-2026-10-24.md): no bot kind won over 41% (steady 27.4%); all-greedy ended Fair in 77.0%, all-keeper Bountiful in 100%.

Judged by watching: do people name the choice untold, adapt to others, want a rematch?

Known limits

Boats move on the server's next frame; slow links feel late.

References

• Fehr & Gächter (2000). Cooperation and punishment in public goods experiments.
• Hardin (1968). The tragedy of the commons.
• Milinski et al. (2008). The collective-risk social dilemma and the prevention of simulated dangerous climate change.
• Ostrom (1990). Governing the Commons.
• Salen & Zimmerman (2004). Rules of Play.
• Lüshi Chunqiu, "Yi Shang".