The mechanism

Nobody decides the answer

There is no oracle to trust here and no admin who calls the result. The price lives in a pool on this chain, the contract reads it, and anyone can recompute the same answer from the same public data.

30 MIN 30 MIN OPENING AVERAGE CLOSING AVERAGE the session · six and a half hours PICKS LOCK THE BELL

Higher at the close than at the open is up. Both ends are averages, not instants, and that is the whole defence.

01

Why an average, not a price

A single price settles on one instant, and one instant is cheap to buy. The shallowest market here holds around a quarter of a million dollars — enough that someone could shove the price through it for the length of one block and put it straight back, deciding the day for the cost of the spread. The markets page shows what each pool actually holds, read live rather than written down here.

Against a time-weighted average that stops working. To move the settlement you have to hold the price away from the wider market for a real share of half an hour, paying arbitrageurs on every block you hold it — and then do it again at the other end of the day, in the same direction, without anyone taking the free money in between.

The pools are Uniswap v3-shaped but were deployed without the observation array, so observe() reverts and there is no built-in average to read. The contract keeps the same accumulator outside the pool instead: anyone may call poke(), which credits the last observed price for the time it actually stood. A price that appears and vanishes between two pokes is credited no time at all.

02

Where the money comes from

The pot is the token's own trading fees. Picking is free — you pay gas and nothing else — and that is deliberate rather than generous: a player who stakes nothing is entering a contest with a prize, not making a wager on a security.

There is no entry fee, and no function to add one.

Everyone who calls it right splits that day's pot. If nobody does, it stays where it is and tomorrow is worth more.

03

What stops a thousand entries

A holding requirement, checked when you pick and again when you claim. One stack of tokens walked between a hundred wallets fails the second check, so a hundred entries need a hundred stacks held at the same time.

It is a blunt tool. It does not stop someone genuinely wealthy from entering many times — it only makes it cost what it should.

What this can't do

Every design has edges. These are the ones worth knowing before you play, and they are on this page rather than buried because finding them yourself later is worse.

The six are not equally deep
Averaging makes moving a price expensive in proportion to how much money sits in the pool, and these pools are very different sizes — the largest holds several million dollars, the two smallest around a quarter of a million each. Those two are the cheapest place to try it. The markets page shows what each one holds, live, so the differences are visible rather than described.
Market holidays
Weekends are skipped. The nine or so holidays a year are not — several have their own weekday rules and weekend shifts, and getting them wrong would be worse than not claiming them. On Thanksgiving a round still runs and settles on whatever the token did in 24-hour trading, which is thin.
Daylight saving
The contract computes the New York close from the rule that has held since 2007. If that rule is ever changed, the arithmetic quietly becomes wrong and there is no owner who can correct it. That is the cost of having no owner.
The token, not the share
These tokens track the real stock because they are redeemable and arbitrage closes any gap. If that link ever broke, the game would keep settling — on something that was no longer the stock.
A flat day is down
The contract asks whether the closing average is strictly higher than the opening one, so an exactly unchanged price — and anything the arithmetic rounds to unchanged — resolves as lower. It is a coin very slightly not fair, and better said than discovered.
Nobody has reviewed this
The contracts compile and pass their tests, and running them end to end on a local chain found three bugs the tests had missed. That is a reason to expect a fourth, not a reason to be confident. Treat it as unproven software, because it is.