Pulse

Backing held is the one that matters. In an ordinary NFT market the floor is whatever the most motivated seller is asking, and it evaporates the moment they change their mind. Here it is ETH already locked behind specific pieces by people who committed to buying them back.

Weight is the counter-intuitive part. Odds are inverse to backing: a lightly backed piece is drawn often and leaves quickly, while a heavily backed one is drawn rarely and sits in the pool. Backing is not a bid for attention. It is a price on being left alone.

The winner gets 24 hours to choose: keep the work, keep it and relist it under their own backing, or sell it straight back into the depositor's standing bid, which pays 85% of backing by default, in ETH or settled as $FWA. Miss the window and the depositor can resolve it; miss that too and anyone can finalise the default after seven days, which is simply that you keep the piece.

Fees split three ways: roughly 1% to the protocol, a 5% tithe to whichever deposit holds the crown, and the rest divided equally across every active position.

The crown is the detail worth knowing, and it does not quite do what it says. One deposit holds it at a time, earning a slice of every acquisition fee, and taking it from a sitting holder does mean out-backing them by 10%. But an empty crown is free: whenever the holder is drawn, withdraws, or trims their backing, the seat opens to whoever asks first, at any size. So a depositor can win it with a large deposit, pull the backing straight back out, and re-take the vacated seat in the same transaction, keeping the tithe on a fraction of what they won it with. It rewards speed at least as much as conviction.

We say this having found it in our own fork of the same code, not by reading theirs uncharitably: the two were identical here. Our pool prices the empty seat instead: it reopens at the backing it was last held at and melts to zero over a week, so the bar comes down with time rather than with another call.

Every so often someone asks the obvious question: why is there no grail in this pool? Put a Punk behind the glass, an Ape, a mega mutant, a Fidenza, and every draw becomes a ticket to something priceless. Lotteries are sold on the jackpot, not the median prize, and volume would go up an order of magnitude overnight. We think the honest answer is that the mechanism cannot hold one, and the reason is worth walking through.

Backing is the only dial, and it does two jobs at once: it funds the standing bid and it sets the odds. A winner keeps anything worth more than 85% of its backing, so a grail is only safe above roughly 118% of its value: back a 50 ETH piece with less than about 59 ETH and every draw is a donation. Back it with more and the weight formula buries it. And on the day it finally hits, the standing bid costs its depositor the piece's full value in cash. There is no backing that makes a grail both winnable and rational to deposit. Priceless is exactly the thing this design cannot price.

That arithmetic is also why the top of the pool is best read as advertising. A deposit backed far beyond its value is not waiting to be won. It is unreachable by construction, and it is paid to be there: it takes the crown and its 5% of every acquisition fee, defensible until someone out-backs it by 10%, on top of the ordinary fee split and an emissions accumulator that runs on all that backing. Whoever holds the headline prize is not cheating the protocol so much as reading it correctly. The design pays its biggest prize to be bait, and the players fund the campaign.

As built, the pool cannot hold a grail. Anything truly priceless inside it is either unreachable or being donated. What sits at the top is not a prize but a picture of one.

If we were TokenWorks, this is the one place we would break the one-number elegance. Give the grail its own slot with published, fixed odds, funded by a rake on every draw that escrows toward the owner's ask. Then the pool underwrites the prize instead of the owner underwriting themselves: consigning a grail becomes a forward sale at your own price that yields while the escrow fills, and players get honest odds and a visible jackpot instead of an implied one. The cost is real: it needs someone's number for what the piece is worth, the oracle everything else here elegantly avoids. But that is the trade. FWA stays honest by never asking what anything is worth, and that honesty is exactly why there is no mega mutant behind the glass.

The settlement data above already gives away half of this, so let us write the other half down. Playing is a fixed-rate donation. A ticket is priced at 1.1 times the harmonic mean of all backing, which, because odds run inverse to backing, is exactly the expected backing your draw lands on. Take the standing bid every time, as almost nine in ten winners do, and you recover 85 ÷ 110, or 77.27% of what you paid, in expectation, before the $FWA on top. Not an estimate; an identity. The missing 22.7% is the price of the keep-option and the emissions, which is to say the game is honest about being a game.

For depositors the surprise is which asset is being spent. It is not the NFT: a winner who can take 85% of backing in cash almost always does, and the piece comes home. It is the backing: every hit hands it to the winner, and the odds formula makes the arithmetic strangely flat. Hit frequency scales with one over backing; the loss when hit scales with backing; the product is a constant. Fee income is split equally per listing, so it is flat too. Net ETH per day is the same whether you back thin or thick, and the percentage yields farmers quote are one flat number divided by ever-smaller denominators.

Backing is the consumable; the NFT is the tool. Every listing earns the same ETH per day; what backing buys is how often you have to reload.

So the real choice is cycle time. Backed thin, you are hit every few hours: rebuy the floor, eat the slippage, sit out the settlement window, run more transactions than a human can, and once in a while a winner keeps the piece anyway. Backed thick, the same expected ETH arrives with a fraction of the churn. The thin cycle is the headline yield, and its margin is thinner than the formula says, because the formula does not price the operating drag. The thick cycle is the one a person can actually run. The equal split has one more corollary: percentage yield scales inversely with the floor price of what you deposit, so the trade concentrates in the cheapest whitelisted collections, and whitelisting itself has become the catalyst, repricing those sets hard once the pool made them ammunition. Depositor demand is the bid under the floor.

None of this is stable, and the pool says so out loud. Between two readings of this page a day apart in late July, live positions grew by roughly a third and the price to play fell in step: every new listing thins the same fee stream, and the yield decays toward the rake that funds it. The schedule is public: emissions end, $FWA becomes buyable without pulling, and the reason the fees exist retires first. A farm this legible ends on time. The traditional sign-off applies: thank you for the dilution.

During emissions both sides earn 1% of supply per day, depositors through an accumulator on their backing, purchasers through a daily pot split across successful draws. Emissions stop entirely when the pool is empty. Afterwards, fees fund a rate-limited buyback that splits 40% to depositors, 40% to purchasers and burns the remaining 20%.

Worth knowing before you touch it: wallet-to-wallet transfers of $FWA are blocked by design. It moves through the configured pool or protocol flows, and nowhere else.

Everything except the settlement breakdown is a direct contract read at the block shown: no indexer, no cached third-party figure. The breakdown is counted from the core contract's own event logs, split into as many queries as the provider's ten-thousand-log ceiling requires and cached for an hour.

We show nothing we cannot point at a contract for. Where the documentation and the deployment disagree, on the crown tithe and the eligible-collection list, we follow the deployment and say so.