Idea review · the method and eight acts
The mechanism in one sentence: a seller says in public, “this item will sell for price P by deadline T,” and backs the claim with a deposit that slowly drips away to the people who bet against it; supporters bet “for” it, are allowed to help make the sale happen, and share the reward the seller announced if it works. Everyone can see the dispute, and the dispute itself doubles as advertising. This page doesn’t describe the mechanism (there’s a separate page for that). It answers one question: why it should hold up. Short answers to common objections are in the FAQ; this page shows where they came from.
Before you read · The mechanism in a minute
The seller says, “this item will sell for price P by deadline T,” and puts down a deposit. The deposit drips: every day, part of it flows into a shared pot for the people betting against it. The doubter bets “against.” Whatever happens, the doubter’s stake itself is never touched—but their vote isn’t free: they pay an entry fee, and their money stays locked up for as long as the stake is in. That’s why their signal is weaker than the seller’s deposit. The supporter bets “for” and helps make the sale happen—with ads, contacts, their audience. Influencing the outcome is allowed here; that’s the whole point. All they risk is their drop—a small share of their stake that goes to the doubters if the lot doesn’t sell; its size is known in advance. It’s a “prove you believe it” filter.
Sold by the deadline: the doubters don’t get the pot (their stakes are untouched), the seller gets back part of what dripped (40% for NFTs, 55% for expensive items; the platform keeps the rest), and the supporters share the reward the seller announced, paid out of the sale price—each in proportion to how much they put in × how long they held it. Not sold: the doubters split the pot by the same formula, and the supporters get their stakes back minus the drop. The seller can also give up early—withdraw the lot and pay the penalty: hand over what has dripped plus a small compensation payment to the doubters; that is always cheaper than faking a sale. Everything below builds on this.
The first version covers only NFTs and tokens: for them, the sale is itself a blockchain transaction, so the contract sees it directly, with no middleman (more in “Who confirms the sale”). Cars and real estate, which come up below, are the next step.
Act 0 · The method
You don’t have to take the main points on faith—you can check them yourself. Test the defense against cheaters in the demo lot: it runs the mechanism’s real rules, with sliders and two numbers: what cheating earned and what the honest path would have cost. The sources are listed at the end of the page, and you can look up every one.
Every claim below went through the same procedure: first a search for the primary source, then a refutation check. As for who did the checking: three independent AI reviewers—not human experts, but separate AI agents in our research pipeline. Each one tried to refute the claim using the sources, not to confirm it; their job was to find where the claim breaks. Next to each point there’s a score: confirmed 3–0 means “three tried to refute it and couldn’t,” refuted 0–3 means “it didn’t hold up.”
What was refuted wasn’t thrown away—it’s on display, including appealing legends we ourselves wanted to use in telling the project’s story. Research that looks for confirmation finds confirmation; ours looked for refutation. That’s the main thing that sets the document below apart from typical crypto “research.”
Act 1 · Originality
The whole combination—the seller’s dripping deposit that flows to those betting against, plus legal influence on the outcome as the intended activity, plus two opposing sides that split payouts by how much they put in × how long they held it—doesn’t appear in any project we found. At the same time, each piece on its own has been road-tested by others, and that’s good news: what’s new is the combination, not unproven parts.
An important caveat: the classic theory people usually cite to calm worries about manipulation in prediction markets explicitly leaves out influence on the outcome itself.² Here, influence is legal and intended. We’re on ground the theory hasn’t covered, so the mechanism can only be tested by calculation—a simulation with a cheater inside (Act 7)—not by pointing to a ready-made theory.
¹ Alex Tabarrok’s idea; lab tests: Cason, Tabarrok, Zubrickas, 2021. ² Hanson & Oprea, 2009.
Act 2 · Other people’s mistakes
An NFT marketplace that paid rewards for trading volume · 2022
94.5–98% of the volume turned out to be fake: people sold to themselves to collect rewards. For comparison: OpenSea, with no rewards, had 2.4% fake volume. After the rewards were partly cut, fake volume only fell to 69.5%—making an attack a bit more expensive doesn’t kill it.
Why it’s different here: LooksRare paid rewards in new, freshly minted tokens—the printing press paid for the fake volume. Our reward is the seller’s own money: inflating your own reward makes no sense. And the bar this grave sets is written into our requirements: the honest path must be strictly cheaper than cheating on every day the lot runs, with no “almost.”
An exchange for betting on NFT collection prices falling · shut down, winnings wiped out
Bets against illiquid NFTs were paid out of a virtual pool with no real money behind it; in a bear market, payouts exceeded what had been paid in, and the exchange went into debt and shut down.
Why it’s different here: a doubter here doesn’t bet on prices falling—they collect drops from a deposit paid in up front. Payouts can’t exceed what was paid in—the math is built that way.
Prediction markets with ultra-short deadlines · 2026
When an outcome can be moved, people milk it as long as the attack costs less than the prize: 821 wallets extracted $8.2M in two months, and 93% of the losses fell on retail players. With a 15-minute horizon, manipulation almost disappeared.
The lesson: a problem like that is fixed by tuning the design (a minimum length for the deadline), not by moderation. That directly confirms our approach—“send the cheater the bill through the settings, not through a court”—and it’s a reason not to allow ultra-short lots.
Act 3 · Checking ourselves
The most self-critical section of all. We had planned to use these stories ourselves—they didn’t survive the checks.
The Wörgl legend refuted 0–3
Why we wanted this story. At the heart of our mechanism is money that loses value over time. The first question any critic asks: “has money that loses value if you sit on it ever worked anywhere?” The crypto crowd’s standard answer is the story of Wörgl, and we were going to answer the same way.
The legend. Wörgl, a small Austrian town, 1932, the Great Depression. The town hall issues local money that loses 1% a month if you don’t spend it—holding it doesn’t pay, spending it does. The legend goes on: the money circulated 20 times faster than normal money, unemployment fell, the town flourished—until the central bank got scared and banned the experiment.
The check. The famous numbers (“circulation ×20” and the like) aren’t backed by primary sources—three AI reviewers independently found nothing to support them; the story survives on retellings of retellings. It won’t appear in the project’s story.
What survived confirmed 3–0—a fact that’s duller and more useful than the legend. In the US in the 1930s, dozens of towns issued similar money, and their fate came down to one detail. Where the fee was charged for the time you held the money (“once a week, stick a stamp on the bill”), the schemes worked. Where it was charged on every transaction, people instantly saw it as a tax on activity and simply stopped using the money: those schemes failed. Hence an iron rule of our mechanism: the deposit drips only by the calendar, never in response to an action.
“Planted bids earn the seller a 16–44% premium” refuted 0–3
A number from real auctions that circulates in the literature on planted bids didn’t survive a check of its source. We don’t cite it in any form.
Three more killed claims
“Token-curated registries (TCRs) died because the people listed in them didn’t need the token” (a folk explanation of their death—not documented); “the US regulator, the CFTC, came out in 2026 against markets where the outcome can be nudged” (no such document exists); “12% fake accounts in Binance’s BAB token giveaway” (the number doesn’t check out). None of this is used in our reasoning anymore.
Act 4 · The theorem
The central question for any mechanism with a seller inside it: can you fully ban the fake sale—where the seller (alone or through planted helpers) fakes success to take other people’s money? Auction theory answers it in a peer-reviewed 2025 paper:³ there are three desirable properties, and a mechanism can have only two of them. A choice where you can only have two of three is called a trilemma.
One action: place your stake and wait for the deadline. No multi-round bidding and no strategic maneuvering.
Guaranteed for fake sales: buying from yourself always costs more than giving up. Planting stakes “for” yourself isn’t fully ruled out, but it’s capped: a gap of at most a few percent of the price (in the demo, up to 3.6%, Act 6). The paper calls this “shill-proofness” (a shill is a planted participant).
A guarantee that it’s mathematically best for every participant to bet exactly in line with what they believe (the technical term is strategy-proof: “gaming it is pointless”). Nice, but one of the three has to go—and this is the one we give up.
Take an auction where the rules are simple (everyone places one bid) and telling the truth pays best. That only happens when your price depends on other people’s bids, not on yours (the classic case: the winner pays the second-highest bid, so lowballing your own is pointless). But since other people’s bids set the price, the seller can quietly slip in a planted bid of their own just below the real top bid—and pump up the winner’s price at no risk. There goes the protection against fakes.
Want the protection back? There are two ways, and both break something. Make the winner pay their own price, and the seller’s planted bid no longer affects anything— but now everyone is better off bidding low: telling the truth no longer pays best. Or run multi-round bidding where a fake gets exposed along the way—and then it’s goodbye, simplicity. Whichever way you turn, exactly one of the three falls. This isn’t an observation, it’s a proven theorem: you can’t get around it, you can only choose what to pay with.
Before this theorem, our choice looked like a compromise. After it, it’s a forced move: nobody gets all three properties at once; the only question is who gives something up knowingly and who does it without noticing.
A caveat (our checks were specifically hunting for overstatements): the theorem does not say “you can’t ban fakes, so charge for them.” A price instead of a ban is our own conclusion, drawn from a different fact: on a blockchain you can’t verify identity, so a ban like “the seller may not bet” can’t be enforced in principle—only a price can. We keep what the theorem proves separate from what is our own engineering choice.
The same paper separates weak protection (cheating doesn’t pay “on average”) from strong protection (cheating doesn’t pay even if the cheater knows everything about everyone). Our promise “a fake sale never pays off” claims the strong version. And since all stakes and pots on a blockchain are public, for us, a cheater who knows everything isn’t the worst case—it’s the normal one. That’s why the simulation (Act 7) was run against exactly that kind of attacker: one who sees everything.
The economics literature on planted bids gives us one argument and rules out another confirmed 3–0:⁴
³ Komo, Kominers, Roughgarden, “Shill-Proof Auctions,” EC’25 conference. ⁴ Lamy, 2009; Chakraborty & Kosmopoulou, 2004; and lab tests from 2007–2022.
Act 5 · A precedent
Our design, “influencing the outcome is legal, but only in your own favor,” sounds exotic—until you look at the rules of horse racing primary sources, 3–0. In British and Australian racing:⁵ owners, trainers and stable staff are allowed to bet on their own horse and forbidden to bet against it. The rule has been around for twenty-odd years, and you don’t need any theory to see what abuse it guards against.
But the precedent has two levels, and the second matters more than the first: the jockey—the one person who directly makes the outcome happen—is banned from any bets, either way, on any race in the world, including bets through stand-ins (the Australian rule says outright that “asking a friend to bet” counts the same as betting yourself). Translated into our roles: supporter-promoters are the “owners,” and a one-way permission fits them exactly; but given how much control the seller has, the seller is the “jockey.” The racing rules’ wording, “forbidden, including through stand-ins,” is a ready-made template for our catalog of attacks.
One important point: we borrow the model from horse racing, not the ban. The model answers two questions: who is most responsible for the outcome (the seller, like the jockey), and how to describe getting around the rule through stand-ins (“through a friend is the same as yourself”). But we have no bans by role: on a blockchain you can’t check whose wallet is whose, so the seller can technically bet both “for” and “against” themselves. Instead of a ban, there’s a price: every such move is worked out in the catalog of attacks (Act 6); it usually costs more than the honest path, and where a gap remains, we say so plainly.
A contrast, for completeness confirmed 2–1: esports⁶ went the other way: it bans players from betting on their own games at all and punishes violations with disqualifications of varying length, depending on which way the bet went. In other words, where influence can’t be priced in money, all that’s left is a court and a schedule of penalties. We replace the schedule of penalties with a price built into the design—that is exactly what sets our approach apart.
⁵ Britain’s Jockey Club, a 2003 rule; Australian Rules of Racing, rules AR 115 and AR 237. ⁶ The ESIC Anti-Corruption Code (esports).
Act 6 · The price list
The principle behind the whole design: cheating must cost more than the honest path, and the attacker must pay for the attempt themselves. There are nine known ways to cheat: eight have been calculated, one is open. Below is the short version; detailed answers with numbers are in the “Cheaters” section of the FAQ, and you can check for yourself in the demo lot. Percentages are a share of the lot’s price; ranges cover the different kinds of items.
| Attack and status | Who pays, and how |
|---|---|
| 1. Fake sale Buy from yourself through a second wallet to get the deposit back. closed by the math | The platform fee, the part of what dripped that never comes back, and the supporter reward—which is always paid. Costs 2.9–4.7% of the price more than giving up—even on the day that suits the cheater best |
| 2. Planted supporters (shills) Back yourself from other wallets to win the reward back. residual risk, capped | Planted stakes pay the drop and the fee, just like real ones—so the gain is capped at a few percent of the price (in the demo, on an NFT lot, no more than about 3.6%). The fix: checking whether the money came from the seller’s wallets, plus reputation |
| 3. Fake sale + planted supporters Both tricks at once. closed by the math | To keep the combo from paying off, less of what dripped comes back on a sale (for expensive items and real estate, 55%). Costs 0.8–2.1% of the price more than giving up |
| 4. List-and-cancel Keep listing and pulling lots while betting against yourself, pocketing the drops of honest supporters you lured in. residual risk | With no one lured in, the cheater pays 9 times the harm done. If real supporters came to the lot, the cheater makes about 0.4–1% of the price per cycle. The fix: the history of cancellations is visible to everyone, plus a small fee for listing a lot (designed, but not switched on yet) |
| 5. Self-buy ladder Split a fake sale into many small resales. closed by the math | Each piece carries its share of all the costs of a fake. Costs 9.0–14.7% of the price more than giving up (demo lot) |
| 6. Paid panic A doubter whips up fear at the right moment so the lot doesn’t sell. residual risk | What they earn isn’t theirs until the deadline, and the pot is fed by supporters’ drops—scare everyone off and they cut off their own income. Campaigning “against” is legal; the line is crossed by lying about your own position or secretly funding someone else’s campaign |
| 7. Sniper Jump in as a supporter at the last moment to grab a share of the reward without risk. closed | The reward is split by how much you put in × how long you held it, and a stake has a minimum holding time—a late stake weighs almost nothing |
| 8. Betting on your own failure Secretly bet against yourself and let the lot fail. residual risk | The fat pot of a doomed lot attracts real doubters, and their money dilutes the cheater’s share: the cheater loses 14–15% of the price versus 0.75% for giving up (demo lot). It’s closed only if enough outside doubters come to the lot; in an empty market, it isn’t |
| 9. Drop outsourcing A service pays the drop for supporters and splits the reward with them. open | Not calculated. Nobody in the industry has a ready-made fix |
A note on row nine, the only open one. The drop is our entry filter: a supporter shows they believe by risking a small share of their stake. The history of crypto projects shows that costs like this end up getting shifted onto a middleman—the belief filter turns into the middleman’s fee.⁷ Nobody has a ready-made fix; the template for wording a ban “including through stand-ins” comes from the racing rules (Act 5).
A word on activity points: they are awarded in proportion to the fees paid—anyone who churns fake volume for points pays for those points themselves. There is no token, and there won’t be one until the mechanism is proven.
⁷ The Curve/Convex case: voting power that was handed out by how much you put in × how long you held it got bought up by middlemen.
Act 7 · The stress test
After the research, the mechanism was run through the numbers: a simulator engine that reproduces every money rule exactly (the balance matches to the 17th decimal place), with zero transaction costs and a cheater who knows everything—they see every stake and pot, as a real attacker on a blockchain would. Seven numerical experiments; eight of the nine attacks in the catalog were calculated. Three results are worth sharing:
Act 8 · The limits
The research is finished. Two questions remain open:
Beyond that, there are some observations and the limits of the first version:
The full list with answers is under the FAQ question “What have you NOT proven yet?”
Appendix