On 17 September 2026, Upbit listed JPYC, a yen-backed stablecoin, and did something exchanges rarely bother to do. It published the right answer in advance. The listing notice cited an external market price of 8.81 won per token. That number was correct, it was public, and anybody could read it before placing an order.
Trading opened at 18:00 KST. The first execution printed at 12 won, already 36% above the figure Upbit itself had published on the same page. Within minutes it was 27.2 won. Between 19:20 and 19:25 it touched 37.60 won, roughly 4.27 times the price the venue had disclosed about ninety minutes earlier.
Nothing was hidden. Nothing broke. The issuer honoured redemption at one yen the entire time. And the price on a regulated exchange in a developed market was wrong by a factor of four for eight hours.
TL;DR
- Upbit listed JPYC on 17 September 2026 and published a fair value of 8.81 won in the listing notice. The first trade was 12 won and the high was 37.60 won, about 4.27x.
- The peg did not fail. JPYC remained redeemable at one yen throughout, which makes this a different failure from the par-redeemability problem the BIS writes about.
- South Korea’s Virtual Asset User Protection Act has no market-making exemption from its manipulation provisions, so the people whose job is to close that gap could not legally do it.
- Where correction was permissionless, on DEXes, it was still bounded by inventory: executed trades above 1,000,000 JPYC only reached ¥1.31 on Ethereum, while the headline ¥4 prints occurred in trades under 1,000 tokens.
- A published price is a number plus a route. A Satoshie prize is not a price at all, because it is escrowed before the draw and paid in the same transaction that resolves it.
The peg never broke, and that is the whole point
It would be easy to file this as another stablecoin wobble. It is not one. We have written before about par redeemability, the question of whether an issuer will actually give you a unit of the reference asset when you ask. That is a question about solvency and willingness. JPYC answered it correctly on the day. Seven mints totalling 12.48 billion JPYC were executed between 21:37 on the 17th and 03:23 on the 18th. The issuer did its job.
The buyer on Upbit paying 37.60 won was not a victim of a broken promise. They were a victim of distance. The promise was real and it was being honoured somewhere they could not stand.
That distinction is the unasked half here, and it is the seventy-seventh in this series. Redeemability is not a price. It is a price plus a route. Every fairness claim built on “you can always redeem at par” quietly assumes a route exists, is open, and is staffed by somebody with capital at the exact moment you need them.
Three things have to be true, and disclosure covers one
For a wrong price to correct itself, three separate conditions have to hold at once.
Somebody has to know. This is the only one Upbit satisfied, and it satisfied it completely. The true value was on the listing page. Korean traders were not operating on bad information. They were operating against a wall.
Somebody has to be allowed. This is where Korea specifically failed. The Virtual Asset User Protection Act contains no exemption carving market-making activity out of its market-manipulation provisions. Continuously quoting two-sided prices looks, to that statute, like manipulation. So the one category of participant whose entire function is to stand between a panicked bid and a fair value was legally absent. Yoo Young-joon, director of digital finance policy at the FSC, has now said the regulator “will also review the need to introduce systems such as market-making activities to increase the efficiency and stability of the digital asset landscape”, noting that criticism over user losses has expanded demands for discipline. A 2024 Seoul Law Review paper had already identified exactly this gap as the source of Korea’s liquidity problems and the Kimchi premium. The warning was two years old and correct.
Somebody has to have the stuff. This is the condition almost nobody names, and it is the one that survives even when the first two are satisfied. Correction is not an act of will. It is an act of inventory.
Where correction was legal, it still did not work
Here is the part the wire copy missed. While Upbit printed 37.60 won, JPYC was also trading on Ethereum, Polygon and Avalanche, where nobody needs a licence to quote a price. An analysis of 55,293 executed swaps across the event window shows what the correction actually looked like when permission was free.
Peak fifteen-minute VWAPs were ¥3.14 on Ethereum at 18:30, ¥2.92 on Polygon at 19:15 and ¥1.39 on Avalanche at 19:45. So the dislocation reached the permissionless venues too. But the price was a function of size. Trades above 100,000 JPYC hit a ceiling of ¥3.32 on Ethereum and ¥3.51 on Polygon. Trades of 1,000,000 JPYC or more only reached ¥1.31 on Ethereum and ¥2.98 on Polygon. The celebrated ¥4 to ¥6 prints happened almost exclusively in trades under 1,000 tokens.
Read that again. The four-times-peg price existed only in sizes too small for anybody to profitably arbitrage. Mint at ¥1 and sell at ¥4 did not work at size, because at size there was no ¥4.
And the route into Upbit was worse than thin. The exchange supported deposits and withdrawals on Ethereum only, covering roughly 7% of circulating JPYC. Deposits on that path opened at 19:23, which is to say during the peak rather than before it. Issuance was capped at ¥1 million per request with delays between consecutive requests. And a participant running the trade needed both a Korean bank account and a Japanese My Number card. Not a capital requirement. A paperwork requirement, held by approximately nobody.
The supply eventually arrived. Of the 12.48 billion JPYC minted, 9.9 billion landed after 02:00 on the 18th, by which point the price had already returned to roughly ¥1. The inventory showed up after the emergency it was needed for. That is not a market failing to clear. That is a market clearing on a schedule set by a KYC queue.
What this maps to in gaming
Every number a gaming platform publishes is a quoted price in this exact sense, and almost none of them disclose the route.
A jackpot figure is a price. It is a claim about what somebody will hand over, at some size, under conditions not written on the page. A withdrawal limit is a route disclosure in disguise, and usually the only honest one on the site. An “instant payout” badge is a statement about a happy path, quoted in the smallest size, exactly like a ¥4 print in a 900-token trade.
The question that separates a real number from a printed one is always the same: at what size, and through which door? A platform that pays a €50 win instantly and a €50,000 win after a manual review has not published one payout policy. It has published two, and told you about the small one.
This is the same shape as the entry-price problem on a dealer desk, but inverted. There, no price was printed at all. Here the price was printed, publicly, correctly, by the venue, in advance, and it changed nothing. Transparency was total and irrelevant. It turns out you can disclose your way to an informed counterparty and still leave them with no counterparty.
What Satoshie actually claims here, and it is narrow
A Satoshie raffle payout is not a price. It is not quoted, not discovered, and does not depend on anyone being present, permitted and inventoried at the moment you win.
Stakes are escrowed on entry, so the prize pot is a balance sitting in the contract before the draw rather than a figure on a marketing page. Odds are a function of entry count, readable in deployed code before you stake. Randomness is requested against a request ID and the proof is verified on-chain by the Chainlink coordinator before the payout callback is permitted to run, using code we did not write and a key we do not hold. Resolution and payout land in the same transaction. There is no admin key over a draw in flight, no owner function over odds or outcomes, no pause.
The structural point is this: there is no moment in that sequence where a market maker has to exist. The number that settles is not a price somebody has to be willing to meet. It is a balance that already exists, moving under a rule written before you arrived. A route cannot be closed if no route is required.
Honest limits, three of them
The unit we escrow in is somebody else’s liquidity problem. We settle in a dollar-denominated unit on Base. If that unit dislocates on the venue where you eventually convert it, our proof is silent, exactly as we argued about denomination back in the fifty-first instalment. The draw is scale-invariant. The exit is not.
Getting in and out is a market, and we do not make it. Funding a wallet and selling what you won both happen on venues with order books we do not run and cannot staff. Our no-counterparty claim covers the draw. It does not cover the round trip, and any platform telling you otherwise is selling you somebody else’s liquidity as its own feature.
Escrow is a ceiling as well as a guarantee. A prize cannot exceed what has been staked into it. That is a genuine limit on us. It is also the tell: an operator advertising a jackpot larger than any balance you can inspect is quoting a price, not showing you a balance. We would rather name our ceiling than quote a number and hope the route holds.
Three questions worth asking anywhere
Is the prize a balance right now, or a number on a page? Ask for the contract address and read the balance. If the answer involves a support ticket, it was a price.
What happens at the largest size you are allowed to play? Not the advertised payout. The one that has actually settled. Small-size performance tells you as much about a platform as a ¥4 print in a 900-token trade told anyone about JPYC.
When you go to sell what you won, who is permitted to take the other side? If the answer is a single venue in a single jurisdiction, you are one statute away from 37.60 won.
The number was never the problem
Korea is now drafting a consolidated Digital Asset Basic Act covering stablecoins, exchanges, disclosures and internal controls, with the market-making exemption reportedly under review. That is the right fix and it is roughly two years late, which is the normal speed for this sort of thing.
But the lesson generalises past Korea. An integrity rule written against manipulation removed the participants who make prices honest, and the result was the exact harm the rule existed to prevent. Every control has a cost side, and the cost side is never in the rule. We are not exempt from that, which is why the controls worth having are the ones nobody can switch off, including us.
Upbit did the transparent thing. It published the true price and let people trade. Four times wrong for eight hours, with the correct answer visible on the screen the entire time. If disclosure were fairness, that market would have been fair.
📷 Photo by T (@tanyabarrow) on Unsplash
