On 4 September 2026 the Bank of Korea published an issue note, “Links Between Dollar Stablecoins and Foreign Exchange Markets”. Its researchers, Jihyun Kim and Sangheum Cho of the bank’s international finance team, asked one question: what happens to a country’s currency when Binance switches on a direct pair between that currency and a dollar stablecoin.
The answer is that the currency goes down. Not dramatically. Measurably. In Brazil, where you can buy USDT or USDC on Binance directly with reais, the demand translated into real dollar purchases as market makers balanced their books, and the real-dollar rate moved 0.12 per cent against the real. Across twelve currencies with pairing dates running 2019 to 2025, the local stablecoin premium narrowed by 0.33 to 0.38 percentage points for the euro and the Turkish lira once direct trading opened.
Everyone filed this as a stablecoin story, or a central bank story. It is neither. It is the fifty-first unasked half of fairness, and the first one in this series where the person who pays never played.
TL;DR
- A Bank of Korea study found that enabling direct fiat-to-stablecoin pairs on Binance correlates with local currency depreciation, measured at 0.12 per cent for the Brazilian real, with stablecoin premiums narrowing 0.33 to 0.38 points in the euro and lira.
- Every fairness proof in crypto gaming, ours included, proves a ratio. Chainlink VRF proves a selection. A coinflip contract proves a multiple. None of them has a unit attached.
- The unit was chosen by the platform, before you arrived, for liquidity reasons, and it has never once appeared in a fairness disclosure.
- The new shape here: fifty instalments have been about what a proof fails to cover for the player. This is about a cost landing on somebody who never opened the app.
- Honest limit: Satoshie settles in a dollar-denominated unit on Base. We are inside the aggregate this study measured, not outside it.
What the study actually did
The method is the good part. You cannot observe crypto demand directly, so the authors used Binance pair-listing dates as a natural experiment and Google searches for “bitcoin” as a demand proxy. A one-standard-deviation rise in Brazilian bitcoin searches came with a 0.118 per cent depreciation of the real and a 0.109 point rise in the local stablecoin premium: same signal, two prices, moving together exactly as the mechanism predicts.
The mechanism is mundane, which is why it is credible. Somebody in São Paulo wants dollars on-chain, hands reais to a market maker and receives USDT. The market maker is now long reais and short dollars, and does what every market maker on earth does: goes to the FX market and flattens the position. Millions of individually sensible decisions become a standing bid for dollars against the real, and the rate moves.
Korea itself barely registers, because Binance has no direct won pair and the domestic market is retail-dominated. The bank’s conclusion is forward-looking rather than alarmed: widen access to domestic exchanges and the link between the stablecoin market and the FX market strengthens, so digital asset reform should run alongside internationalising the won. That is a careful sentence from an institution with an interest in won stablecoins existing, and I will treat it as one.
The fifty-first unasked half: the unit nobody chose
Here is the thing this series exists to say out loud.
Every fairness guarantee in crypto gaming is a ratio, and a ratio is silent about the unit.
A Chainlink VRF proof says one output is the unique VRF output for one key and one seed. A coinflip contract says the escrow doubles or it does not. A raffle contract says one ticket in n. Every one of those statements is true in dollars, true in won, true in beads, and true in a token invented last Tuesday. They are scale-invariant, which is exactly why a stranger with no context can check them.
You are not scale-invariant. You are paid in something and you owe rent in something, and the numerator of your win is made of a specific substance whose relationship to those things is set in a market you do not participate in. The proof covers the multiple. It has never covered the material.
And the unit is a decision. Somebody picked it, before you arrived, on grounds of liquidity, integrations and which asset the bridges support. Good engineering grounds, and nothing to do with you. I have never seen a “provably fair” page anywhere in this industry name the unit as a term of the deal, because in the platform’s mental model it is plumbing. For the player it is the denominator of every outcome.
Why this instalment is a different shape
Fifty of these have been about a gap between what a proof covers and what a player assumed. The uncovered party was always the player. Read the Brazilian number again and notice who absorbed it.
A person in Brazil who has never bought a token, is paid in reais and buys imported goods absorbed a fraction of that 0.12 per cent. There was no terms of service for them to not read, no opt-out, no support ticket, no pick-a-better-operator advice to give them. Nobody defrauded them and no rule was broken. Every transaction in the chain was legitimate, voluntary and, in the on-chain leg, verifiable to the last wei. A very large number of individually provable transactions produced a price move on a third party who was never a counterparty to any of them.
That is not a failure of verification. Verification worked perfectly. It is the boundary of what verification is for: a proof is a statement about a transaction, it has no vocabulary for the sum of transactions, and the sum is where currencies live.
There is a sting worth sitting with. The Bank of Korea could measure this only because our side of the market is legible: listing dates public, on-chain flows public, premiums computable by anyone. No equivalent study exists for the private FX flows the same market makers run, because those rows are nobody’s to read. Transparency did its job here. It just did it for a central bank, about us, rather than for us, about a platform.
Where Satoshie actually stands
The rule in this series is that we go first, so: we settle in a dollar-denominated unit on Base. We are inside the aggregate this study measured, at the size of a rounding error, and a rounding error is still a term of the deal. We did not choose the dollar on principle. We chose it because that is where the liquidity is, which is precisely the individually rational decision the paper is about.
Three honest limits, stated plainly:
- We cannot denominate in your currency. There is no deep on-chain real, won or naira market to escrow a prize pot in, and “we would if we could” is not an engineering statement.
- The issuer can freeze the unit. A trust assumption we inherit, cannot verify our way out of, and which sits upstream of every contract we deploy. Named here for the same reason we named the Base sequencer.
- Our claims stop at the ratio, on purpose. Ticket price is a constant in the deployed contract. Odds are a function of entry count readable before a ticket sells. The VRF proof is verified on-chain by the coordinator before the payout callback runs. Escrow, resolution and payout happen in one transaction, with no admin key reaching a live draw. Read that list: ratios and constants, and not one claim about what a dollar will be worth on Thursday. That is the honest scope, and the day we imply otherwise you should stop reading.
The thirty-eighth instalment covered par: a peg is a promise made to wholesale redeemers and inherited by everybody else on trust. The forty-sixth covered entry price on an FX desk, where the chain proves what you own and says nothing about what you paid. This is the third corner: the chain proves the ratio, the peg is somebody else’s promise, and the price of the unit against your actual life has never had a line item in any fairness disclosure ever written.
Three questions
What unit is my stake escrowed in, and who chose it? If the answer is a platform token, the exchange rate between your stake and money is a market that the platform makes, which is a house edge wearing a ticker.
Is there a conversion at deposit and another at withdrawal, and at whose rate? Two conversions at an unpublished rate is a cost with no line item, which is the worst kind, because there is nothing to disclose and nothing to compare.
If the unit moved ten per cent against my currency between escrow and payout, would the fairness page mention it? It would not, and that is acceptable, so long as you knew before you played that the proof stops at the ratio. The problem was never the scope. It is that nobody publishes it.
Fifty-one
The Brazilian number is 0.12 per cent, which is small, and I am not going to inflate it into a crisis the authors did not claim. What makes it worth an instalment is where it landed. Fifty of these have argued that a proof covers less than the badge implies, and the remedy was always in the reader’s hands: check the statement, run the verifier, ask before you stake. This is the first one with no remedy in it, because the person carrying the cost is not in the room and never was.
We are not going to solve that, and neither is anyone else in on-chain gaming. What we can do is stop pretending the unit is plumbing. It is the denominator of every outcome we settle, somebody chose it, and it was not you.
Photo by noodle kimm on Unsplash
