The question

The US Treasury is the government’s finance department. It borrows what the government spends by selling bonds, and the interest a bond pays is its yield. In August 2026 the Treasury doubled the size of its buybacks of long-term bonds, the operations where it buys back bonds it sold earlier. Many people suspect the real aim is to push long-term interest rates down.

One piece of vocabulary before anything else. A bond’s price and its yield move in opposite directions, so holding a yield down means holding a price up. Push rates down, defend a price, manage prices and target a yield all mean the same thing on this page.

On 4 November it publishes its quarterly refunding statement, the announcement of how much debt it will sell over the next three months and why. If it decides to sell fewer long-term bonds, it must give a reason.

The Fed is the US central bank, and it sets short-term interest rates. Its own officials have said in public that once the Fed has set that rate, the rest of the government bond market is the Treasury’s business, including the mix of short-dated and long-dated debt. Nobody has said whether the Treasury may use that power to push long-term rates down. So the question we built the game around: who, if anyone, forces the Treasury to admit that aim before 4 November, and what does the reason it gives on that day cost it?

Our answer, after 45 games: nobody forces the Treasury by argument. Speeches from the people who most wanted the question asked moved the public debate part of the way and never far enough. The question came into the open only after the Treasury itself spent enough money that the market noticed. And the reason it then gave on 4 November cost it more credibility than the decision to sell fewer bonds did.

How the simulation works

Seven roles, each played by a computer program that follows a written policy, not a person and not a chatbot. The game runs five turns of one to two weeks each, 9 September to 4 November. Each turn every role tries a move, a referee program sets the odds, and a dice roll decides whether it works. A game ends with the Treasury’s statement on 4 November and the market’s response to it.

We gave the Treasury three plans for 4 November and played each plan 15 times, 45 games in all.

Each plan’s 15 games split two ways. Three ran on the same dice as the other plans, so the plans compare move for move. We call those the matched games. Twelve ran on fresh dice, the fresh-dice games. Only plan 3 includes the cash draw, so 30 of the 45 games have none.

Every game then ran a second time with one rule switched off: the rule that lets a role react to the reason another role gives, rather than only to the money it moves. That second run is the control. No ending, no Fed vote and no dissent count, the published number of Fed officials voting against the majority, changed with the rule off. One thing did, the Fed chair’s late move, and the page comes back to it. Everything up to 3 September 2026 is real and sourced in the deep mode. Everything after that date is the game. None of it is a forecast.

Three scales we wrote track the story. All run 0 to 100, and all are bookkeeping rather than measurements.

The five turns in five lines

Teal dates hold a real event on the calendar. Grey dates hold moves the roles made in the game.

The Fed and the Treasury agree, and the agreement leaves one thing unsaid

Everything in this section is on the public record as at 3 September 2026. The commentary describes a fight between an independent central bank and an activist Treasury. The record shows agreement. A former regional Fed president says it plainly: once the Fed has set the overnight rate, everything else about the Treasury market belongs to the Treasury. Stephen Miran, a Fed governor, defends a smaller Fed balance sheet, the stock of bonds the Fed owns, because it keeps the mix of long and short government debt a matter for the Treasury.

That agreement settles who owns the mix. It says nothing about whether the Treasury may use the mix to push long-term rates down. Nobody argues that, because both institutions have written down the same answer to the only question anyone has asked.

The Treasury operates in exactly that gap. Its own refunding statement says any change to bond sales will be judged on trends in structural demand. Scott Bessent, the Treasury Secretary, told a Treasury market conference in November 2025 that his job is to be the nation’s top bond salesman and that Treasury yields are a strong barometer of his success. The day after the August increase, CNBC reported, he said the level of yields did not factor into that decision.

Both statements are on the record, and the game is about which one the 4 November statement is written under. The reading that there is no clash holds only while the Fed’s own bond holdings stay flat, meaning the Fed releases no long-dated debt onto the market itself and leaves the mix to the Treasury. They were flat in early September, and a Fed task force is studying their mix.

Speeches moved the debate a little and never enough

In the two plans where the Treasury keeps its wording or cites structural demand, the question of its real aim never came into the open before 4 November: in none of those 30 games. The Fed presidents made their speech in almost every game, on the same turns each time, and it changed nothing downstream. The endings, the count of Fed officials voting against the majority, and the reason given on 4 November came out the same whether the speech landed or missed.

Look at who does the arguing. The only players willing to say in public that the Treasury is managing prices are the three who want a rate rise, so the question arrives already sounding like a bid for something else. That is true of the real Fed as well as the game.

The Treasury’s own spending opened the question

In the third plan the Treasury drew $40bn from its cash account at the Fed, an amount we chose, to make its buybacks bigger than the $4bn-per-operation ceiling it had published. It changed not one word of what it said. The draw shows up in the Treasury’s own daily cash statement, and the size of each buyback is published with its result, so the market saw both.

In 9 of those 15 games the question then came into the open within a turn or two. Never after a speech. Always after the extra spending, and usually with several players reacting at once: a routine buyback read as a defence of a price, a Fed vote that went badly, and hedge funds and dealers betting the Treasury could not hold the level.

The exposure index is our own scale, so here is a move the dice decided instead. Hedge funds and dealers bet against the long-term yield as a defended price in 11 of the 15 games in the third plan and in none of the other 30, always at turn 3, always after the cash.

We then played the three plans once more, one game per plan, with a language model, the kind of AI behind a chatbot, in every seat, meaning each of the seven roles, and a stronger model as referee. The seven rule-following programs cannot argue, change their minds or find a move nobody wrote down. This replay checks whether players that can do all three open the question by the same route, and they did.

Told to keep its wording or to cite structural demand, the Treasury drew on its cash account anyway and funded bigger operations while saying nothing new. In all three games hedge funds and dealers bet against the level at turn 3 and the question came into the open. The three plans differed only in what the Treasury said on 4 November.

The reason cost more than the cut

The Treasury’s credibility starts at 70 on our scale. In the three matched games, keeping the wording cost five to ten points. Cutting bond sales and citing structural demand cost fourteen to twenty-five. Cutting by the same amount and admitting a target for the yield cost thirty-seven to forty-two.

“Structural demand” was believed in 11 of the 12 fresh-dice games. It has no middle setting: the market either accepts it or names the real aim for the Treasury. In the twelfth game the market decided the real aim was the yield anyway, and credibility landed at 50, against 54 where the reason was believed. In one of the three matched games the same thing happened and credibility ended at 45. Figure 4 is the same finding as a picture, and it is the one line this page asks you to keep.

A reason your own stated policy already permits is a good reason right up until the day it is not believed, and there is no partial credit.

There is a second effect. The Fed chair’s best move, saying in public that long-term rates belong to the market and the debt mix to the Treasury, only becomes available once the question is open. In this game that happens only after the Treasury has spent the money. In the games where he said it, he ended six points of authority above the same game with the reason rule switched off.

What it means

The reusable idea is not about bonds. When an institution is scored on a number it only partly controls, the fight everyone expects is about the decision, and the fight that happens is about the justification. Judge the justification, not the decision.

Three questions follow, and they work in any organisation managing a number under scrutiny.

What this does not prove

The control run, the second run of every game with the reason rule off, is the most important check in the report. With the rule off, every ending, every dissent count, both Fed votes and every path of the 30-year yield in the matched games came out the same. So we can tell you who forces the question and what each plan costs. We cannot tell you that the reason changes what the Treasury or the Fed decides inside these eight weeks, because in this game the reason is given last.

Nor is any of this a forecast. The exposure index is ours, start value and threshold both. The odds on a president’s speech and on the Fed Board’s silence are ours, and they set how fast speeches alone can work. The $40bn is our number, and the cash balance it comes from is press reporting rather than the Treasury’s own daily statement.

We chose the small drift in the 30-year yield ourselves, which is why nothing here is scored on the yield. The third plan did not lower that yield in a single matched game. And the largest player in the story is not on the board: the White House enters only through outside events we scripted, and made no move in the eight weeks, because on the public record it has made none. If the White House acts inside the window, we re-run the game.

The evidence pack behind the report lists 57 sourced claims, and the deep mode marks every sentence that rests on one. Where a sentence leans on press reporting rather than an institution’s own release, it says so.

What to watch

Nine real-world signals, and where each is published.

The Treasury’s statement on 4 November will contain one reason, and the reason is the thing to read.