At 8:27 on Friday morning the market gave a September rate hike a 59.5% chance. At 8:31 it gave it 79.5%. Nothing happened in between except the Bureau of Labor Statistics publishing one decimal place.

Core CPI printed +0.3% month over month. Consensus was +0.2%. That is the entire event.

Here is what makes it worth your morning. The Wall Street Journal surveys economists before every CPI release, and for August core, all seventeen of them were wrong in the same direction. The median forecast was 0.22%. The full range ran 0.16% to 0.24%. Every single estimate rounded to 0.2%. Not one person in that survey left room for the number that actually arrived. The day before the release, Kalshi was pricing a core print above 0.2% at a 28% chance.

The repricing, minute by minute, from Polymarket's own trade history:

8:27 a.m. ET, 59.5%

8:30 a.m., release, 71.0%

8:31 a.m., 79.5%

9:09 a.m., 82.5%, the peak

early Monday, 79.5%

Twenty points in four minutes, and three more over the next thirty-eight.

A hike on Wednesday would be the first since July 2023. The Federal Reserve has not been able to say one word about any of this. Its blackout began at 12:00 a.m. Eastern on Saturday 5 September and runs until 11:59 p.m. on Thursday 17 September, which is its own published rule. Across that silence the hike leg has gone from 50.5% to 79.5%. Twenty-nine points, with the committee under its own gag rule the whole way.

Keep one human detail in view. Kevin Warsh chairs this meeting. Jerome Powell is still on the Board as an ordinary governor and sits at that table on Tuesday, no longer running it, one vote among the rest.

And now the part we cannot tidy up.

Bitcoin rallied on the news. Hourly marks: $76,996 at 12:00 UTC, $77,961 at 13:00, $79,157 at 14:00. That is 2.8% in ninety minutes, on a release that made a rate hike substantially more likely. By 19:00 UTC it was $77,037, having given back every cent of it.

The day before, in the same 8:30 slot, it did the exact opposite. PPI printed hot on Thursday and bitcoin went from $77,831 to $76,869, down 1.2%. Hot producer prices sold it. Hot consumer prices bought it for ninety minutes and then unsold it.

We do not have a clean explanation for that and we are not going to build one. Two hot inflation prints, one day apart, same release time, opposite reactions, both fully retraced. Anyone telling you confidently which one was the signal is reading a chart backwards. What we can say is that the ninety minutes of buying was not a view about the Fed, because the thing being priced in the other market at the same moment was a hike.

Three metrics.

Metric 1, Fear and Greed: 57. Greed, per alternative.me, stamped 00:00 UTC on Monday. The index updates once daily, so that is still the live reading as you open this. A week ago it was 71. There is one sharp move in the series, 69 down to 56 on Friday's stamp, and it would be easy and wrong to file that as the CPI reaction: each stamp covers roughly the preceding twenty-four hours, so Friday's 56 is Thursday's PPI selloff. It bounced to 63 on Saturday, then 61 on Sunday and 57 on Monday. Greed, in a week when spot fell 2.9%.

Metric 2, ETF flows: September is now about +$307 million. Four consecutive negative sessions gave back $462.7 million: minus $46.6M on the 8th, minus $120.2M on the 9th, minus $282.6M on the 10th, minus $13.3M on the 11th. That cuts the month from the roughly $770 million we reported through 4 September down to about $307 million, a 60% haircut to the margin on our 7 October dated test. Worst day was the 10th, which is also the day PPI printed and the thirty year yield closed at its highest since the 2000s. (Sourcing, stated rather than assumed: Farside, which our test names, returns 403 to us from this network, so these are SoSoValue's prints. They match our own published Farside figures for 1 to 4 September to within $0.2 million.)

Metric 3, BTC dominance: 58.9%. CoinGecko's global measure, pulled at half past midnight Eastern on Monday, against a total crypto market cap of $2.64 trillion. Up from the 58.4% we published Wednesday, which is what dominance does when bitcoin falls and everything else falls harder. Strategy belongs in this metric too. Its 8 September filing, covering 31 August through 7 September, shows 845,050 BTC at an average cost of $75,412 on a $63.73 billion book, and states verbatim that the company "did not sell any shares under its at-the-market offering program and did not purchase or sell any bitcoin." At $77,560 that leaves it 2.8% above cost, roughly $1.8 billion unrealised, with the breakeven about 2.8% under spot. The line to read is the cash line. USD Cash fell from $1.61 billion on 31 August to $1.44 billion on 7 September, and the same filing says where it went: $176.3 million bought back 1,810,885 shares of STRC, its variable rate preferred, at an average of about $97 against a $100 stated value. A week in which Strategy bought no bitcoin was a week in which it bought its own paper at a discount. The next 8-K is due today.

The book is flat for the twelfth consecutive issue. The STACK below states what would change that in three different ways, because quoting only one of them would be a choice about how we look, and it carries a correction to our own record that we would rather you heard from us.

📣 THIS SPOT IS OPEN

Baseline Crypto reaches nearly 7,000 Bitcoin-focused readers every Monday and Wednesday morning, opening just under half of what we send. Direct sponsor placements are open, premium position, top of the issue, one per send.

Reply to this email for rates.

Credit is not calm about the hike. Credit has priced the hike, and priced it working.

Take the position seriously first, because it has been right for three years. High yield OAS is the most honest stress instrument in finance. It is marked continuously by people with real money at risk, and it has led equities into every serious credit event of the modern era. It is not a survey. Anyone who spent the last three years calling a credit event off macro plumbing lost money doing it. And right now it is flat: 2.72% on 11 August, 2.70% on 10 September. Two basis points tighter over thirty days.

Here is what happened to the price of money over those same thirty days.

The ten year real yield closed at 2.55% on 10 September, the highest since 28 November 2008, verified against the full FRED series rather than a snippet, and Treasury's own table has it at 2.60% for Friday, which FRED posts on Monday. The thirty year hit 5.37% on Thursday, a level it last closed at in the 2000s. The ten year nominal printed 4.95%, highest since October 2023. So the risk free rate reached levels most people currently working in high yield have never traded through, and the extra compensation for owning junk instead got smaller.

Now ask the incentive question, because it is the one that actually resolves this. What is the person setting a high yield spread paid to forecast?

Not real rates. Defaults. A spread is compensation for the probability a borrower fails to pay, and the dominant input to that probability over the next twelve to twenty-four months is nominal revenue. Which is the uncomfortable part. Headline inflation at 3.4% is nominal revenue. A levered borrower with fixed coupons and rising prices sees interest coverage improve, not deteriorate, for as long as it does not have to refinance. Real yields do their damage at the maturity wall, and for most of this market that wall is years out, not weeks. High yield is short real rates on a long lag and long nominal growth right now. Friday's print delivered the second and deferred the first. That is not blindness. That is a position doing exactly what it is paid to do.

And the bond market itself is demonstrably not confused, which is the detail that settles it.

Over two sessions the 2s30s curve flattened 13 basis points, from 85bp on the 9th to 72bp on the 11th. On CPI day the two year rose 7bp to 4.63% while the thirty year fell 2bp to 5.35%. And the ten year breakeven, which is the market's own inflation forecast with money behind it, fell from 2.40% to 2.36% on a hot inflation print.

Read that sequence back. Short rates up, long rates down, expected inflation down, on the day inflation surprised to the upside. There is one coherent translation and it is four words. He hikes, it works.

Which is also why both debasement trades lost money in a week headline inflation printed 3.4%. Bitcoin fell 2.9% over the seven days to Monday morning. Gold, via PAXG, fell 1.6% over the same window. Neither of those assets is priced off a CPI release. They are priced off the real yield, and the real yield just printed its highest level since November 2008. That is the mechanism, and it is the one people substitute a story about money printing for.

Now the hole in our own argument, stated rather than buried. FRED has not posted a high yield OAS value for 11 September. The latest print is the 10th, the day before CPI. So the single observation that would most directly test everything above, how credit actually closed on the day the forecast broke, is the one we do not have. It should post Monday. If spreads widened materially on Friday, the steelman gets stronger and we will say so on Wednesday rather than waiting to be asked.

Three numbers that define the next three weeks.

2.70%, the number doing three jobs at once. That is high yield OAS for 10 September, the latest FRED has posted. It is 30 basis points from the 3.00% level that puts us back on the board. It is 2 basis points tighter than thirty days ago, which is the wrong direction for us. And it is 10 basis points wider than the 2.60% low of 28 August, which is the right one. It is also the Branch B floor on our 7 October dated test, exactly. What to do with it: watch Monday's print specifically, because that is the first reading that contains CPI day, and it is the single observation that tests the FILTER above. One number, one day, and it either supports us or it does not.

$75,412, the level where every Strategy headline changes register. That is the average cost across 845,050 BTC, from the 8 September filing. Spot was $77,560 early Monday, so the company is 2.8% above water on a $63.73 billion position, roughly $1.8 billion unrealised, with the breakeven about 2.8% below where bitcoin trades this morning. A 2.8% move is what bitcoin did in ninety minutes on Friday. What to do with it: read the 8-K due today, and read the cash line rather than the coin line. Last week's filing showed $176.3 million of USD Cash going into STRC buybacks in a week with no bitcoin bought. If that repeats, the company is telling you, in its own capital allocation, that its preferred stock is the cheaper asset. Nobody at Strategy will ever say that sentence out loud, which is why the table is worth reading.

Plus 4.00%, the line our difficulty test crossed overnight. The retarget at block 967,680 is about 766 blocks out, projected for Saturday 19 September, and it grades a week today, on Monday 21 September, not in this issue. Average block time this epoch is about 578 seconds against a 600 second target, which is where the increase comes from. On Sunday evening our own computation off the block timestamps put the estimate at 3.86%, with two mempool mirrors at 3.83% and 3.92%, all just under the Branch A line. By half past midnight, after a run of fast blocks, mempool.space itself read plus 4.24%. That is the estimate crossing our line in the hours before we send, and it will cross back if the next few dozen blocks come slow. What to do with it: nothing, and that is the point. The estimate has been above 6% and below 2% inside a fortnight, so anyone treating a single reading as the answer has not watched it move. We grade on the figure at the retarget.

Forget analyst predictions. Polymarket is a real-money prediction market, where traders put actual dollars on outcomes. Scorecard first, then the board. Not financial advice. Our read. Disclosure: we hold personal positions in Polymarket itself and may earn a commission from Polymarket referrals. These markets are thin and the odds below move, so read direction over ticks.

Scorecard: the book is flat for the twelfth consecutive issue, and first, a correction to our own record that is not small.

On 31 August, with the September hike contract at 52.5%, we declined to buy it. We also published a way back in. A fall below 40% with Warsh's position unchanged, we wrote, "would be a mispricing claim rather than a threshold, and we would act on that."

That condition was met, and we missed it. On the morning of Thursday 3 September, during Governor Waller's remarks backing a hold, the contract traded as low as 37.5%, and shares changed hands between 37 and 39 cents for about twenty minutes, with more at 39 cents on Friday morning. Warsh had said nothing new. Our 7 September issue reported that the contract fell to 40.5%, and our 9 September issue said the rule had not fired. Both were wrong. We were reading a coarser price series than the one the market actually traded, and our own trigger fired on a morning we were not looking closely enough to see it.

The contract was 79.5% early Monday, and it resolves on Wednesday. If it resolves yes, a contract our published rule told us to buy below 40 pays 100. We will record that as a missed trade that our own rule called, which is worse than a trade we chose to pass on, and the ledger will carry it that way. The fix is procedural and it starts now: every published trigger gets checked against the full trade history, not a daily or hourly figure, before we say whether it fired.

The only entry condition on the board is high yield OAS widening through 3.00% into the late October Treasury cash peak. Stated three ways, because all three are true and quoting one of them is a decision about how we look. The latest print is 2.70% for 10 September, so the trigger is 30 basis points away. Against the value thirty days earlier, 2.72% on 11 August, spreads have moved 2 basis points the wrong way. And off the 2.60% low of 28 August they have widened 10 basis points. Neutral, bad for us, good for us, same series, same week.

A branch change, and we are calling it that rather than a drift. On 9 September we told you the high yield OAS test was tracking Branch C at 2.68%. FRED's latest print is 2.70%, which is the Branch B floor exactly. Branch B is 2.70% up to but not including 3.00%. Branch C is 2.40% up to but not including 2.70%. At precisely 2.70% it resolves B, not C. So that test crossed a boundary since Wednesday, and the time to say so is now, not on 7 October when the grade is already written. Two things are true about 2.70% at once, and the second cuts against us: it is inside Branch B, and it is exactly the line in our separate 2 September commitment, that spreads at or inside 2.70% when the Treasury cash peak arrives in late October would mean our framing was wrong and we would say so in those words. That commitment is still live. Nothing in this issue retires it.

And a stale number in our own ledger, which is a worse failure than a wrong forecast. Our ledger file had been carrying CLARITY at 20.5%. The contract was 14.5% on the day we published on 9 September, 21.5% on Sunday evening and 34.5% by half past midnight. So the file was wrong in one direction and then wrong in the other inside five days, because nobody re-pulled it. The published issue had the right number on Wednesday; the record behind it did not. Corrected and logged, and we would rather you heard it from us.

Four dated tests are open, none of them grades today, and a fifth is set below. September's spot ETF monthly total grades 7 October: Branch A is still live, but the margin collapsed 60% this week, from roughly $770 million to about $307 million, with thirteen sessions left to run. The 30 September Treasury cash close grades 5 October: the TGA was $818.1 billion on 10 September, down $205.5 billion from 31 August, which is further into Branch C. Carry the caveat we set with it: a month start drop is routine entitlement spending and the account normally rebuilds through the month, so this is not yet evidence of a Treasury undershoot. The difficulty test at block 967,680 grades Monday 21 September, and it is covered above. The high yield spread test grades 7 October on the 30 September value, and it is the branch change above.

One new dated test, grading in the Wednesday 7 October issue. The published number is the ten year Treasury inflation indexed constant maturity yield, FRED series DFII10, the value published for 30 September 2026.

Branch A: 2.80% or above. Branch B: 2.55% up to but not including 2.80%. Branch C: 2.30% up to but not including 2.55%. Branch D: below 2.30%.

One series, one date, four ranges, no gaps and nothing outside them. The 10 September value of 2.55% sits exactly on the Branch B floor, and Treasury's own table shows 2.60% for Friday, so today's reading is Branch B either way. This is the number the whole issue rests on. If the FILTER is right that credit has priced a hike that works, the real yield holds here or pushes higher into the meeting and then stops climbing. Branch A is 20 basis points above Friday. Branch D is what it looks like if the market decides on second thought that Wednesday was a mistake.

Fed decision in September, 25 bps increase | Market: 79.5% ⚪ NO TRADE, and the trigger we missed is above
No change is 20.5%. Volume on the hike leg itself is $30.4 million. The event across all five outcomes is $147.4 million, and those two numbers are not interchangeable, though they get quoted as if they were. The arc since we last published: 53.5% on 9 September, 62.5% on the PPI print, back to 57.5% by Friday morning, 79.5% one minute after CPI, a peak of 82.5%, and 79.5% early Monday. Net plus 26 points in three sessions. At 79.5% you are staking 79.5 cents to win 20.5 on a binary that resolves in a little over two days while the committee is under its own blackout and cannot say anything that would move it. That is not a price with edge in it. It is a price with a countdown on it.

CLARITY Act signed into law in 2026 | Market: 34.5% ⚫ EXITED at roughly 33 to 37% after entering at 41% and 56%, and the re-entry rule is ruled on before Tuesday
34.5% at half past midnight, on $16.1 million of volume with about $900,000 traded in the last day, genuine depth for a legislative contract. It was 14.5% on 9 September and 21.5% at ten o'clock on Sunday night. Thirteen points arrived in the two hours after The Block reported that Senate Republicans had released what they call a final text, with the White House accepting most of the ethics provisions Democrats wanted. Our published re-entry has three legs: the merged text survives markup, cloture carries, and the contract is still under 30%. We are not going to claim a text released ahead of a floor vote settles the first leg, because a release is not a markup, and we can find no committee markup of either the July text or this one. And at 34.5% the third leg is not met either. So the ruling, made before the vote rather than after it: a cloture win on Tuesday does not put us back in on its own. If the price falls back under 30% and a committee marks the text up, the rule is live again. If no committee ever marks it up, the re-entry we published is unreachable as written, and we will retire it in print rather than quietly rewrite it.

Fed rate hike in 2026, any meeting | Market: 86.5% ⚪ NO TRADE, but the seven point gap is the number to check Thursday morning
$9.2 million of volume. This is a hike at any 2026 meeting, not September's, and the two get quoted interchangeably by people who should know the difference. The useful thing is the spread between them. September is 79.5%, this is 86.5%, so the market is paying seven points for the entire scenario in which Warsh holds on Wednesday and hikes later in the year. Seven points is very little room. If the Fed holds and this contract does not fall hard on Thursday, the market is telling you the hike is delayed rather than cancelled, and everything priced off the real yield has to sit through the wait. If it collapses toward 50%, then the whole repricing of the last three sessions was about one meeting and nothing structural.

Three entries, not four. The fourth slot would have gone to the bitcoin versus gold markets, which are quoting on $186 and $142 of daily volume. That is not a price, it is two people, and we would rather run three contracts that say something than pad the board.

Track everything live at polymarket.com, free, no account required.

On 3 August a reader named Darryl sent us a complete, falsifiable forecast and gave us written permission to publish it under his name. One leg of it has now been tested. Verbatim:

"I believe he gets the Straight open again (not necessarily ending military operations in Iran), this brings oil down, brings down inflation significantly enough that there will be no rate hikes from the Fed this year."

Oil went the other way. Brent is up 14.0% in seven trading days and gasoline was over a third of Friday's CPI increase. The oil leg has resolved against him and there is no reading of it that does not.

That is one leg. The Fed leg is not settled until Wednesday afternoon, and 79.5% is not 100%. We are flagging the miss because the whole point of publishing a call is that somebody checks it, and that cuts both ways: our own record this year carries a CLARITY loss, a short we covered for a loss on a rule we wrote ourselves, and, as of this issue, a trigger of our own that fired while we were not looking. Darryl named a mechanism, named the consequence and put a date on it, in public, where it could fail. Almost nobody does that. The people who do are the only ones who learn anything from being wrong.

One question: if the Fed hikes on Wednesday and high yield spreads are still inside 3.00% on Friday, is credit right or is credit late?

Pick one and give us the sentence behind it. A hedge is not an answer, and the point of a binary is that it can be wrong.

Hit reply. We read every response, and the best calls run Wednesday.

See you Wednesday morning, a few hours before the decision, with the first high yield spread print that includes CPI day, the setup into 2 pm and the dot plot, and whether the CLARITY cloture vote found its sixty.

The Baseline Crypto Team

Our three free tools are down for a fix. A formatting error broke all three pages, the wallet check, the custody audit and the public scorecard, and we did not catch it until this weekend. The scorecard was also still carrying data from 10 August, including the stale CLARITY number we corrected above, which is its own small embarrassment. We are repairing all three and will say here when they are back.

In the meantime, this week's is easy, and it needs no tool. Somebody you know completed KYC at an exchange years ago and has not thought about it since. Revolut just showed what sits in that file and how thin the thing guarding it is: a passport scan, a selfie, and a full bitcoin transaction history, released because an email arrived from the right domain. Ask them to list every venue that still holds that same file on them, and to close the accounts they no longer use. Closing an account does not delete the file, but it stops the file growing.