On Monday this newsletter gave you one thing to watch. Last week's ETF selling had been concentrated in two funds, Fidelity's FBTC and ARK's ARKB, which together accounted for 86.5% of a $678.9 million outflow. We read that as a few large decisions rather than a crowd, labelled it an inference, and set a test: if the reading was right, those two funds should go quiet this week.

Two days in, on Farside's table as printed when we wrote this, they did more than go quiet.

Monday 12 October: FBTC zero, ARKB zero. The day's total was minus $65.4 million, and all of it was BlackRock's IBIT, the fund that had finished last week flat.

Tuesday 13 October: FBTC plus $103.1 million. ARKB zero. MSBT plus $6.7 million. IBIT had not reported, so Tuesday's total of plus $109.8 million is partial and will move.

The fund that put out $380.3 million in the five days to Friday took in $103.1 million on Tuesday. That is consistent with the reading we gave you on Monday, and we would rather say it plainly than oversell it: two days of five, and a reading about which funds are selling is not a reading about why.

Because the price did not wait for the funds. On Coinbase daily closes, on UTC days, bitcoin ended Sunday at $83,678.29 and Monday at $81,742.50, down 2.3%, on a day when the two funds we had named did nothing. That close was $49.71 above last Thursday's. The whole bounce we described on Monday morning was gone by the end of the same day. Then it turned: $83,101.99 at Tuesday's close, up 1.7%. It was $83,235 at 10:30 p.m. New York time on Tuesday.

The hourly path matters here, because it does not flatter anybody's story, including ours. Monday's fall came in pieces. Bitcoin drifted from about $83,900 before dawn to about $82,200 by mid-afternoon New York time, climbed back to about $83,000 by 4 p.m., then slid again through the evening, with no cause we can name for either leg. The sharpest drop came at 8 p.m. New York time on Monday, the hour Seoul's market opened on Tuesday. South Korea's KOSPI fell about 5.5% that session on worries about chip makers and AI valuations, and bitcoin touched $80,927.48 in that hour. CoinDesk reported about $800 million of leveraged crypto positions forcibly closed over the 24 hours to early Tuesday, around 86% of them bets on prices rising; that figure is CoinDesk's compilation of exchange data, and totals vary by provider. Then bitcoin was back above $83,000 within three hours, overnight in New York, before any US fund had traded a share.

So Fidelity's $103.1 million arrived after the recovery, not before it. It may have helped the price hold. It could not have caused the turn.

That is now three explanations in nine days that this newsletter has tested and found unable to carry the price by itself: liquidity on 5 October, the rates picture on Friday, and on Monday the ETF sellers, who behaved as predicted while the price did something else. The best fit for the sharpest hour of the week is a stock market on the other side of the world. Each of those inputs is real. None of them, on its own, tells you where bitcoin goes over a few days.

Consumer prices for September land at 8:30 this morning, an hour after this email. The FILTER deals with the claim that the print decides bitcoin's week, and the STACK writes down, before the number exists, what would count as evidence either way.

Three metrics.

Metric 1, Fear and Greed: 62. The reading stamped 14 October, posted the evening before. Monday printed 69. Since then: 57, 62. Twelve points down on a 2.3% fall, five back on a 1.7% recovery. The gauge is following the price closely, which is what half of its weighting is built to do.

Metric 2, ETF flows: minus $65.4 million on Monday, plus $109.8 million on Tuesday before IBIT. October to date, 1 to 13 October, before IBIT's Tuesday figure: minus $341.9 million.

Metric 3, Strategy's stack: 848,250 BTC. Its filing dated Monday covers 5 to 11 October: 250 BTC bought for $21.5 million, average cost now $75,444. Against $83,235 that is 10.3% over cost. What else it spent money on is in the RADAR.

No new exchange hack we can verify, and nothing new from Ledger beyond its reseller notice. CLARITY has not moved.

"CPI at 8:30 decides bitcoin's week. A hot print means higher yields, and higher yields sink bitcoin."

Possibly. But we have watched that chain fail in both directions for two weeks, and it is worth writing down what would count as evidence before the number exists.

Start with what is right in it. Inflation data moves the bond market, and the bond market sets the price of money that everything else is measured against. Real yields, the return on Treasuries after inflation, were 2.91% on 9 October, close to the highest of the year. An asset that pays no yield has to compete with that. Over months, that competition is real.

Now the part the claim skips. It has two links, and only the first one is reliable. A surprise in CPI moves yields; that happens most months and you can watch it within minutes. Whether yields then move bitcoin is the link that has failed recently. In the week to 4 October yields hit their highest level in more than two decades and bitcoin rose 3.4%. From 5 to 8 October yields fell and bitcoin fell 5.6%. On Monday the bond market was shut and bitcoin fell 2.3%. On Tuesday the ten-year eased three basis points from Friday and bitcoin bounced 1.7%, which fits the textbook for exactly one day.

Then the bar the print has to clear. The bond market's own ten-year inflation expectation was 2.33% on Tuesday, the same as Friday and three basis points below the 2.36% it held at the start of October. On the bond market's own gauge, long-run inflation fear is not what has been pushing rates up this autumn. A single month's print that leaves that number where it is will be read by the bond market as noise, whatever the headlines say.

So here is what would actually be evidence, written down before 8:30. If CPI surprises, the ten-year moves five basis points or more by the close, and bitcoin moves 0.5% or more the opposite way over the same day, that is one day for the old textbook. If the ten-year moves five basis points or more and bitcoin moves the same way, or less than 0.5% either way, that is one more day of the link failing. If the ten-year moves less than five basis points, the print was not a surprise to the bond market and it tells you nothing about bitcoin. The STACK records it exactly that way, so nobody has to take our word for which it was.

Three numbers that define the rest of the week.

8:30 a.m. today, September CPI. The STACK below fixes, in advance, what the print would have to do to the ten-year and to bitcoin to count as evidence either way. The bond market's ten-year inflation expectation was 2.33% on Tuesday and the ten-year real yield 2.91% on Friday. Producer prices follow on Thursday and import prices on Friday, on the BLS calendar.

$103.1 million, Fidelity's FBTC on Tuesday. The fund that did more of last week's selling than any other was the largest buyer on Tuesday before IBIT reported. Three trading days remain in the test. If FBTC and ARKB stay at zero or better, last week was a few decisions that have been made. If the selling comes back, it was not.

Friday 16 October, MSCI's own deadline. MSCI says it will announce the result of its consultation on a rule that would remove Strategy from its global indexes by then. Separately, the credit question we opened on 5 October runs out on the 19th, with spreads at 3.11% on Monday, between the two exits we named: above 3.46%, or under 3.00%.

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. Nothing grades this morning. One test is running, and one record is set.

The running test, set Monday. The rule: the combined net flow of Fidelity's FBTC and ARK's ARKB from Monday 12 to Friday 16 October, on our sum of Farside's printed daily figures for those two funds. Branch A is zero or above. Branch B is below zero, down to and including minus $100 million. Branch C is below minus $100 million.

After two days of five: plus $103.1 million. Monday: FBTC zero, ARKB zero. Tuesday: FBTC plus $103.1 million, ARKB zero, on the figures printed when we wrote this. That is Branch A so far, and it is the direction the reading implied. Three days remain, and on Monday's issue we said one quiet day settles nothing. The awkward part is in the SIGNAL: the funds behaved as the reading predicted, and the price moved on its own timetable anyway.

The record, set before 8:30 this morning. This is not a trade and not a branch test. It is a written description of what September's CPI would mean for the claim in the FILTER, fixed before the number exists. Measured on Treasury's ten-year from Tuesday's 5.21% to Wednesday's close, and on bitcoin's Coinbase close for Wednesday 14 October against Tuesday's $83,101.99:

The numbers to check it with are all public, so you can grade it yourself on Wednesday evening. That is the point of writing it down first.

The board has no live rule. Nothing has been added since the October entry was retired on 5 October.

Fed decision in October, 25 bps increase | Market: 19.5% ⚪ NO TRADE, no rule
No change is 80.5%, as of 10:30 p.m. Tuesday New York time. It was 16.5% at midnight on Sunday, so three points went in over two days. We found no Fed speech posted in the window that would explain it, and we are not going to guess at a cause for three points in a thin market the night before CPI. What would move it: this morning's CPI, and any voter saying October out loud before the 28th.

Fed decision in December, 25 bps increase | Market: 75.5% ⚪ NO TRADE
No change is 21.5%, as of Tuesday evening. A point up from Sunday's 74.5%. The hike is still the clear base case. What would move it: CPI this morning, then the 28 October statement.

Zero Fed rate cuts in 2026 | Market: 96.15% ⚫ EXITED at 88.75%, a loss
As of Tuesday evening, up from 95.55% on Sunday. Short from 78%, trimmed at 85%, covered at 88.75% on a published stop on 3 August, and recorded as a loss in every issue since. What would move it: a dot below the current midpoint in the December projections.

CLARITY Act signed into law in 2026 | Market: 6.15% ⚪ NO TRADE, nothing to trade on
As of Tuesday evening, up from 5.05% on Sunday. We found no news behind the move, and a point in a market this thin is not a signal. What would move it: a markup of the substitute text, and a new rule written before the fact.

The book. Flat for the twenty-first consecutive issue, recomputed from the archive: twenty issues published after the 3 August exits, so this is the twenty-first.

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

On Monday we asked where your bitcoin lives: fund, exchange or keys.

Nothing came back. Our file of reader replies still has nothing new since 3 August.

This issue is about a reading that held for two days and still did not explain the price. The question goes with it.

One question: when bitcoin moves 2% in a day, what do you check first? Reply with one word: price, news or nothing.

"Nothing" is a real answer, and possibly the best one.

Hit reply. We read every response.

The Baseline Crypto Team

Two tools, both free, no account needed.

The hardware wallet checker takes your device and firmware version and tells you whether you sit inside a population with a known problem, and what to do if you do. The custody audit is nine questions about what happens to your coins when something goes wrong, and it is deliberately unkind about backups and inheritance.

This issue spends a lot of words on why the price moved and ends up not knowing. That is a fair place to end up after a week like this one. It is also a reminder of where your attention pays. Nothing you do this morning changes what CPI prints or what the ETFs do tonight. Whether your coins survive a lost device or a family emergency depends mostly on choices you make in advance.

Monday's issue carried Ledger's warning to buyers from one reseller in South East Asia. The checker tells you whether your device and firmware sit inside a known problem, and the audit asks what happens if the person holding the keys is not you. Neither can tell you whether a device was tampered with before it reached you. That part is decided at the moment you buy it, by buying from the maker or a reseller the maker lists.

Send the hardware wallet checker to the friend who bought their device from a marketplace listing because it was cheaper. The checker is a start. Where they buy the next device matters more.