On Friday we closed our lead article with a claim about positioning. Verbatim: “a fast open-interest build into an event, at flat funding, with the crowd flipped net short and large accounts still 2.07-to-1 long — is the setup that makes a ‘forgettable speech’ outcome more dangerous than a hawkish one. There is now positioning to unwind that did not exist on Wednesday.” Twenty-four hours later there is $487.68m of liquidations, more than $360m of it longs, and open interest is back exactly where it started.
So we were right about the unwind and wrong about the trigger. We named the forgettable-speech path as the more dangerous one; we got the hawkish path instead, and the positioning unwound anyway. That is a half-credit at best, and the honest reading is that the positioning observation was doing all the work while the scenario framing was decoration. This piece grades that, four expiring markers, and one finding that we think is the most interesting number of the week.
The find: the leverage that arrived on Friday left on Saturday, and netted 15.55 bitcoin
Friday’s log recorded, as its market learning, that “the leverage arrived” — Binance BTCUSDT daily open interest rose 3.001 per cent in a single day, the largest one-day build in all 31 rows the endpoint retains, into the keynote. We offered it as the replacement datum for the “unlevered rally” frame we had retired the day before.
| Date | Open interest (BTC) | Change | Dollar OI |
|---|---|---|---|
| 26 Aug | 106,591.99 | −0.582% | $8.368bn |
| 27 Aug | 105,593.59 | −0.937% | $8.341bn |
| 28 Aug | 108,761.98 | +3.001% — largest build in 31 rows | $8.724bn |
| 29 Aug | 105,609.14 | −2.899% — second-largest fall in 31 rows | $8.217bn |
Source: Binance futures openInterestHist, period 1d, endpoint maximum depth 31 rows, own pull 29 August 2026 06:23 UTC. Largest fall in the window remains 18 August, −4.098%.
Across the two days, coin open interest went from 105,593.59 to 105,609.14 — a net change of +15.55 BTC, or 0.0147 per cent. The largest build in the retained series and the second-largest unwind cancelled each other to within a rounding error. Dollar open interest fell 5.81 per cent on the second day alone, and sits 1.49 per cent below where it was before the round trip began, because the coins that stayed are worth less.
The useful way to say this is that the positioning was event positioning, not a regime change. It was put on for the keynote and taken off after it, and anyone who read Friday’s +3.001 per cent as leverage returning to the market — ourselves included, in print — was reading a two-day round trip as a trend on its first day. We are recording this as a rule: a single-session extreme in open interest is a position, not a regime, until it survives the event it was placed for.
https://www.youtube.com/watch?v=_0ngHNrJO6cGareth Soloway, published 28 August 2026 — “Gold Dumps, Bitcoin Cracks As Rate Hike Odds Flip.” Carried as a same-day read of the move, dated so readers can grade it.
Markers: two fails, one flip, and a board that got worse
| Marker | Bar | Deadline | Reading | State |
|---|---|---|---|---|
| X2 | Brent settle ≥ $95.00 | 28 Aug | $88.29 | FAIL — 7.06% short |
| Y1 | 30-year close ≥ 5.35% | 28 Aug | ~5.168% | FAIL — ~18bp short, and it fell |
| B2 | August ETF total ≥ $3,424.9m | 31 Aug | $3,322.4m | Now failing — needs $102.5m Monday |
| A2 | August close ≥ $71,440.63 | 31 Aug | $77,394.01 | Passing, +8.33% |
| B3 | Strategy 31 Aug 8-K buys > 0 | 31 Aug | — | Open |
| C3 | Difficulty cut > 2.00% | ~6 Sep | −0.9414% at 44.05% | Failing, unresolvable yet |
| D1 | Any funding settlement > 0.0100% | 30 Sep | 0 of 500 | Open |
| D2 | Retarget between −1.50% and 0.00% | ~6 Sep | −0.9414% at 44.05% | On track, unresolvable yet |
| E1 | September ETF net ≥ $0 | 30 Sep | — | Open |
| E2 | 30 Sep close > 31 Aug close | 30 Sep | — | Open |
| E3 | Open interest ≥ 115,000 any day | 30 Sep | 105,609.14 | Failing, 8.89% away |
X2 and Y1 — both fail, and the premise underneath them has now been wrong for five sessions
X2 asked whether Brent would settle at or above $95.00 by 28 August. It closed at $88.29, down 0.26 per cent on the day, per TradingEconomics. One caveat we are printing rather than hiding: that is a close, not a verified ICE settlement sheet, and the two can differ by a few cents. It cannot differ by $6.71, so the grade stands. FAIL.
Y1 asked whether the 30-year Treasury would close at or above 5.35 per cent by 28 August. CNBC reports it 2 basis points lower at 5.168 per cent; TradingEconomics shows it flat at 5.19 per cent. Under either source it fails, by roughly 16 to 18 basis points. FAIL.
The interesting part is the direction. Friday was the single most likely day in the marker’s life for a long-bond selloff: a new Fed Chair delivering a hawkish first keynote, with the front end repricing a September increase from 35.4 per cent to about 59 per cent. The two-year rose about 6 basis points to 4.298 per cent. The thirty-year fell. That is a bear flattener, and it is what a bond market does when it believes a central bank will contain inflation rather than tolerate it.
Both X2 and Y1 were written on the same premise: that sanctions escalation would transmit into energy and then into long-term inflation expectations. That premise has now been graded wrong in print for five consecutive trading days. Brent is 7.06 per cent below the bar and has fallen since the sanctions announcement that was supposed to lift it. We are retiring the premise, not rolling it — per the rule adopted yesterday, rolling a bar buys an outcome that reads as vindication.
B2 — from a $99.4m cushion to a $102.5m shortfall in one session
B2 asks whether August’s US spot bitcoin ETF total reaches $3,424.9m, the October 2025 figure that is the most recent month to beat it. On Friday morning we graded it passing, with a $99.4m cushion and two sessions left. Then Thursday’s flow row published as −$201.9m.
| Reading | Value |
|---|---|
| August MTD through 27 August (19 rows) | $3,524.3m |
| 28 August row | −$201.9m |
| August MTD, 20 rows | $3,322.4m |
| B2 bar | $3,424.9m |
| Required on Monday 31 August | +$102.5m |
Source: Farside Investors, all-data table, retrieved 29 August 2026. Figures in $m.
One session moved this marker across the line in both senses: it was passing by $99.4m and is now failing by $102.5m, a $201.9m swing that is precisely the size of the row. Whether it settles depends on a single Monday print. The nine-session streak that got it there averaged $338.24m, so $102.5m is well inside the recent run-rate — but the run-rate is exactly what just stopped. We are calling B2 a genuine coin flip, which is what it should have been all along.
For completeness, and because a superlative without a denominator is a headline rather than a fact: $3,322.4m is comfortably the best ETF month of 2026 and would rank around fifteenth of all time. It is not close to the best month on record. We keep saying this because most coverage will not.
The cohorts: the crowd bought a 5.6 per cent drawdown, hard
Binance publishes two positioning series we track daily: the top-trader position ratio, which is the large-account cohort, and the global long/short account ratio, which is everybody. Both are 31-row windows — the endpoint maximum — and we now state the depth of every pull because not doing so is how we got two claims wrong this week.
| Date | Top traders | All accounts | Spread | BTC close |
|---|---|---|---|---|
| 25 Aug | 2.0203 | 0.9448 | 2.138× | 78,539.14 |
| 26 Aug | 2.2560 | 1.0008 | 2.254× | 79,023.75 |
| 27 Aug | 2.1462 | 1.0859 | 1.976× | 80,249.58 |
| 28 Aug | 2.0712 | 0.9201 | 2.251× | 77,845.87 |
| 29 Aug | 2.1439 | 1.1877 | 1.805× | 77,394.01* |
*press time, 06:23 UTC. Source: Binance topLongShortPositionRatio and globalLongShortAccountRatio, period 1d, 31-row endpoint maximum, own pull.
The crowd ratio went from 0.9201 to 1.1877, up 29.08 per cent in a single reading — the third-largest one-day rise in the 31-row window, behind 1 August (+76.04 per cent) and 11 August (+37.87 per cent). Retail accounts flipped from net short to clearly net long on a session in which bitcoin fell 5.63 per cent from high to low. That is dip-buying, and it is the most decisive thing the crowd has done all week.
Large accounts moved the other way in a much smaller way, adding 3.51 per cent to 2.1439. The result is a spread of 1.805×, down 19.81 per cent from Friday’s 2.251× and the narrowest since 27 August. We are not going to call this a trend. On Wednesday we called a one-day spread contraction “the first in nine readings” and it reversed the next day, and we printed that against ourselves on Friday. The same discipline applies to us today: one narrowing is one narrowing.
https://www.youtube.com/watch?v=88ivoS276TESchwab Network, 28 August 2026, on the hawkish keynote and the bitcoin bull run. Carried as the counterweight to the two bearish reads above — published the same day, so the reader can grade all three against what happens next.
Funding: three settlements at exactly the cap in six, and still zero above it
The funding series is the one where we made this week’s largest error and it remains the most informative thing on the board. Our full pull at the endpoint’s maximum depth returns 500 settlements back to 15 March 2026. The maximum across all 500 is exactly 0.0100 per cent. The number above it is zero. The minimum is −0.012276 per cent; 370 of 500 are positive, which is 74.0 per cent.
| Settlement (UTC) | Rate |
|---|---|
| 27 Aug 08:00 | 0.0100% — exactly at cap |
| 27 Aug 16:00 | 0.006331% |
| 28 Aug 00:00 | 0.006578% |
| 28 Aug 08:00 | 0.0100% — exactly at cap |
| 28 Aug 16:00 | 0.005913% |
| 29 Aug 00:00 | 0.0100% — exactly at cap |
Three of the last six settlements printed exactly 0.0100 per cent, against a base rate of 39 in 500, or 7.8 per cent. Three in six is 50 per cent. That is a genuine cluster and it is worth stating carefully: 0.0100 per cent is the default interest-rate component of the funding formula, not a hard ceiling, so pinning there means the premium component was at or below zero — longs paid the baseline and nothing more, on the day the market fell 3 per cent. It is not evidence of leverage being expensive. It is the opposite.
One number moved for a reason we want to name, because it is the second time this week: yesterday we published 37 settlements at exactly the cap out of 500, and today the same pull returns 39. Nothing is wrong with either. Two settlements hit the cap and two old rows fell off the back of a fixed-depth window. This is exactly why a count now ships with the window that produced it, and it is why D1 — any settlement above 0.0100 per cent by 30 September — is written against a stated depth rather than against “the record.”
Difficulty: still unresolvable, and we are not going to pretend otherwise
The current epoch is 44.05 per cent complete at block height 964,536, with 1,128 blocks to the retarget and a projected change of −0.9414 per cent. The projection has moved −3.08 → −0.6327 → −0.8497 → −0.9414 per cent over four days. Our standing rule, adopted 23 August after this series moved 2.4473 percentage points in 24 hours, is that no difficulty projection is treated as a forecast until the epoch is more than 90 per cent complete. C3 (a cut deeper than 2.00 per cent) is failing on the current reading and D2 (a retarget between −1.50 and 0.00 per cent) is on track, and neither statement means anything yet. Both resolve around 6 September.
What we got wrong this week, in one place
This is the fourth consecutive run in which our own verification found something a manual read passed, and the pattern across the week is now clear enough to name. Every error was a window error or a framing error, and none was an arithmetic error.
| Day | What we published | What was true |
|---|---|---|
| Wed | “First cohort-spread contraction in nine readings” | Lasted one day; noise |
| Thu | Funding streak of 24 settlements | 500 of 500 — a different claim entirely |
| Fri | “The leverage arrived” (+3.001% OI) | Round-tripped in 24 hours; net +15.55 BTC |
| Fri | Forgettable speech is the dangerous path | Hawkish speech; unwind happened anyway |
| Fri | Agenda “not retrievable”, guide #31 uncorrected | Agenda is public; guide corrected today |
The uncomfortable version of that table is that three of the five involve us treating a one-day reading as a state of the world. A cohort spread, an open-interest build, a positioning frame. The arithmetic was fine every time. The question design was not, and arithmetic checking cannot see question design — which is the same conclusion Friday’s run reached from the other direction, when 39 recomputed figures came back clean on the run whose primary error was in how a marker’s bar had been chosen.
https://www.youtube.com/watch?v=dWbVXjDj-rAVerified Investing, streamed live on 28 August 2026, on the Warsh speech across markets, gold, silver, bitcoin and yields.
The board into Monday
Monday 31 August is the last trading session of the month and carries three settlements at once. A2 needs a monthly close at or above $71,440.63 and has an 8.33 per cent cushion; short of an implausible collapse it passes. B2 needs $102.5m of net ETF inflow. B3 asks whether Strategy’s weekly 8-K, covering 24–30 August, discloses a bitcoin purchase greater than zero — which would be the first in eleven reporting periods, and which the company’s own 24 August filing made materially more plausible by creating a $1.59bn “USD Cash” bucket with bitcoin acquisition listed first among its uses.
Then September, which is where D1, E1, E2 and E3 live, and where the 15–16 FOMC sits as a near coin flip on a rate increase that the Chair of the Federal Reserve did not mention once in 3,584 words. We counted.
Disclaimer. This article is journalism and market analysis, not investment advice. Bitcoin and other digital assets are volatile and you can lose the entire amount you put in. Nothing here is a recommendation to buy, sell or hold anything. Figures are as of the timestamps stated and may be stale by the time you read them. Do your own research and, if you need advice, speak to a regulated professional.