Four markers expired today. Two passed, two failed, and one of the failures failed for a reason that indicts the whole board: its bar was copied out of the series it was supposed to be testing. That is now the second time in four days we have found the same defect — and the marker that passed today, C1, passed despite having exactly the same defect. Yesterday this column said C1’s bar was “badly chosen” and could only clear on a strict break of a level the tape had touched once. Nine hours later the tape cleared it by 1.62%.
Both facts are true at once and they do not cancel. A badly designed marker that happens to pass is still badly designed. What follows is the full board, the finding, and three new markers written against the lesson.
The board, 28 August 2026
| ID | Bar | Deadline | Reading now | State |
|---|---|---|---|---|
| C1 | US-session print > $79,563.71 | 28 Aug | $80,848.74 | PASS, +1.62% |
| C2 | Top-trader ratio ≥ 2.0000 at final read | 28 Aug 00:00 UTC | 2.0712 | PASS, +3.56% |
| Z3 | ETF net 24–27 Aug ≥ $0 | 27 Aug | +$1,126.4m | PASS, settled |
| Z2 | Open interest ≥ 111,988 BTC | 28 Aug | 108,761.98 | FAIL, −2.88% — and a bad bar |
| X2 | Brent settle ≥ $95.00 | 28 Aug | $89.28 | Failing, needs +6.41% today |
| Y1 | 30-year close ≥ 5.35% | 28 Aug | 5.199% | Failing, needs +15.1bp today |
| A2 | August close ≥ $71,440.63 | 31 Aug | $79,885.13 | Passing, +11.82% |
| B2 | August ETF net ≥ $3,424.9m | 31 Aug | $3,524.3m | Passing, $99.4m cushion, 2 sessions |
| B3 | Strategy 31 Aug 8-K buys > 0 | 31 Aug | — | Open |
| C3 | Difficulty cut > 2.00% | ~6 Sep | −0.8497% projected | Failing |
| D1 | Any funding settlement > 0.0100% | 30 Sep | 0 of 500 | Open |
| D2 | Retarget lands −1.50% to 0.00% | ~6 Sep | −0.8497% projected | On track, unsettled |
Two settlement caveats, stated up front. X2 settles at the ICE Brent close (19:30 UTC) and Y1 at the Treasury close (21:00 UTC), both after this piece publishes. Neither is graded here as final. X2 needs a 6.41% single-session move in crude; Y1 needs a 15.1 basis-point single-session rise in the 30-year. We are calling both failing and will publish the settlements tomorrow rather than pretend a morning reading is a close.
THE FIND — Z2’s bar is the maximum of the series it measures
Z2 asked whether open interest on the Binance USDT-margined perpetual would reach 111,988 BTC by 28 August. This morning we pulled the endpoint at its maximum depth rather than the default page. The endpoint retains 31 rows — 29 July to 28 August 2026 — and that is all it will ever serve.
| Statistic, full retained series | Value | Date |
|---|---|---|
| Rows available | 31 (endpoint maximum) | 29 Jul – 28 Aug 2026 |
| Maximum | 111,988.29 BTC | 15 August 2026 |
| Minimum | 103,253.30 BTC | 29 July 2026 |
| Latest reading | 108,761.98 BTC | 28 August 2026 |
| Z2’s bar | 111,988 | — |
The bar is the maximum. To the hundredth. Z2 was not a question about whether leverage would return; it was a question about whether a 31-row series would exactly re-attain its own all-time high inside the window, and it was written by reading that high off the screen. That is the identical construction as C1, whose $79,563.71 was selected by taking the best US-session print of the 25 August spike.
Set alongside B1, retired yesterday as unfalsifiable, and A3, a bet on a rounding boundary beneath a series that had not moved in two months, the pattern is now four for four:
| Marker | Bar | Where the bar came from | Outcome |
|---|---|---|---|
| A3 | July core PCE ≤ 3.2% | A rounding boundary under a flat series (3.3% June, 3.3% July) | FAIL, by 0.1pp |
| B1 | Funding > 0.0100% | A ceiling the series had not touched in 500 settlements | FAIL, unfalsifiable |
| C1 | Print > $79,563.71 | The maximum of the series | PASS — on a bad design |
| Z2 | OI ≥ 111,988 | The maximum of the series | FAIL |
The common cause is the one we named yesterday and then failed to generalise: we were reading numbers off whatever window the default request happened to return. Yesterday it was funding — a 24-settlement streak that was actually 500. Today it is open interest — a 14-row window whose oldest row, 15 August, happens to be the series peak. In both cases the default page was not a sample of the data. It was the data, and we treated it as scenery.
The rule, extended. Yesterday: every streak, series high, series low and “consecutive” count ships with the depth of the pull that produced it. Today: no marker bar may be a value the target series has already printed. If the bar equals a historical reading, the marker is asking whether history repeats to two decimal places, which is not a question anyone holds a view on.
C1 passed, and here is exactly when
The 27 August US session, hour by hour, on Binance BTCUSDT (highs, UTC):
| Hour (UTC) | Hour (ET) | High | vs C1 bar $79,563.71 |
|---|---|---|---|
| 13:00 | 9:00 a.m. | $79,568.00 | +$4.29 |
| 14:00 | 10:00 a.m. | $80,388.00 | +$824.29 |
| 15:00 | 11:00 a.m. | $80,799.91 | +$1,236.20 |
| 16:00 | 12:00 p.m. | $80,624.00 | +$1,060.29 |
| 17:00 | 1:00 p.m. | $80,848.74 | +$1,285.03 |
| 18:00 | 2:00 p.m. | $80,442.47 | +$878.76 |
| 19:00 | 3:00 p.m. | $80,294.11 | +$730.40 |
Seven consecutive US-session hours above the bar, peaking at $80,848.74 between 1:00 and 2:00 p.m. ET. This was not a wick. C1 passes cleanly. It passed on the opening day of Jackson Hole, hours after the host bank’s president said the policy rate is “very accommodating.” The day then closed at $80,249.58, the highest daily close of the advance, and the Asian session extended to $81,478.87 — the highest Bitcoin print since May 2026.
https://www.youtube.com/watch?v=t1z4ks3u1NgMore Crypto Online, “Bitcoin Hits Major Resistance: Will It Hold?”, published 26 August 2026 — a resistance read published the day before the break, included because the question it asks is the one C1 answered.
Open interest: the largest one-day build in the retained series
Z2 failed on the bar and yet the underlying series did something genuinely new. Open interest went from 105,593.59 BTC on 27 August to 108,761.98 BTC on 28 August — +3.001% in a single day, the largest daily build in all 31 rows the endpoint retains.
| Rank | Date | One-day change in open interest |
|---|---|---|
| 1 | 28 Aug 2026 | +3.001% |
| 2 | 12 Aug 2026 | +2.803% |
| 3 | 1 Aug 2026 | +2.746% |
| 4 | 15 Aug 2026 | +2.195% |
| — | 18 Aug 2026 | −4.098% (largest fall) |
This is the datum that starts rebuilding the frame we demolished yesterday. For four runs this column argued the advance was unlevered on the strength of a funding observation that turned out to describe 166 days in both directions rather than August. We retired that inference. Here is a replacement built from something else: leverage was absent through the run-up and is arriving now, in one day, into an event.
Funding: 90 of 90 positive, 0 of 500 above baseline
The two statements in that heading are both true and they are not the same statement. Conflating them is the single most common error in derivatives commentary, and we found a live example of it this week — a widely repeated line that funding has been “positive in 88 of the prior 90 eight-hour windows.” Our own pull of the endpoint at maximum depth:
| Window | Settlements | Positive | Negative | Mean | Max | Above 0.0100% |
|---|---|---|---|---|---|---|
| Full series (14 Mar – 28 Aug) | 500 (endpoint max) | — | 136 (27.2%) | 0.00261% | 0.0100% | 0 |
| Last 90 (29 Jul 08:00 – 28 Aug 00:00) | 90 | 90 (100%) | 0 | — | 0.0100% | 0 |
| August only | 82 | 82 (100%) | 0 | 0.006339% | 0.0100% | 0 |
| Last 30 | 30 | 30 (100%) | 0 | 0.007934% | 0.0100% | 0 |
So the correct figure is 90 of 90, not 88 of 90, and every one of August’s 82 settlements has been positive. And yet the number of settlements above the 0.0100% baseline across the entire 500-row window remains zero. Longs pay every eight hours; they have never once paid a premium for the privilege. The mean August settlement, 0.006339%, annualises to roughly 6.9% — positive carry, but the boring kind.
A correction against ourselves, and it is instructive. Yesterday we published that the series touched exactly 0.0100% “36 times, 7.2%” of the 500. Today the same pull returns 37 times, 7.4%. Nothing was wrong yesterday and nothing is wrong today: a new settlement hit the cap and the oldest row rolled off. That is precisely why the depth of the pull now ships with every count. A rolling-window statistic without its window is a number with a shelf life nobody told you about.
Read the two series together and the picture is coherent for the first time in a week. Open interest +3.00% in a day; funding uniformly positive; funding never above baseline. Positions are being added, longs are paying, and the perpetual still is not bidding for leverage. That is what an event hedge looks like, not what a crowded long looks like — but it is also positioning that did not exist on Wednesday and therefore can unwind on Friday.
Cohorts: C2 passes on its final read, and the spread snaps back to the extreme
| Date (00:00 UTC) | Top traders | Crowd | Spread | Crowd positioning |
|---|---|---|---|---|
| 25 Aug | 2.0203 | 0.9448 | 2.138× | Net short |
| 26 Aug | 2.2560 | 1.0008 | 2.254× | Dead flat |
| 27 Aug | 2.1462 | 1.0859 | 1.976× | Net long |
| 28 Aug | 2.0712 | 0.9201 | 2.251× | Net short |
C2 passes on its final scheduled reading: 2.0712 against a 2.0000 bar, a 3.56% cushion. But the shape of the pass deserves attention. The top-trader ratio has now fallen for three consecutive days, −8.19% from the 26 August high, and closed toward its bar rather than away from it. Large accounts are 67.44% long, down from 69.29% two days ago.
The crowd did the opposite, violently. Retail flipped from net long (1.0859) straight back to net short (0.9201) in one reading — 47.92% long against 52.08% short — on the day price made its highest close of the advance. The spread re-widened to 2.251×, within 0.14% of the 26 August 30-day maximum, one day after we noted its first contraction in nine readings.
So the honest read on the cohort table is: Wednesday’s contraction was noise, not a turn. We flagged it as potentially meaningful yesterday. It lasted one day. Written down here so it counts against the record.
https://www.youtube.com/watch?v=RwxmkukaScwTradeProsTV, “Bitcoin LIVE BTC/USD Technical Analysis & Price Prediction | Massive Breakout Incoming 16 AUG 2026,” streamed live 15 August 2026 — the session that coincided with the open-interest peak of 111,988.29 BTC that became Z2’s bar.
ETF: Z3 settles, B2 flips to passing, and breadth collapses
Z3 is settled and passed. The 24–27 August window required a net flow at or above zero; it delivered +$1,126.4m across four sessions ($337.6m, $314.3m, $232.2m, $242.3m).
B2 has flipped from “needs $47.6m per session” to passing. August month-to-date is $3,524.3m across 19 rows against a $3,424.9m bar — a cushion of $99.4m with two sessions left (28 and 31 August). It is not settled: two sessions of the kind of outflow August has already produced would erase it.
The 27 August composition is the finding, and it is uncomfortable:
| Fund | 27 Aug flow | Note |
|---|---|---|
| IBIT | +$277.6m | 114.6% of the complex total |
| FBTC | −$83.6m | Largest single-session outflow of August |
| ARKB | +$29.7m | |
| BITB | +$21.7m | |
| GBTC | −$27.2m | Second consecutive outflow |
| Grayscale Mini (BTC) | +$11.7m | Offsets only 43.0% of GBTC |
| HODL / MSBT | +$5.7m / +$6.7m | |
| Total | +$242.3m | Ninth consecutive positive session |
| Ex-IBIT | −$35.3m | Everything else, net |
The ninth-best-looking session of the streak was an IBIT-only session. Strip BlackRock out and the other eleven vehicles gave back $35.3m. On Wednesday this column noted that breadth outside the two giants had widened 52.8% on a smaller headline day; Thursday reversed it completely. The lesson holds in both directions: the headline number and the breadth number are separate facts, and they routinely disagree.
And our Grayscale finding weakened on its very next observation. We published yesterday that GBTC’s outflows are substantially migration into Grayscale’s own cheaper Mini Trust, on an offset series of 71.3% (10 Aug), 107.2% (13 Aug) and 92.9% (26 Aug). Thursday’s pair came in at 43.0% — less than half. August still nets the complex a receiver (GBTC −$137.2m, Mini +$200.9m, net +$63.7m), but the offset ratio is not the stable structural constant we implied. Four observations, range 43.0% to 107.2%. That is a tendency, not a mechanism, and yesterday’s copy was firmer than four data points support.
Difficulty: C3 dead, D2 on track, and the rule holds a fourth day
At block 964,393, with 1,271 blocks to the retarget at 965,664, the epoch is 36.95% complete and the projected adjustment is −0.8497%. Estimated retarget date is roughly 6 September 2026. The previous retarget was −1.3122%.
| Reading | Epoch progress | Projected change |
|---|---|---|
| 26 Aug | ~15% | −3.08% |
| 27 Aug, 06:12 UTC | 29.86% | −0.8409% |
| 27 Aug, 06:27 UTC | 30.01% | −0.6327% |
| 28 Aug, 06:20 UTC | 36.95% | −0.8497% |
The projection has now travelled 2.45 percentage points in 48 hours and is still not half way through the epoch. Our 23 August rule — a difficulty projection is not a forecast until the epoch is past 90% complete — is holding for a fourth consecutive day. C3 (a cut deeper than 2.00%) is failing badly. D2 (the retarget lands between −1.50% and 0.00%) is currently inside its band, and by our own rule that means nothing yet. Three-day average hashrate: 890.40 EH/s (26 Aug), 964.39 (27 Aug), 894.73 (28 Aug, partial).
Three new markers, written against today’s lesson
Each of these is chosen so that its bar is not a value copied from the series it tests. Where a bar is arbitrary, we say so and give the rationale in advance rather than after the grade.
| ID | Claim | Deadline | Why the bar is defensible |
|---|---|---|---|
| E1 | September 2026 US spot bitcoin ETF net flow ≥ $0 | 30 Sep | Zero is a natural boundary, not a reading lifted off the chart. 2026 has produced four negative months and four positive. |
| E2 | Bitcoin’s 30 Sep UTC daily close is above its 31 Aug UTC daily close | 30 Sep | The bar is a fact that does not yet exist. It cannot be reverse-engineered from history. |
| E3 | Binance BTCUSDT daily open interest prints ≥ 115,000 BTC on any day | 30 Sep | A round number 2.69% above the retained series maximum of 111,988.29. Stated in advance: this is a level the series has never printed, so the marker is hard by construction and we are not pretending otherwise. |
Z2 is retired rather than rolled. Rolling it forward would carry the defective bar into September. If the question is whether leverage returns to its August peak, E3 asks a cleaner version of it with the bar set deliberately rather than by transcription.
What today actually established
One. The pattern in our marker design is not anecdotal. Four expiring markers, four bars that were either copied from the target series, set at an unreachable ceiling, or aimed at a rounding boundary beneath a flat series. The corrective is a construction rule, not more care.
Two. A defect and an outcome are independent. C1 passed on a bad bar; Z2 failed on the same bad bar. If we only audited the failures we would have learned half of this.
Three. The market did something new underneath all of it. The largest one-day open-interest build in the retained series, at flat funding, with the crowd short and large accounts trimming, into a keynote at 10:00 a.m. ET. That is the fact worth carrying into tomorrow — not the score.
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. The markers described below are a public record of this desk’s reasoning, not trade signals, and several of them have failed. Figures are as of the timestamps stated. Do your own research and, if you need advice, speak to a regulated professional.