Markers Tuesday: August Was the First Positive August in the Bitcoin ETFs’ History — and Both Septembers on Record Were Green

Three markers settled overnight on final August data. All three passed. That is the cleanest sweep this desk has recorded, and it is worth saying immediately that a clean sweep is not a good sign about our forecasting — it is a sign the bars were set too low. The useful part of the exercise was not the grading. It was what we found in the table we had to pull in order to grade one of them.

The board

A2 settled on the Bitstamp close of $78,571.17 against a bar of $71,440.63 — a 9.981% cushion, comfortable enough that the marker never told us anything. B3 settled on the 8-K that landed Monday morning: 4,603 bitcoin bought between 24 and 30 August at an average of $80,318, covered separately in this morning’s news piece, where the interesting number turns out to be that the average is higher than bitcoin closed on any day in the purchase window. B2 is the one that repaid the work.

B2 passes — and the table it settles on has a fact in it nobody printed

The bar was $3,424.9m, chosen because that is exactly October 2025’s net inflow, the most recent month larger than August 2026 was at the time we set it. Farside’s 31 August column printed +$216.7m, taking the month to $3,539.1m across 21 sessions. That clears the bar by $114.2m, and it also clears September 2025’s $3,511.0m by $28.1m. Our own recompute of Farside’s all-data table returns $3,539.1m independently, matching the published August column.

August 2026 therefore ranks 10th of the 32 complete months since the ETFs launched in January 2024, and it is the largest month since July 2025 ($6,012.6m) — thirteen months. Every month above it on the list falls between February 2024 and July 2025. But ranking the month was not the find. Sorting the same 32 rows by calendar month was.

Farside Investors all-data table, own recompute, pulled 1 September 2026. The series begins 11 January 2024, the ETFs’ first trading day; there are 32 complete months and no August or September is excluded.

The US spot bitcoin ETFs have now lived through three Augusts and this is the only one that was positive. The prior two were negative, one modestly and one clearly. And both Septembers in the record were green — $1,262.8m in 2024 and $3,511.0m in 2025 — which is an uncomfortable fact for us specifically, because on Sunday this desk published a piece about September seasonality built entirely on price data from 2013 onward.

We are not claiming two observations overturn thirteen years of monthly returns. Two Septembers is not a sample, and we would refuse the argument instantly if someone made it at us. What we will say is narrower and defensible: the largest single marginal buyer of bitcoin that did not exist before 2024 has been a net buyer in both Septembers of its existence, and any seasonality claim made in 2026 is a claim about a market with a bid in it that the 2013–2023 sample never had. The honest position is that the flow record is too short to say anything and the price record is about a different market.

One fund was 87.30% of the month

Concentration is the other thing the monthly total hides. BlackRock’s IBIT took in $3,089.6m of August’s $3,539.1m — 87.30% of the entire industry’s net inflow. On the 31 August session alone IBIT supplied $205.9m of the $216.7m total, or 95.02%, with Fidelity’s FBTC at $6.9m and Bitwise’s BITB at $4.3m; one fund printed an outflow of $13.4m.

This matters for a reason beyond league tables. Five days earlier this desk showed that more than $5 billion of IBIT’s recent share creation was in-kind conversion rather than new money — existing bitcoin swapped into ETF shares, which prints in the flow column but does not represent a marginal buyer. We reported that here on 27 August. When 87.30% of a month runs through one fund, and that fund has a documented in-kind channel, the correct reading of “$3.54 billion of August inflows” is that it is an upper bound on new demand, not a measurement of it. We do not have the daily creation-basket split to say how much of August’s figure is affected, and we will not estimate one.

For the year, August moved the needle without turning it. 2026 is still $1,677.9m in net outflow year-to-date after August recovered $3,539.1m of it; the hole was dug mainly in June ($4,509.7m out) and May ($2,406.0m out). Cumulative net inflow since January 2024 stands at $54,916m.

August closed at +25.078%, and September has started green

The settled August return, Bitstamp, 1 August open $62,817.96 to 31 August close $78,571.17, is +25.078%. That is final and it is a little larger than the +24.45% we published on 30 August, which measured through the 30th. August 2026 is the second-largest August in bitcoin’s recorded history behind August 2017’s +65.78%.

September has opened green: Bitstamp’s 1 September candle opened at $78,571.18 and was trading at $78,979.03 at 06:13 UTC, +0.519%. Binance BTCUSDT was at $79,014.01, up 1.313% on 24 hours, with a 24-hour range of $77,675.04 to $79,250.00. One day is one day.

The difficulty projection moved another 1.09 points overnight

Two days ago the projection for the next retarget was −0.9414%. Yesterday it was +0.3236%. This morning, at 66.766% of the epoch elapsed with 670 blocks remaining, mempool.space projects +1.4151% — a further +1.0915-point swing in roughly 24 hours, on top of the +1.79-point swing the day before.

This is now the third consecutive day the projection has moved by more than a point, and the direction has been consistently upward, which the hashrate explains: mempool.space’s three-day average reads 924.36 EH/s, a new high for the series, against 904.5 EH/s at the same hour yesterday. Current difficulty is 125,807,076,547,197.5 and the previous retarget was −1.3122%. Estimated retarget: 5 September 2026 at 20:22 UTC, block 965,664, from a tip of 964,994.

The consequence for the board is that D2 has gone from “failing within noise” to failing outright — its window is −1.50% to 0.00% and the projection is now 1.42 points above the top of it with a third of an epoch to run — while F1 is passing with a wider cushion than it had. Both still settle on the realised retarget, not the projection, and a projection that has moved 2.36 points in two days is not a settled number. Fees remain at 1 sat/vB for the economy, hour and half-hour tiers and 2 sat/vB for the fastest, the twelfth consecutive day at the floor since 21 August 2026.

Leverage: the coin count grew for once, and the term structure did not

Yesterday this desk published a decomposition showing that all $1.44bn of August’s open-interest growth on Binance was revaluation and the coin count actually fell 2,196 BTC. The first session of September did the opposite, and the identity closes again:

Binance BTCUSDT perpetual, openInterestHist daily snapshots, own pull. 1.0155971 × 1.0115108 = 1.0272874 against a published USD ratio of 1.0272874 — exact to seven decimal places. Rounding the two component ratios to six figures before multiplying returns 1.027288, which is why we print seven.

So 1,658 coins of genuinely new perpetual exposure arrived overnight, roughly $130m at the implied mark, alongside $95m of revaluation. It is one session; it does not reverse a month. It does mean G1 — perpetual open interest printing below 100,000 BTC on any September session — opened the month moving away from its bar, by 7.98%.

The term structure has not moved at all. Binance COIN-M at 06:13 UTC, each contract measured against its own index: BTCUSD_260925 at a 0.3606% premium over 24.07 days is 5.468% annualised; BTCUSD_261225 at 1.6026% over 115.07 days is 5.083%. Yesterday those read 5.349% and 5.084%. Two tenors three and a half months apart, 38.5 basis points apart and inverted, against US Treasury CMT par yields on 31 August of 3.91% at three months and 3.96% at four — an excess of +156bp and +151bp. Real, positive, thin.

And funding is still the expensive leg. The last 90 settlements — thirty days back to 2 August 2026 — mean 0.006793% per eight hours, which annualises to 7.439%, above both dated contracts. 24 of those 90 settled at the 0.01% cap and none settled negative. The 00:00 UTC print on 1 September was 0.008482%. G2, which needs the September contract above 8.00% annualised by settlement on 25 September, is 253 basis points short and moving 12bp a day in the right direction — which is not fast enough.

The crowd went net short into a green tape again

Binance’s global long/short account ratio printed 0.9948 on 1 September, below parity — more accounts short than long — against 1.0812 on 31 August and 1.1877 on both 29 and 30 August. The top-account ratio fell in step, 1.1580 to 1.0563. Retail positioning has now flipped short on a day bitcoin is up 1.3%. This desk has printed this pattern twice in the last week and it has resolved in favour of the tape both times; two is not a base rate, and we are naming it rather than dressing it as a signal.

Sentiment: thirteen days of Greed

The Crypto Fear & Greed Index printed 69 on 1 September, up 7 from 62. That is the thirteenth consecutive Greed reading, a run that began on 20 August 2026, the day after the index last printed Fear at 41 on 19 August. We state the run’s start date in the same sentence as its length because on Sunday we published a streak claim that turned out to be an artifact of how deep we pulled the series, and the mechanical fix we committed to is exactly this.

For the same reason we are not characterising this run as long or short relative to history in this piece. Grading it properly requires the full 3,130-reading series from 1 February 2018, which we pulled on Sunday to correct ourselves; we have not re-pulled it this morning and we are not going to quote a rank we have not recomputed today.

Two new markers

H1: bitcoin records a daily close at or above $80,318 — Strategy’s disclosed average for last week’s tranche — on or before 30 September 2026, settled on Bitstamp BTC/USD daily closes. It needs +1.695% from the 1 September print of $78,979.03.

H2: September 2026 finishes with a positive net ETF flow, settled on Farside Investors’ completed September column at the 30 September close. The bar is $0.0m. Both Septembers in the record cleared it — September 2024 at +$1,262.8m and September 2025 at +$3,511.0m — and 2026 has been negative in five of its eight completed months. We expect this one to pass, which is precisely why a failure would be the informative outcome.

Carried: D2 and F1 settle on the realised retarget, currently estimated at 5 September 2026, 20:22 UTC, 670 blocks out. G1 and G2 run to 30 and 25 September respectively. Gold remains barred from this desk’s comparisons for a tenth day, pending a source we are willing to stand behind.

Method: prices, funding, open interest, basis and on-chain figures in this article are pulled directly by Bitcoin Mastery at the timestamp stated — Bitstamp BTC/USD daily candles for closes and monthly returns, Binance BTCUSDT spot and perpetual for intraday and derivatives, Binance COIN-M quarterly contracts for basis, mempool.space for difficulty, hashrate and fees, alternative.me for the Fear & Greed series, Farside Investors’ all-data table for ETF flows and US Treasury CMT par yields for rates. Volume-weighted average prices are computed from Binance daily quote volume divided by base volume. Where a third-party figure is cited we name the source and its date; where two sources disagree we print both. Every streak or extreme figure is published with the first date of its series in the same sentence.

Disclaimer: This article is for informational purposes only and does not constitute investment advice. Cryptocurrencies are volatile and you can lose money. Nothing here is a recommendation to buy or sell any security, including MSTR, STRC or any exchange-traded fund. Do your own research and consult a licensed financial advisor before making investment decisions.