Markers Tuesday: the Difficulty Projection Flipped Sign, and R3 Was Right Three Days Late

This column writes its calls down before the events and grades them in public afterwards, whichever way they land. Today we have a marker that has quietly inverted, a dead marker that turned out to be correct on the substance and wrong only on the calendar, and an interim reading on a filing that did the opposite of what everyone expected. Each of the three teaches something different about how to write a falsifiable claim.

U2: the projection crossed zero

On Monday we wrote U2 down as follows: the settled difficulty retarget at block 963,648 prints at or above 0.00%. We set the bar deliberately at zero because the projection at the time was −0.28% and the day before it had been −1.63% — both negative, both implying the network was losing hashrate into a falling price. A marker set at zero was a bet that the direction of travel mattered more than the level.

This morning, same endpoint, same hour, third consecutive day: +0.49%. Our own mempool.space pull at 06:09 UTC reports a projected adjustment of +0.485%, with 67.7% of the period elapsed, 651 blocks remaining to height 963,648, and an estimated retarget around Saturday, August 22. The three-day series reads −1.63%, −0.28%, +0.49% — a swing of 2.12 percentage points in 48 hours, and, more to the point, a change of sign.

Source: own mempool.space difficulty-adjustment pulls, same endpoint and same hour on three consecutive days. Previous settled retarget: +0.99%.

Before anyone treats this as a trend, understand what the number actually is. The projection is not a forecast of future hashrate; it is an arithmetic extrapolation of the average block interval so far in the current period, and every additional block dilutes the influence of the blocks already mined. Early in a period the estimate is noisy enough to be nearly useless. At 67.7% elapsed it is starting to mean something, and by 90% it is close to final. So the honest reading is not “hashrate surged 2.12 points in two days” — it is “blocks have been arriving fast enough over the last two days to drag a two-thirds-complete average from below the line to above it.” mempool.space put the current average block interval at roughly 9 minutes 57 seconds against the ten-minute target. That is a thin margin, and 651 blocks is enough runway for it to close again.

The hashrate estimates cut slightly against the projection, which is the part worth sitting with. Our pull this morning reports a three-day average around 932 EH/s, against the 1,025 EH/s daily-average print we reported on Monday — the first reading above one zettahash on this board. Those two numbers are not in conflict; they are different windows over a series so noisy that a single lucky run of fast blocks moves the daily estimate by 10%. This is precisely why we grade U2 on the settled retarget rather than on any projection: the settled number is the only hashrate statistic in bitcoin that is not an estimate. It is a definition.

What has not changed: zero of the last fifteen blocks signal bit 4 for BIP-110 (own pull, tip height 962,997), and fees remain at 1–2 sat/vB. The signalling window closes at the same block that triggers the retarget. Whatever the difficulty prints on Saturday, roughly 100% of the hashrate that produced it voted no.

R3: right thesis, wrong deadline — and why that is a distinct grade

On August 8 we armed R3: Brent settles above $90 before Friday, August 14. Brent was near $82 at the time and the reasoning was that the Hormuz fee dispute would eventually break the wrong way. On August 14 we graded it FAIL, in public, with Brent at $88.52 — $1.48 short.

On Monday, August 17, Brent settled at $90.87. On Tuesday it traded to $91.59. The thesis was correct. The deadline was three days too tight.

We are not going to regrade it — a marker you regrade after the fact is not a marker, it is a memoir. R3 stays FAIL. But the failure mode here is genuinely different from the failure mode of, say, T1, where the price simply did not do the thing. R3 failed on timing, not on direction, and lumping the two together destroys information about which parts of our process actually work. So we are adding a third dimension to the board rather than a third grade: every marker from today carries an explicit note of whether its deadline is scheduled or arbitrary. T1’s deadline was scheduled — the weekly candle closes when it closes, and nothing about that date was our choice. R3’s deadline was arbitrary; we picked a Friday because Fridays are tidy. Attaching an arbitrary deadline to a geopolitical catalyst that has no schedule is not a test of the thesis, it is a coin flip on timing wearing a thesis as a costume.

This is a refinement of, not a retreat from, the date-anchoring rule we adopted after the SEC cancelled its August 14 meeting and handed this board its first VOID grades. Date-anchoring is still right: markers must not depend on an event happening as scheduled, because events get cancelled. But a date-anchored marker on an unscheduled catalyst needs its window sized to the catalyst, not to the calendar’s convenience. Had R3 read “before August 29” it would have graded PASS on the substance it was actually about. That is our error, and it cost us a real call.

U1 interim: the filing showed no coin sale at all

U1 reads: Strategy discloses a further bitcoin sale in a filing dated on or before Friday, August 21. The reasoning was that three consecutive weekly sales would convert a funding tactic into an operating policy. Monday’s 8-K arrived and disclosed no bitcoin purchases and no bitcoin sales. Holdings sit unchanged at 840,447 BTC. U1 is therefore not fired, with three days left on the clock; it grades Friday.

The more interesting thing is what the filing did contain, because it tested a rule we wrote down twenty-four hours ago. Monday’s lesson, drawn from the August 10 filing, was: when a filing shows both a sale and an issuance, the pair is the story — never report the sale alone. This week the pair is 3.46 million common shares sold for about $333.7 million against zero coins, with $132.2 million routed to STRC preferred repurchases, $52.4 million to STRC dividends and $149.1 million into a dollar reserve that now stands at $4.8 billion. A desk that had been reporting only the coin line would have written “Strategy stops selling bitcoin” and missed that the company simply pulled a different lever harder. The rule held on its first live test, which is more than most rules manage.

Note the market’s verdict: MSTR closed at $97.68, up 4.99% from Friday’s $93.04, on a day it announced it had issued a third of a billion dollars of stock. Dilution that buys optionality is being priced as good news, at least this week. Today’s Update takes the filing apart properly.

U3 got a lot closer, and nobody should be comforted

U3 reads: a daily close at or above $65,000 on or before Friday, August 28 — the morning of Chair Warsh’s Jackson Hole keynote. We wrote it on Monday when it sat about 2.2% away and described it as deliberately uncomfortable. After Monday’s 2.59% rally to a $64,532.10 close it is 0.73% away, with ten sessions to run. That is a large improvement in a marker’s odds produced by a single session, and it is worth naming the trap it sets: the temptation now is to talk about U3 as though it is nearly settled. It is not. Bitcoin has not closed a day above $65,000 since August 3, and the two macro events between here and the deadline — Wednesday’s minutes and next Wednesday’s core PCE — are both capable of removing 0.73% before lunch.

The board

Three new markers

• • X1 (deadline: scheduled). The FOMC minutes released at 2:00 p.m. ET on Wednesday, August 19 show that more than the three named dissenters judged an increase in the target range appropriate at the July meeting. This is a test of the single most misread feature of the minutes: dissents are cast by the twelve voting members, but the minutes report the views of all nineteen participants, and the participants’ section routinely reveals that the recorded vote understated the hawkish or dovish weight in the room. Grades Wednesday, on the text.
  • X2 (deadline: arbitrary, sized to the catalyst). Brent settles above $95 on or before Friday, August 28. Brent settled $90.87 on Monday and traded $91.59 on Tuesday; the June memorandum has now expired with no replacement. Per the R3 lesson above, the window is eleven days rather than three, and is sized to the diplomatic calendar rather than to a tidy Friday.
  • X3 (deadline: scheduled). At least one of the August 17 or August 18 sessions prints a positive net flow across US spot bitcoin ETFs, per settled tracker data, graded at Friday, August 21’s close. This is the direct test of whether Monday’s 2.59% rally was accompanied by the ETF bid or happened in spite of its absence — the question left open by last week’s −$389.7 million, the heaviest week since June.

What would change our mind

The bear case for everything above is a single sentence: Monday was a short squeeze on an oil headline and it will be given back. It is a good case. Bitcoin rallied 2.59% on a session in which the inflation impulse strengthened on two fronts and September hike odds should logically have risen; a market that rallies on hawkish news is either seeing something the macro tape is not, or it is mechanically covering. If Tuesday and Wednesday give back the gain, if the ETF prints come in negative for a fourth and fifth day, and if the difficulty projection slides back under zero as the last third of the period fills in, then the correct reading of Monday is “noise” and we will say so on Friday. The value of writing U2, U3 and X3 down in advance is that we will not get to choose the story afterwards.

What is the projected bitcoin difficulty adjustment for August 2026?

As of 06:09 UTC on August 18, 2026, mempool.space projects a +0.49% adjustment at block 963,648, with 67.7% of the period elapsed and 651 blocks remaining, for an estimated retarget around August 22. The same endpoint at the same hour showed −0.28% on August 17 and −1.63% on August 16.

Did Strategy sell bitcoin in its August 17, 2026 filing?

No. The filing covering August 10–16 disclosed no bitcoin purchases and no sales, leaving holdings at 840,447 BTC. The company instead sold about $333.7 million of common stock, directing $132.2 million to STRC repurchases, $52.4 million to STRC dividends and $149.1 million to its dollar reserve.

How far is bitcoin from $65,000?

After Monday’s daily close of $64,532.10 on Binance BTCUSDT, bitcoin sits 0.73% below $65,000. It has not recorded a daily close above that level since August 3, 2026.

Why does a failed marker still count if the thesis was right?

Because a marker with a deadline is a test of both the thesis and the timing, and regrading one after the fact removes the discipline that makes it useful. R3 stays graded FAIL. The correct response is to fix the process — sizing deadline windows to the catalyst rather than to the calendar — not to move the goalposts.

Disclaimer: This article is for informational purposes only and does not constitute investment advice. Cryptocurrencies and crypto-linked equities are volatile and you can lose money. Do your own research and consult a licensed financial advisor before making investment decisions.