Strategy (MSTR) · weekly ledger · as of 2026-07-20

The Gap Nobody Measures

Published 2026-09-06

Reported Bitcoin per share counts the coins. Common Equity Bitcoin Exposure counts the coins that common equity still owns after senior claims. The two are different quantities, and the difference is measurable per share, week by week, from filed figures.

The measured figures below come from the corrected MSTR weekly series and each one carries the basis it was struck on. The narrative sections are stubs: each states in one line what it will say, and the writing is held for the author.

Section 06 stays unwritten by the study's own rule, which holds findings until every findings slot returns verified. The slot table at the foot of this page shows which slots the series fills and which are still open, with the desk each open one routes to.

Basis

Series: gap-series-mstr-2026-09-06.csv, 70 ledger rows, reported dates 2025-01-06 to 2026-07-20.
Bitcoin price: 46 rows priced at the site close for the row's own date (the daily Bitcoin close the tracker serves, struck as the 00:00 UTC print of the following day), every one at zero lag. 24 back-period rows priced at the Yahoo BTC-USD daily close for the same date, labelled at the row, because the site close series opens 2025-08-10 and the public tier will not serve earlier prints. No price is ever carried forward.
Claims: Debt plus preferred less cash less STRC holdings, from the latest MSTR snapshot dated on or before each row. A step function across 7 distinct snapshots, never interpolated. The construction matches config/compute-drag.js.
Shares: Basic shares from the ledger's own derived column, which the sheet states in thousands. The series carries both the stated column and the corrected count so the factor of 1,000 is visible rather than silent.
Floor: None applied. The ratio form of Senior Claims % clamps to a 0 to 1 range; a per-share quantity does not, because clamping it would hide a net cash issuer.
USD figures: The per-share claim in dollars is claims divided by basic shares and uses no Bitcoin price at all. It is complete on all 70 rows and is unaffected by the price question above.

The chart that belongs here

Chart 12 is the hero of this page: the measured series over its full window, with the price basis printed on the face. It installs when the back-period amendment to the series lands on the main line, which is what fills the 24 early rows that carry no site close of their own.

One more thing has to move with it. The series file lives under a path the site rewrites to a 404, so a served chart cannot read it where it sits today. The chart and its data path install together or neither does.

Gap, latest row

79,980.1 sats

Per share, at 2026-07-20. Reported 222,535.0 sats against 142,554.9 sats of common equity exposure. Priced at the site close for that date, $65,185, zero lag, against claims from the 2026-06-30 snapshot.

Gap in dollars, latest row

$52.14

Per share, at 2026-07-20. Net senior claims of $19,767,777,000 over 379,165,000 basic shares. No Bitcoin price enters this figure.

Cumulative, this window

51,140.9 sats

Sum of the 69 row-to-row changes from 2025-01-06 to 2026-07-20, which equals the change in level across the same window. It is not the level, and the two are labelled apart everywhere on this page.

At a held Bitcoin price

72,409.8 sats

The latest row with claims converted at a fixed $72,000 rather than at the print of the day, computed here from the series' own claims and share columns. The window change on the same basis is 31,523.0 sats.

01The Question

Opens on the distinction the page exists to measure: reported Bitcoin per share is read as ownership, ownership is what is left after senior claims, and the difference has a size. Prose held for the author. The measured answer is the figure block above.

02What This Replaces

This study supersedes a prior claim published on this site.

In February 2026 this site published a directional accuracy comparison reporting CEBE at 62 percent and Bitcoin per share at 43.5 percent, with McNemar p below 0.01. That study used basic Bitcoin per share rather than fully diluted, which violates the comparison standard this site holds itself to: fully diluted is the market convention on the reported side, and CEBE uses basic shares with all non-cash senior claims in the numerator.

Re-run under the corrected standard, quarter-end to quarter-end across 89 transitions, the result was CEBE 49.4 percent against fully diluted Bitcoin per share 46.1 percent at p equal to 0.69. Not significant. A per-purchase reconstruction returned 45.6 percent against 43.7 percent at p equal to 0.617. The original finding does not replicate. The most likely explanation is that 62 percent was a favorable rolling window rather than a durable structural property.

The correction matters less than the diagnosis. Directional accuracy was the wrong test. Capital structure does not move inside a quarter. Both metrics therefore absorb the Bitcoin price the same way, and both collapse toward the same question, which is whether Bitcoin rose that week. Testing two metrics on a variable they share is not a test.

The framework never claimed to predict price direction. It claims something mechanical: that a specific class of senior claim appears in one metric and not the other. That claim is measurable directly, and this study measures it.

The three figure sets in this section are historical results from completed studies and are not exposed to the corrections that produced the series above. They are carried here as the brief writes them and still owe a citation check against the source workbooks before this page leaves draft. Slots V-02 through V-04.

03What Fully Diluted Bitcoin Per Share Gets Right

Concedes the convertible notes: if-converted treatment puts their shares in the denominator before the dilution arrives, so the convert stack is largely visible in the reported number, with two second order seams at moneyness crossings and at the cash repayment branch. Prose held for the author. Slots V-05 through V-07 are open; the series carries total debt and does not separate convert principal.

04The One Thing It Cannot See

States the mechanism: preferred stock that cannot convert produces no shares for a denominator and no reduction in coins for a numerator, so the claim expands while the reported metric does not move. Prose held for the author.

Measured	Value	Basis
Aggregate preferred face	$15,462,056,000	Snapshot MSTR_20260630, carried on every ledger row through 2026-07-20.
Debt, and cash netted against it	$6,753,703,000 / $2,447,982,000	Same snapshot. Debt is the total figure and is not separated into converts here.
Preferred share of net senior claims	78.22%	At 2026-07-20, against 16.52% at 2025-03-31, the first row in this window carrying a preferred balance. Shares of a net figure, so the components can exceed 100 percent before cash nets.
Senior Claims %	35.94%	Claims in Bitcoin over coins held, at 2026-07-20, priced at that date's site close. 15.84% at the first row of the window, 22.43% at 2025-03-31.

05Method

States the event set, the per-event measurement, the share count convention, the claims measurement, and the normalization, in the brief's order. Prose held for the author. The construction below is the series as computed and is stated so it can be checked.

claims_usd(t)  = debt + preferred - cash - strc_holdings, from the latest MSTR
                 snapshot dated on or before t. Step function, never interpolated.
claims_btc(t)  = claims_usd(t) / btc_price(t) + btc_claim_static
BPS_sats(t)    = cumulative_btc(t) / basic_shares(t) * 1e8
CEBE_sats(t)   = (cumulative_btc(t) - claims_btc(t)) / basic_shares(t) * 1e8
gap_sats(t)    = BPS_sats - CEBE_sats, which reduces to claims_btc / basic_shares * 1e8
gap_usd(t)     = claims_usd(t) / basic_shares(t), which needs no Bitcoin price at all
gap_change     = the difference between consecutive ledger rows

Series property	Value	Basis
Rows, and the window	70	Reported dates 2025-01-06 to 2026-07-20, ledger counter 50 through 119. One reported date, 2026-07-06, carries two ledger rows.
Rows priced at the site close	46	Each at the print for its own date, maximum lag 0 days.
Rows priced on the labelled back-period close	24	Yahoo BTC-USD daily close for the same date, labelled per row, superseded when the site close series can serve the range.
Distinct snapshots behind the claims figures	7	2024-12-31 through 2026-06-30, each carried forward until the next one files.
Range, sats	27,618.1 to 84,319.6	Low at 2025-01-21 on a back-period price, high at 2026-07-13 on a site close.
Range, dollars	$28.59 to $53.53	The low holds across 2025-02-24, 2025-03-17 and 2025-03-24; the high holds every week from 2025-10-13 to 2025-12-01. Both are plateaus, because the figure moves only when claims or the share count move.
Largest single-row change	+13,766.5 sats	Into 2025-03-31. The largest contraction is 6,185.0 sats into 2026-03-16, on a rising Bitcoin price against fiat-fixed claims.

06Findings

Reports the terminal number, the shape of the series before and after the first preferred issuance, the contribution by series introduction and by funding type, and the Senior Claims % progression. Held unwritten by the study's own blocking rule, which keeps this section closed until every findings slot returns verified. The measured shape below is stated as a measurement and is not a finding.

What the series shows across the eight rows that precede the first preferred balance: the gap runs 28,839.2 sats to 32,670.9 sats, and in dollars $29.44 to $28.59. Whether that reads as flat is the directional judgment the study routes to the company anchor rather than settling from the arithmetic.

Two of the brief's directional claims cannot be tested from this series at all. Contribution by series introduction needs the preferred stack split by instrument, and contribution by funding type needs the deals ledger. Neither is in the file this page reads.

07Limitations

Sets out price sensitivity in the sats series, convert moneyness discontinuities, intra-quarter share counts, liquidation preference treatment, path dependence, the measurement-not-prediction boundary, the single issuer scope, and the independence statement. Prose held for the author. The mechanical limits of the series itself are stated in the basis block and in section 05.

08What Would Falsify This

Closes the ring: the study predicts that the measured quantity accumulates while the preferred stack expands, plateaus when it stops, and reverses on retirement, and it names the three outcomes that would show the mechanism wrong. Prose held for the author.

09Slot inventory

Every figure slot the brief opened, with what the corrected series fills and what it does not. A slot the series cannot reach is named as open and routed, never estimated.

Slot	Quantity	Value	Basis, or where it is open
V-01	Headline, cumulative across the window	51,140.9 sats	Sum of 69 row-to-row changes, 2025-01-06 to 2026-07-20. The level at the last row is 79,980.1 sats and is a different quantity.
V-02	Prior published 62 / 43.5, p below 0.01	carried	February 2026 study workbook, per section 02. Citation check outstanding.
V-03	Quarter to quarter 49.4 / 46.1, p 0.69, n 89	carried	Corrected study workbook, per section 02. Citation check outstanding.
V-04	Per purchase 45.6 / 43.7, p 0.617	carried	Run 2 workbook, per section 02. Citation check outstanding.
V-05	Aggregate convert principal	not yet verified	The series carries total debt only. Routes to the instruments tab through the company anchor.
V-06	Count of outstanding convert series	not yet verified	Not in the series. Company anchor.
V-07	Convert moneyness crossing dates	not yet verified	Needs strikes against a daily price series. Company anchor.
V-08	Preferred stack by series face	not yet verified	The series carries the aggregate only. Instruments tab.
V-09	Aggregate preferred face	$15,462,056,000	Snapshot MSTR_20260630.
V-10	Senior Claims %, latest row	35.94%	Claims in Bitcoin over coins held at 2026-07-20, site close $65,185.
V-11	Event count	70 rows	Ledger rows in this window, counter 50 to 119. The brief's event set runs from the first purchase and is wider than this window.
V-12	Fully diluted rows flagged above 5 percent deviation	not yet verified	A balance sheet spine question. Not in the series.
V-13	Rows on a carried-forward input	70 of 70	Claims are carried from the latest snapshot on or before each row, 7 snapshots across 70 rows. The brief's share-count carry-forward question is separate and remains open.
V-14	Preferred claim basis	anchor ruling	Stated face against liquidation preference, including the floating STRK preference. Not a lookup.
V-15	Convert claim basis	anchor ruling	Face against carrying value. Not a lookup.
V-16	Terminal at spot	79,980.1 sats	Level at 2026-07-20, priced at that date's own print.
V-17	Terminal at the held price	72,409.8 sats	Claims converted at $72,000, computed here from the series' claims and share columns. Window change 31,523.0 sats.
V-18	Contribution by series introduction	not yet verified	Needs the preferred stack split by instrument.
V-19	Contribution by funding type	not yet verified	Needs the deals ledger. Not in the series.
V-20	Senior Claims % at the first preferred row	22.43%	At 2025-03-31, the first row in this window carrying a preferred balance. The issuance date itself is the anchor's figure.
V-21	Composition shift	16.52% to 78.22%	Preferred as a share of net senior claims, 2025-03-31 against 2026-07-20. Debt moves 84.10% to 34.17%. Converts are inside the debt figure and are not separable here.
D-01	Shape before the first preferred balance	measured	8 rows, 28,839.2 to 32,670.9 sats, $29.44 to $28.59. The reading is the anchor's.
D-02	Step at each series introduction	not testable here	Needs V-18.
D-03	Common and cash funded contribute near zero	not testable here	Needs V-19. Its failure blocks publication of the findings section.

Every figure on this page is struck on one series and one basis, both named above. Figures move with each week's filings; the page carries an as-of date for that reason. Nothing here is investment advice, and this is a measurement rather than a forecast.