CEBE Under the Hood: How the Framework Matured as the Preferred Era Arrived

The CEBE framework I published in February made a structural argument and a statistical claim. The structural argument was that fully diluted Bitcoin Per Share, the market convention for measuring BTCTC dilution, cannot see what senior claims do to common equity ownership. That argument was correct then and is more correct now. The statistical claim, a 62% / 43.5% directional accuracy result with a McNemar significance test, used basic BPS rather than fully diluted BPS as the comparator. Basic BPS is not the metric the market actually uses. Comparing CEBE against basic BPS produced a result that overstated the framework's edge against real-world practice.

This piece refines the comparator. Every measurement that follows uses fully diluted BPS, which folds convertible debt into the share count via the if-converted method whenever those converts are dilutive. That is the standard published in 10-Q and 10-K filings, the standard analysts cite, and the standard against which any new metric has to earn its place. Against fully diluted BPS, the original structural argument doesn't weaken. It sharpens, because the part of FD BPS that works (convertible debt treatment) gets credited to it cleanly, and the part it structurally cannot handle (the rest of the senior claims stack) becomes the unambiguous space where CEBE contributes.

The maturation I'm publishing is a move from a directional accuracy claim against the wrong comparator to a structural divergence claim against the right one. The original insight that senior claims matter to common equity ownership survives intact. What changes is the evidence I bring for it. Instead of a single statistical headline, the case rests on 58 weeks of verified MSTR 8-K filings covering the entire preferred era, a cumulative gap expansion that FD BPS structurally cannot register, and a price-normalized analysis demonstrating that preferred events are the sole source of that gap expansion. Non-event weeks compress the gap. The finding is binary and threshold-insensitive.

What follows is one short methodology section disclosing how the analysis was constructed, then the structural argument itself, then the data, then a single worked example from the most recent week of MSTR activity, then the synthesis. The framework's foundation is stronger, not weaker, for the comparator change.

Three disclosures govern the analysis that follows.

The CEBE measurement and the FD BPS comparator use different share count denominators. CEBE uses basic shares outstanding. FD BPS uses fully diluted shares, which fold in convertible debt via the if-converted method when those converts are dilutive. The asymmetry is by design, not oversight, and the reasoning is unpacked in Sections 3 and 4. Briefly: each metric uses the share count appropriate to what it measures, and the divergence between the two share counts is itself part of what the framework is built to surface.

Two BTC price references appear in the analysis, used for different purposes. The cumulative gap chart in Section 6 uses weekly 8-K implied market prices (BTC reserve value divided by cumulative holdings) across 58 weekly snapshots. The event analysis in Section 7 normalizes to a constant $77,500 BTC to isolate capital structure changes from price movement. The two serve different purposes. Longitudinal measurement requires the gap as investors experienced it. Event analysis requires the structural signal stripped of price noise.

The gap series joins two data sources. Weekly data from verified 8-K filings provides BTC holdings, share counts (basic and assumed diluted), preferred face values, and implied BTC price. Quarterly data from 10-Q and 10-K balance sheets provides debt and cash via a step function: hold the most recent quarter-end value until the next quarter replaces it. Debt and cash do not appear in weekly 8-Ks. The step function is not interpolation. It reflects when these values are actually reported. Every other input is real weekly data from the specific 8-K filing for that row, with a source URL tracing to the SEC Archives path.

CEBE and FD BPS differ in one foundational choice: which standard the capital structure is measured against.

A Bitcoin treasury company succeeds or fails on whether it accretes Bitcoin exposure to the common equity holder. That is the terminal measure. Every other measure is intermediate. Dollar value of holdings, mNAV, market capitalization, share price, all of them are derivatives of how much Bitcoin a common share actually owns once senior claims have taken their cut. Measuring a Bitcoin treasury strategy in any unit other than Bitcoin produces a metric that can move favorably even as the underlying outcome deteriorates. The strategy is denominated in Bitcoin. The evaluation frame has to be denominated in Bitcoin.

CEBE puts the entire capital structure on the Bitcoin standard. Every senior claim, regardless of native denomination, gets translated to BTC at the prevailing reference price at every snapshot. Cash, the offset to the claims stack, also gets translated. The basic-shares denominator stays in shares because shares are what common equity holders actually own, but the entire claims numerator is converted to the standard the strategy itself is denominated in.

FD BPS lives on a hybrid standard. BTC holdings get measured in BTC. Fully diluted shares get measured in shares, including the convertible debt component folded in via the if-converted method. Preferred stock, non-convertible debt, and cash sit outside the calculation entirely. The Bitcoin standard governs two components of FD BPS and no standard governs the rest.

The maturation of CEBE depends on giving FD BPS clean credit for what it does on the Bitcoin standard before showing the categories where it has no standard at all. Convertible debt is what FD BPS handles cleanly. The if-converted method is the mechanism.

When a company issues a convertible bond, the bond carries a contractual right to convert into common shares at a defined ratio. If those converts are dilutive at the reporting date, the if-converted method folds the underlying share count into the diluted denominator. The convertible bond simultaneously appears as debt on the balance sheet and as potential shares in the EPS footnote. FD BPS picks up the share-side effect through the denominator. CEBE picks up the claim-side effect through the numerator. Both metrics see the same instrument from different angles, and both metrics adjust as the instrument's status changes.

Strategy's convertible debt stack is the cleanest place to see this work. The 2027 notes, the 2028 notes, the 2029 notes, the 2030 notes, the 2031 notes, the 2032 notes. Each carries a strike and a conversion ratio. Each enters the FD denominator when dilutive. Each enters the CEBE numerator as a fiat-denominated senior claim. When BTC rises and the converts move further into the money, FD BPS reflects more potential shares. When CEBE recalculates against current BTC price, the dollar claim compresses in BTC terms. Two different mechanics capturing the same underlying reality, with both metrics registering it.

This is the part of the framework that gets understated in the discourse. CEBE does not displace FD BPS on convertible debt. Both metrics handle converts. The market-standard practice of citing FD BPS for converts is correct, and CEBE's contribution on converts is supplementary rather than corrective.

The structural divergence shows up in the rest of the senior claims stack.

The senior claims stack contains four categories of items that rank above common equity. Convertible debt. Non-convertible debt. Preferred stock. Cash, which offsets the other three. FD BPS handles the first category through the if-converted method, as Section 3 established. The other three categories have no path into the FD denominator, and they have no representation anywhere else in the FD BPS calculation either.

Preferred stock has no conversion mechanism into common shares in any normal sense. STRK, STRF, STRD, STRC, and STRE are perpetual instruments with dividend obligations, liquidation preferences, and seniority ranks. They are not convertible bonds. They do not trigger if-converted treatment in the EPS footnote. They sit on the balance sheet as senior claims and they stay there. The if-converted method has nothing to fold in. The treasury stock method has nothing to net out. The fully diluted share count moves not at all when a preferred series gets issued, modified, or expanded.

Non-convertible debt has the same problem from a different angle. A straight bond, a term loan, a credit facility, none of them carry a conversion right into common shares. None of them touch the FD denominator. They sit on the balance sheet as senior claims, ranking above common equity in liquidation, and FD BPS records the BTC the company holds as if those obligations were not there.

Cash is the offset to the senior claims stack. Cash on the balance sheet can be used to service debt, pay preferred dividends, retire claims. CEBE nets cash against the claims stack at every snapshot because the net senior claim is what actually competes with common equity for the BTC reserve. FD BPS has no mechanism to net cash against anything, because there is nothing on the FD BPS side of the calculation that cash would offset.

The result is a metric that handles convertible debt cleanly and handles nothing else in the senior claims stack. A preferred issuance can grow the senior claims stack by billions of dollars. A non-convertible debt drawdown can do the same. A cash burn can erase the offset that previously cushioned common equity against those claims. FD BPS will print the same number it would have printed if none of those changes had occurred. The Bitcoin per share calculation produces a quantity of sats per fully diluted share that treats the modified balance sheet as identical to the prior one for every category of senior claim except convertible debt.

CEBE handles all four categories because CEBE's numerator is built to handle them. The framework subtracts net senior claims (debt plus preferred minus cash) from total BTC before dividing by the share count. A preferred issuance increases senior claims. The senior claims convert to BTC at the reference price. The BTC available to common equity falls. CEBE falls. The same mechanic applies to non-convertible debt drawdowns and to cash depletion. The metric registers what happened in each case.

This is the unambiguous space where CEBE contributes. Not a marginal improvement on FD BPS. Not a refinement at the edges. A structural gap in what FD BPS can see across three of the four categories of senior claim, filled by a metric built specifically to see them.

The next section walks through the wrapper that has grown outside the Bitcoin standard.

Strategy's preferred stack as of April 20, 2026 carries five series with a combined face value of approximately $13.39B. STRC at $8.54B. STRK at $1.40B. STRD at $1.40B. STRF at $1.26B. STRE at $0.81B (EUR-denominated, converted at the prevailing EUR/USD rate).

The stack did not arrive whole. STRK debuted in February 2025. STRF followed in March 2025. STRD in June 2025. STRC in late July 2025. STRE in November 2025. Each issuance added a layer to the wrapper. Each layer carried its own dividend rate, its own payment mechanism (cash, shares, or both), its own seniority rank, and in the case of STRE, its own currency denomination.

STRC is the layer that reshaped the stack. As of the terminal snapshot, STRC alone accounts for 64% of the aggregate preferred face. The July 2025 debut at $2.80B has grown to $8.54B through sustained ATM issuance. STRC's variable rate adjusts monthly. STRC pays cash only. STRC sits second in the preferred seniority hierarchy behind STRF.

The wrapper is not a single instrument. It is an evolving multi-series capital structure that grows independently of any FD BPS denominator change. Every dollar of preferred face that gets added is a dollar of senior claim that FD BPS does not see and CEBE does. The cumulative gap between the two metrics across the preferred era is the cumulative effect of the wrapper's growth on common equity ownership.

The chart in Section 6 traces that growth across 58 weeks.

Every week since STRK debuted in February 2025, I have tracked two numbers against each other. FD BPS, which is the industry-standard Bitcoin per share metric using fully diluted shares. And CEBE, which uses basic shares in the denominator and subtracts senior claims in the numerator before dividing. For most of 2025 the two numbers moved roughly in parallel. Over the last four months, that parallel broke.

The first chart shows the cumulative gap between FD BPS and CEBE across all 58 weeks of the preferred era. It starts at 4,533 sats per share on January 27, 2025. It ends at 51,974 sats per share on April 20, 2026. That is a 47,442 sat expansion in what FD BPS structurally cannot capture.

The chart uses weekly 8-K implied market BTC prices, not normalized, because it shows the gap as it actually unfolded. The event analysis in Section 07 normalizes to a constant $77,500 BTC to isolate the structural signal from price movement.

The chart reads cleanly. For most of 2025 the gap oscillated between 5,000 and 25,000 sats as STRK, STRF, and STRD came online and Bitcoin price chop moved the fiat-denominated senior claims in BTC terms. Starting in late January 2026 the line bends upward and does not return. By April 20 the gap has roughly doubled from its Q4 2025 range.

At that terminal point, 51,974 sats per share translates to $40 per share at the April 20 8-K reference price of $77,453. Across Strategy's 349.3 million basic shares outstanding (as of the April 20, 2026 8-K), the aggregate works out to approximately $14.1 billion of common-equity exposure that FD BPS does not account for. That is not a theoretical gap. It is a measurable economic position sitting outside the metric the market quotes every day.

A single line chart cannot prove what drove the expansion. The gap could have widened because Bitcoin fell, or because share count grew, or because senior claims grew. To distinguish those, I need to show what was happening inside the capital structure across the same 58 weeks.

The second chart plots each of Strategy's five preferred series as an independent line on a shared y-axis. STRK and STRF debut in Q1 2025. STRD joins at the end of Q2. STRC debuts in late Q3. STRE debuts in Q4 2025 at EUR 775 million, denominated in euros and shown here in USD at the prevailing rate.

Four of the five lines have the same structural shape. Each grows during its issuance phase, then plateaus. STRK plateaus at $1.40 billion. STRF plateaus at $1.26 billion. STRD matches STRK at $1.40 billion. STRE debuts and holds flat at $0.81 billion. From Q4 2025 onward these four series are horizontal lines across the chart. No further issuance.

The STRC line does something different. It debuts in late Q3 2025 near $2.8 billion, holds flat through year-end, and then ramps through Q1 2026 and into Q2. In the final fourteen weeks of the dataset alone, STRC grew by over $5 billion. None of the other four series moved during that span.

At $8.54 billion, STRC accounts for 64% of the aggregate preferred stack. When Strategy grew the wrapper through Q1 2026, it grew STRC. When I look at Chart A and see the gap line accelerating through that same window, the causal explanation is sitting directly below in Chart B.

FD BPS has no mechanism to register that any of it happened. CEBE registers every dollar of it, in BTC terms, every week. The gap between the two metrics is the measurement of what the wrapper built and where FD BPS went blind.

The week ending April 20, 2026 sits inside the chart's pattern as the clearest instance of the structural argument in the dataset.

Strategy disclosed via 8-K the acquisition of 34,164 BTC during the week. The purchase was funded through $2.18B of STRC ATM activity and $366M of MSTR common stock. Both CEBE and FD BPS rose. Both increases were accretive at the per-share level. The week's contribution to the normalized cumulative gap was +7,053 sats per share, the third-largest single-week expansion in the entire 58-week dataset.

I define a preferred-event week as a week in which a new preferred series debuted or an existing series expanded materially through ATM activity exceeding 0.5% of total preferred face outstanding at the start of that week. Rate modifications are excluded because they affect wrapper cost, not claims size. The gap measures claims size.

Across 58 weeks of verified, price-normalized data, the finding is binary. The cumulative gap between FD BPS and CEBE expands only during preferred-event weeks. During non-event weeks, the gap compresses. Preferred events are the sole source of gap expansion. This holds whether the materiality threshold is set at any amount above zero, at 0.5%, or at 1.0% of preferred face. The result is threshold-insensitive.

The evidence goes beyond the binary.

Event weeks expand the gap 76% of the time. Non-event weeks compress it 75% of the time. The directional bias flips with preferred activity. Not 100%, because BTC accumulation timing and share dilution create noise on any given week, but the base rate is clear in both directions.

The correlation between preferred issuance size and gap expansion is r = 0.94 with p < 0.0001. The top five gap expansion weeks are the top five issuance weeks. The relationship is mechanical. Larger preferred face directly increases net senior claims in BTC terms, which widens the gap FD BPS cannot see. This is not a statistical pattern. It is arithmetic.

IPO debut weeks expand the gap 100% of the time at an average of +6,564 sats. ATM expansion weeks expand it 71% of the time at +616 sats. Different profiles. IPOs land as large discrete events. ATM issuance accumulates week over week. Both mechanisms contribute materially across the full period.

April 20 at +7,053 sats is one observation inside that pattern, behaving exactly the way the framework predicts a large preferred-event week should behave: $2.18B of STRC face entered the structure, net senior claims in BTC terms grew, and the gap expanded by the third-largest increment in the dataset.

A reader who saw only the @cebetracker panel from that week saw a single transaction with two accretive metrics and a large gap expansion. A reader who sees the panel through the framework's verified dataset sees that same transaction as one observation in a pattern that has been running for 15 months. The panel and the dataset describe the same structural reality at different scales.

Scale is one dimension the framework operates across. Currency denomination is another. STRE is Strategy's EUR-denominated preferred series and the only FX-denominated claim in the current capital structure. The face value at issuance was set in EUR. The dividend obligations are paid in EUR. The liquidation preference is denominated in EUR. To enter the CEBE calculation, the EUR face value must be converted to USD at the prevailing EUR/USD rate, then converted to BTC at the prevailing BTC/USD rate.

That two-step conversion introduces a second variable. A USD-denominated claim moves only with BTC price. A EUR-denominated claim moves with BTC price and with EUR/USD. If BTC rises 10% in USD terms while the EUR strengthens 5% against the dollar, the BTC-equivalent of a USD claim falls roughly 9%, while the BTC-equivalent of a EUR claim falls roughly 5%. The compression behavior diverges.

The divergence is not large at any single moment for a $0.81B series in a $13.39B stack. The mechanism still matters, because the framework is built to absorb it without modification. CEBE recalculates STRE's BTC-equivalent claim at every snapshot using the prevailing FX rate at that snapshot. The compounding dimension is captured automatically. FD BPS sees none of this because FD BPS does not see STRE at all.

The mechanism STRE illustrates applies across the BTCTC universe. Metaplanet's MERCURY bonds are JPY-denominated, with the yen's ongoing weakness against the dollar adding a second compression layer that runs in the same direction as BTC appreciation. H100 Group's convertible bonds are SEK-denominated, with similar Nordic-currency dynamics. Capital B's OCAs occupy a separate category entirely: the convertibles are EUR-denominated for accounting purposes, with the redemption mechanics tied to BTC. The BTC-indexed redemption produces a static drag profile across BTC prices rather than the compression-as-BTC-rises behavior that fiat-denominated claims exhibit. That distinction was confirmed by Capital B board director Alexandre Laizet in direct response to the framework's earlier classification work, and it is the cleanest illustration in the universe of why claim denomination is a structural feature of CEBE and not a footnote to it.

The point is not to compare those companies to Strategy on capital structure quality. The point is that claim denomination is a category, not an exception, and any framework built to measure common equity Bitcoin exposure across the universe of BTCTCs has to absorb FX and BTC-indexed claims as first-class variables. CEBE does, while FD BPS does not.

Ownership acceleration is what the framework measures.

Common equity ownership of a BTCTC's Bitcoin holdings is not static. It moves with BTC price, with the size of the senior claims stack, with FX rates on non-USD claims, and with share count. Four variables, each independent, each capable of moving at its own pace.

CEBE tracks all four because the Bitcoin standard established in Section 03 requires it. BTC holdings in the numerator. Senior claims in the numerator, converted from any native denomination to BTC at every snapshot. Cash in the numerator as the offset, also converted to BTC at every snapshot. Basic shares outstanding in the denominator. The composite recalculates at every snapshot because the composite is what determines common equity ownership at that snapshot.

FD BPS tracks two of the four. BTC holdings move through the numerator. Fully diluted shares move through the denominator, capturing convertible debt via if-converted. Senior claims outside the convert category, and FX on non-USD claims, are not in the calculation at all. The two variables FD BPS tracks move with their own dynamics. The two variables FD BPS does not track move with theirs. Across a 58-week dataset, the tracked and untracked variables drift apart.

The cumulative gap chart in Section 6 is the visual record of that drift. The cumulative expansion is not a measurement curiosity. It is the accumulated divergence between a metric that puts the full capital structure on the Bitcoin standard and a metric that puts only part of it there. Every week in the chart is one increment of the divergence. Price-normalized analysis confirms that preferred-event weeks are the sole source of that divergence. Non-event weeks compress the gap. The driver is not ambiguous.

A snapshot of ownership at a moment in time is what FD BPS provides for the share-and-converts portion of the capital structure. The rate at which ownership changes as the full cap stack evolves is what CEBE provides. That rate of change is the framework's contribution.

The rate of change is measurable. What remains is whether the measurement holds.

Across 58 weeks of price-normalized data, the evidence converges on a structural claim with two testable dimensions. Both are staked against defined thresholds on a defined timeline.

The first claim is the binary finding. Preferred-event weeks are the sole source of cumulative gap expansion between FD BPS and CEBE. Non-event weeks compress the gap. The finding holds at every materiality threshold tested. Event weeks account for 135% of total normalized gap expansion; the excess over 100% reflects non-event compression that event weeks must overcome to produce net growth. The mean normalized gap change is +1,518 sats for event weeks and -542 sats for non-event weeks. The directions do not overlap.

The falsifiability stake on the binary finding: non-event weeks should remain compressive on average through Q2 2026 and Q3 2026, computed at each quarter-end using the same price-normalized methodology and the same 0.5% materiality threshold applied to the full accumulated dataset at that point. A result where non-event weeks show positive average gap expansion through both Q2 and Q3 would constitute a falsification. Two checkpoints rather than one, because a single quarter could deviate for reasons unrelated to the underlying structure. Two consecutive quarters outside the pattern is a signal, not a fluctuation.

The second claim is the magnitude relationship. The correlation between preferred issuance size and normalized gap expansion is r = 0.94 across the current 33 event weeks. The top five gap expansion weeks are the top five issuance weeks. The relationship is mechanical: preferred face enters the claims stack in dollars, converts to BTC at the reference price, and widens the gap proportionally. Larger issuance, larger gap expansion.

The falsifiability stake on the magnitude claim: the Pearson correlation should remain above r = 0.85 through both Q2 and Q3 2026 quarter-end checkpoints. The threshold is set below the current 0.94 to accommodate a structural feature of the near-term dataset: Strategy's preferred issuance is currently concentrated in a single series (STRC ATM), which produces less variance in issuance size than a period with multiple IPOs and ATM programs running simultaneously. Reduced variance in the independent variable compresses correlation coefficients even when the underlying linear relationship is unchanged. Setting the threshold at 0.85 accounts for that compression without surrendering the claim. A result where r falls below 0.85 through both Q2 and Q3 would indicate the magnitude relationship has weakened beyond what reduced variance alone can explain, and would require the framework to re-examine the dose-response claim.

The data to test both claims will be available without dependency on quarterly filings. Weekly 8-K snapshots continue to populate the dataset. Preferred events continue to be disclosed as they occur. Both dimensions can be tested at any point during Q2 and Q3 2026 from data that is already in the publication pipeline.

If both claims hold, the framework's structural prediction stands. If one breaks, the framework narrows. If both break, the distinctive-driver claim needs re-examination. The maturation that this piece publishes is durable to any of those outcomes. The structural argument that FD BPS misses three of the four categories of senior claim survives regardless. What collapses if the pattern breaks is not the framework. What collapses is the specific quantitative claim that preferred events are the exclusive driver of gap expansion, at the consistency and magnitude currently observed.

That is the position I am willing to be wrong about in public, on a defined timeline, with the tests specified before the data arrives.