How this site measures Bitcoin treasury companies, and why
Every figure on this site rests on a document that can be read. This is the whole corpus, in three parts: the construction, the studies that test it, and the surfaces where the measured figures are published. Each entry is quoted from its own page.
What the figures are, and the rules they are built under. These change by ratified ruling, and the changelog below records every change.
Each study fixes its criteria and its data before it sees a result, and reports the result it got. A study that fails is published as a failure and the surface it would have changed does not change.
Where the measured figures are published, and what each surface is for. Every figure traces to a primary filing.
In a single week of June 2026, two of the sector's largest issuers publicly disagreed about what mNAV means, a third shipped a per share metric whose construction matched no published standard, and the chief executive of one major treasury company spent ten minutes on stage answering a definition question from the chief executive of another. None of these people were being careless. They were using the same words for different mathematics, because the sector has labels and conventions but no standard.
This page is the standard this site holds itself to. Every number published on cebetracker.io is produced by the rules below. The rules are stated so they can be checked, attacked, and improved. Where a rule involves a judgment call, the judgment is named and defended rather than absorbed silently into the output.
One sentence of positioning before the rules. CEBE is a measurement, not a verdict. It is an instrument panel, not a price target. It tells you what common shareholders own. It does not tell you what they should pay for it.
Everything on this site lives in one of four layers, and the boundaries between them are load bearing.
Measurement is the CEBE formula applied to verified filing data. It contains no assumptions about the future. Every input traces to a primary source, and every figure carries a VERIFIED or EST flag.
Mechanics is the modeler. It is deterministic arithmetic that computes whatever assumptions a user feeds it. The mechanics belong to the site. The assumptions belong to whoever sets the sliders.
Scenarios are projections built from measurements plus stated assumptions. A scenario belongs to the person who states the assumptions. This site publishes mechanics that anyone can use to build scenarios. It does not publish return predictions of its own.
Valuation is the structural risk layer, Claims Grade and Adjusted Claims %, which characterizes the quality and behavior of the claims the measurement counts. It grades instruments. It does not grade futures.
When you see a return projection described as a CEBE model, what you are seeing is a scenario built on CEBE measurements. The measurements are this site's responsibility. The scenario is its author's. The rules below are grouped under these layers, so you always know whether you are reading a measurement rule or a valuation judgment.
CEBE = (Total BTC - Net Senior Claims in BTC) / Basic Shares Outstanding x 100,000,000
Claims in BTC = (Debt + Preferred - Cash - STRC Holdings) / BTC Price + BTC-Indexed Claims
Net Senior Claims in BTC = max(0, Claims in BTC)
The construction is one sentence. Convert the balance sheet to BTC and net the results. Fiat claims, debt plus preferred, convert at the snapshot's own BTC price. Cash nets against the total, and STRC holdings net the same way as a cash equivalent. BTC-indexed claims enter as BTC directly and never touch the price. The floor at zero applies to the netted total, not to any single bucket, so an issuer holding more cash than fiat claims uses the surplus to offset its BTC-indexed claims.
Cash nets against senior claims because it is fungible until it is paid. A dollar of treasury cash can retire a dollar of debt, and until it does, that dollar of claim is already provisioned for. The netting reports the balance sheet as filed rather than an assumed transaction, so nothing is repaid and no timing is projected. State consistency requires the same treatment. Carrying cash gross while carrying claims at face would describe a company that holds the money and owes the full claim at once, a position no filing reports.
Declared but unpaid preferred dividends do not enter the claims line as a separate item. The claim standing ahead of common is the accrued liquidation preference, and where a preference accrues or compounds, the accrued figure is the carrying value. A declared dividend is settlement in transit: the claim converting to cash, not a claim added on top of one already counted. Carrying both the accrued preference and the payable would count the same obligation twice. The disclosed dividends-payable figure is retained as a non-netting disclosure field. Today no tracked series has a suspended dividend, so accrued equals face everywhere; the case where they diverge is scoped in the arrears work and will be carried through the preference itself.
The claims line has a stated perimeter, and accrued preferred dividends are tracked at the instrument level, with declared and unpaid amounts carried inside the accrued liquidation preference as described above. Accrued interest on the convertible stack is excluded, immaterial at coupons running from 0.00 to 2.25 percent. Deferred tax liabilities and CAMT are excluded, and the reasoning sits with the wider question of which liabilities a residual measure should refuse, which this page will take up separately.
See the formula in motion on the framework page and the learn course.
The unit is satoshis per basic share.
Satoshis are a convention, chosen rather than derived. The strategies under measurement are stated in Bitcoin accumulation, so per share Bitcoin is the unit the goal is already written in, and measuring in it keeps the metric and the objective in one denomination. That is a reason for the choice, not a claim that it is the only valid frame. A dollar frame answers different questions and answers them well, which is why dollar renderings appear throughout this site alongside the sats figures. The numéraire is stated here so a reader always knows which one they are reading.
Market capitalization in the multiples below is struck on basic shares outstanding, the same fact-based denominator the formula uses. This matters because market cap can be built on basic, diluted, or proximity weighted share counts, and an unstated choice hides the same modeling the rest of this page works to keep out of the measurement. Basic on both sides keeps the multiple in one coherent state.
Three construction choices define the metric, and each follows from one principle. The principles are stated first because they decide every hard case that follows.
Three principles decide every hard case in the formula. They are stated before the rules because the rules follow from them. Each principle names the part of the construction it governs.
Principle oneClaims sit at their legal accrued value until an actual transaction changes them. In both directions.
A convertible note that is deep in the money stays at face value in the claims line. It is never marked up to its conversion value, because conversion has not happened, and a share cannot be diluted by an event that has not occurred. A preferred trading at a discount to its liquidation preference stays at the accrued preference. It is never marked down to market, because the claim standing ahead of common shareholders is the legal preference, not the price at which the instrument changed hands today.
The symmetry is the point. Marking claims to market in either direction imports someone else's risk into the common shareholder's ownership measure. When a company retires a claim below face, the gain is real and CEBE records it on the date it happens. Until then, an unrealized discount is an opportunity, not property.
Cash netting follows the same logic in miniature. Cash offsets claims until it pays them, which is why Net Senior Claims subtracts cash. A dollar of treasury cash standing against a dollar of debt is a claim already provisioned for.
Principle twoA per share measurement must describe a single coherent state. The claims in the numerator and the shares in the denominator have to come from the same world.
CEBE describes the standing state. Claims exist at face, conversion has not been assumed, and the share count is the basic shares outstanding. Every instrument appears exactly once, in its current form.
FD BPS, the market's convention, describes the converted state for convertible instruments. The note has become shares, the shares are in the denominator, and the claim is gone. For converts, that is internally coherent. For everything that cannot convert, straight debt and most preferreds, FD BPS is simply blind. Those claims exist in the converted state too, and FD BPS counts them nowhere. This is why a preferred issuance can raise FD BPS while leaving the common shareholder's net position exactly unchanged.
The test catches the constructions that fail it. A metric that subtracts a convertible note as debt in the numerator while also counting its conversion shares in the denominator is describing a world where the same note is simultaneously repaid and converted. No shareholder can live in that world. Prudence is not a defense for incoherence. A number can be conservative or it can be a measurement, and when forced to choose, this site chooses measurement and labels conservatism as the scenario it is.
Principle threeBasic shares are a fact. Every fully diluted count is a model, and the issuer holds the knobs.
The basic share count is one observable number from the register. A fully diluted count requires choosing a method, Treasury Stock Method or as converted, deciding which instruments are rationally exercisable, and applying anti dilution exclusions. Two competent analysts running fully diluted on the same company at the same moment can produce different numbers and defend both. The label is one word wearing many methods.
Worse, the most common method is reflexive. A TSM denominator shrinks in a selloff as instruments fall out of the money and drop out of the count, which means the per share metric improves mechanically while shareholders lose money, and degrades mechanically in rallies. A measurement whose denominator is a function of the stock price has a feedback loop where its foundation should be.
So CEBE keeps the model out of the measurement. The denominator is basic shares. Dilution scenarios get their own clearly labeled instruments, conversion overhang, As Converted pro formas, and the issuer's own audited diluted count, which this site reports from EPS footnotes as the issuer's figure and never computes in house. The moment an analyst picks a fully diluted method, they have joined the drift instead of measuring it.
The denominator counts economic classes only. Where a company's structure includes a share class with no dividend rights and no liquidation rights, a pure voting instrument, that class is excluded, matching the issuer's own two class treatment in its audited per share calculations. Share counts measure ownership of economics, and a share with no economics is a vote wearing a share's costume.
Three constructions circulate under names close enough to be mistaken for variants of one number. They are three different objects. Basic shares outstanding is a fact from the register, and it is CEBE's denominator. Market convention fully diluted adds every dilutive instrument without asking whether it is currently exercisable, and it is the count this site carries in the comparison column. The third is the moneyness partitioned construction an issuer published in July 2026, which converts only in the money instruments into the count and deducts out of the money converts and preferred at notional instead. That third construction is internally coherent, its rule is stated, and it is not the same object as the second, which is why this site names it separately rather than folding it into a fully diluted column. Its properties are worked through on The Net Effect.
The comparison column uses the all dilutives convention deliberately, and the reason is that it is the harsher of the two. A comparison a measurement builds against itself should not be tuned in the measurement's favor. All dilutives is the widest denominator applied in practice, so the gap it opens against CEBE is a floor on the difference rather than a flattering midpoint, and a reader who prefers a partitioned count will find the gap narrower than shown, never wider. Wherever the counts appear together, the construction of each is named, on the rule that a metric label without a stated formula is a genre rather than a number.
What the formula counts, and in what state.
This section applies realize first and one state of the world to convertible instruments.
A convertible instrument has two possible futures. It stands until it is settled as a claim, or it converts and becomes shares. The measurement holds every convert in the standing state, claim at face, conversion not assumed, per the realize first principle.
The converted state is not a second headline number. It is shown where it is decision relevant, as a per instrument pro forma computed under the same rules, the As Converted CEBE for that instrument. The convert's claim leaves the numerator, its conversion shares enter the denominator, and every claim that cannot convert stays exactly where it was, because straight debt and preferreds do not vanish when a note becomes stock. Each convertible instrument therefore carries its own envelope, the standing state on one side and its converted state on the other, and the width of that envelope is the swing the instrument represents. For claims denominated in Bitcoin, the pro forma has a useful property, each tranche's step is independent of price and can be published in advance as a deterministic schedule.
The interpolation across each envelope is Adjusted Claims %, which weights each convertible claim by moneyness, on the observation that an instrument deep in the money behaves as equity and will most likely exit through dilution, while an instrument out of the money behaves as debt and will most likely be settled from the treasury. The bands, the weights, and the stress window are published in the Claims Grade rubric, stated as priors and open to challenge. This is the internal rule for the mixed case, and it lives in the valuation layer where probability belongs, not in the measurement where facts live.
FD BPS is not an honest measurement of the converted state. It is the market's incumbent convention, and it is reported on this site for comparability and for audit, not as a bound. It assumes conversion for convertible instruments while ignoring every claim that cannot convert, which means it measures no coherent state at all. The gap between FD BPS and CEBE is still information, but it is information about the convention, not about the company. It quantifies everything the incumbent metric hides, and it decomposes into two parts, the claims FD BPS never counted and the conversion it assumed.
A convert appearing as a claim in CEBE and as dilution in FD BPS is therefore not double counting. Double counting is one number counting an instrument twice. These are two numbers, each counting the instrument once, in different worlds, and only one of those worlds is internally consistent everywhere.
This section applies realize first to the preferred claim.
Three different numbers can describe the same preferred, and each answers a different question. Par answers what the instrument was issued at. Market answers what it trades at. Accrued liquidation preference answers what stands ahead of the common shareholder if seniority is ever enforced.
CEBE strips at accrued liquidation preference, because CEBE is measured from the common shareholder's seat, and the residual claimant's question is the third one. The market price of a preferred reflects the preferred holder's risks, rate risk and credit risk, and a selloff in the preferred does not reduce the claim standing ahead of common by one dollar. What a discounted preferred does represent is a retirement opportunity, and per the realize first principle, the gain is recorded when the company captures it. What that capture is conditional on, the discount standing at execution and the ranking against the other uses of the same treasury dollar, is set out in the crossover section of the retirement page.
Where a preferred's preference accrues or compounds, the accrued figure is used. Deferral of dividends does not shrink a claim. It grows one.
This section applies realize first to both sides of the balance sheet.
Held treasury securities net at fair value, at the mark the holder's own disclosure provides, with basis and as-of stated. Issued obligations carry at liquidation preference and are never marked to market.
Both sides are valued at what the common holder can reach: an issued claim stands ahead of the common at its full preference whatever it trades at, and a held asset is worth what it sells for. Only one input in the stack reprices with the market, treasury securities held as reserves.
What the counted claims are made of, and how the market prices them.
The measurement treats every claim at face. The valuation layer characterizes what the face is attached to, because identical dollar amounts can carry opposite behavior.
Fiat denominated claims compress as Bitcoin rises. The claim is fixed in dollars while the treasury floats, so Claims % falls in a rally and expands in a drawdown. Claims denominated in other currencies add a second variable, and can compress twice or expand twice depending on the currency pair. Bitcoin denominated claims do not compress at all. Their Claims % is static across every price, responding only to discrete conversion events, which produces a staircase where fiat claims produce a curve. And collateral backed claims with forced liquidation triggers can convert a drawdown into a forced sale, which is a different kind of risk than any ratio expresses, and receives a categorical treatment in the Claims Grade rubric rather than an arithmetic one.
Two companies with identical Claims % can therefore carry entirely different structural risk. The number is the beginning of the analysis, not the end of it.
CEBE mNAV is market capitalization divided by the value of the common equity Bitcoin position, the premium the market pays per unit of net per share exposure. Cycle mNAV is the forward version of the same multiple, adjusted for the wrapper. It compounds the annual cost of the capital structure, dividends, debt service, compensation, net of operating income, across the cycle window, so that a company burning capital to maintain its wrapper faces a multi year hill built into its multiple. A premium that looks identical to a cheaper company's premium at spot is not identical forward, because one of them must outgrow its own wrapper inflation just to stand still. CEBE Implied P/E divides the premium by the demonstrated CEBE growth rate, expressing the price of the stock in years of the engine's measured output. It prices the debate between snapshot and trajectory without resolving it, which is the correct amount of resolving for a measurement site to do.
None of this is a new invention, and it is not borrowed either. Price to book is what you get when you price equity against what remains after liabilities, which is the oldest equity question there is. CEBE mNAV is not a variant picked from a menu of multiples. It is the metric that emerges when you measure the claimant's residual correctly, and traditional finance already had a name for the result. It is price to book derived from first principles for a Bitcoin treasury, where two things book gets wrong materially move the answer. Book carries claims at par and carrying value, while CEBE strips them at real liquidation seniority. Book carries Bitcoin at cost or the lower of cost and fair value, while CEBE marks it at fair value. The two coincide exactly when the claims are simple, no preferred off par, no convert optionality, debt at face, which is the tell that book price to book is the frictionless special case and the claims aware version is the general one.
The same logic extends to the share count. An analyst who wants to weight convertible instruments by their proximity to conversion, between basic and fully diluted, is doing on the denominator what Adjusted Claims % does on the claims line. That proximity weighted share count is a valuation layer extension, and it is defined with the moneyness bands in the Claims Grade rubric, not in the measurement.
One concession is owed and gladly paid. For enterprise questions, credit analysis, acquisition framing, cross company enterprise comparisons, an EV construction is the right tool, and this site does not pretend otherwise. Metrics answer questions, and the honest taxonomy is a matrix, the question on one axis and the capital structure on the other, rather than a tree that assigns each company one number. The equity seat question exists for every company. So does the enterprise question. They are different questions, and a metric that answers one is not wrong for declining to answer the other.
A note on vocabulary, offered without heat. As of this writing, the sector uses the word amplification in at least three incompatible senses, a claims to reserve ratio, a leverage multiplier on net ownership, and dilution free Bitcoin growth. This site uses Claims % for the ratio and states the formula wherever a multiple appears, on the view that a metric label without a stated formula is a genre, not a number.
Each instrument is scored on four dimensions, zero to three points each. Lower is better for common equity. The sum maps to a letter, with one categorical override that sits outside the arithmetic. Because one dimension moves with the stock price, a grade is as dynamic as Claims percent, and every published grade carries its date, its Bitcoin price, and its stock price.
| Score | Character |
|---|---|
| 0 | Self extinguishing. A convertible with conversion events occurring, or a Bitcoin denominated convert moving through a tranche conversion staircase. |
| 1 | Maturing. A defined maturity date, repayable or refinanceable when it arrives. |
| 2 | Perpetual but managed. Callable or redeemable at the company's option, or a rate that adjusts to hold the instrument near par. |
| 3 | Perpetual fixture. No economic call, no conversion path, cumulative. It does not leave. |
Applies to convertible instruments, measured as stock price divided by conversion price at grading time. A non convertible instrument inherits its Permanence score here, because its lack of an exit is already counted once and should not be counted twice.
| Score | Character |
|---|---|
| 0 | Deep in the money. Stock at or above 130 percent of the conversion price. Conversion is near certain, the claim behaves as equity. |
| 1 | In the money. Between 100 and 130 percent of the conversion price. |
| 2 | Out of the money but within reach. Between 70 and 100 percent. |
| 3 | Deep out of the money. Below 70 percent of the conversion price. The claim behaves as debt and will most likely be settled from the treasury. |
| Score | Character |
|---|---|
| 0 | No expansion. A Bitcoin denominated claim. Its Claims percent is static across every price, responding only to discrete conversion events. |
| 1 | Standard fiat expansion, no trigger. The claim grows in Bitcoin terms as Bitcoin falls, but carries no covenant, no compounding pressure, and no engaged maturity falling due inside the stress window. |
| 2 | Fiat expansion plus a treadmill. Cumulative dividends compounding through the drawdown, or an out of the money claim whose cash maturity falls inside the stress window. |
| 3 | Trigger risk. Collateral backed with forced liquidation rights, margin maintenance, or covenant acceleration that can force Bitcoin sales. |
Currency denomination is recorded as a flag, not a score. An FX denominated claim can expand twice in a drawdown, but that is a market movement rather than a contractual claim growth mechanism, so it never enters the treadmill and never moves Downside.
A maturity date does not mean the same thing to two converts at the same Bitcoin price. An in the money convert leaves through conversion, so its maturity is a date its claim has already exited. An out of the money convert, and any non convertible debt, leaves through repayment from the treasury, so its maturity is a bill the reserve has to meet. The earlier rubric carried one flag, a maturity inside a fixed window, and fired it the same way for both. That charged a near date against a claim that was never going to need the cash. The maturity term is now gated by the same Moneyness reading used above.
The maturity is one input to the treadmill that defines Downside two, alongside cumulative compounding and double expansion. It is read only after the trigger and Bitcoin denominated tests have passed, so it never overrides them. When the claim is in the money, stock at or above the conversion price, the maturity contributes nothing and the score rests on the claim's other behavior. When the claim is out of the money, or is non convertible debt with no conversion exit, the maturity engages on the tenor schedule below.
| Out of the money tenor | Maturity contribution |
|---|---|
| Beyond 36 months | none |
| 24 to 36 months | none |
| 12 to 24 months | none |
| Under 12 months, the stress window | treadmill |
The under twelve month band is the stress window, narrowed to twelve months from the grade date so that it marks a repayment that is imminent. A claim that reaches it contributes the maturity term to the treadmill, which carries Downside to two for a claim that is not a trigger and not Bitcoin denominated. The schedule is a contribution within that branch, not an unconditional outcome. A trigger still reads three, a Bitcoin denominated claim still reads zero, and an in the money convert ignores the maturity entirely.
| Score | Character |
|---|---|
| 0 | Zero coupon. No cash cost. |
| 1 | A coupon under 5 percent, or any rate payable in shares at the company's option. Dilution is a cost, but a flexible one. |
| 2 | A coupon between 5 and 9 percent, cash only. |
| 3 | A coupon of 9 percent or more cash only, or any cumulative rate that compounds when unpaid. |
The second word is the severity of the claim burden on common equity. It tracks the letter grade and nothing else: it does not describe any single instrument's terms.
The total runs zero to twelve. Every grade pairs a letter with a word. The letter sorts, the word classifies. A grade stated as a bare letter reads as a report card and invites an argument about whether the letter is fair. A grade stated with its word names the severity the score measured, not a verdict open to dispute.
The word is not borrowed from the instrument. It is a reading of how heavily the claim sits on common equity, nothing more. A claim graded Elevated weighs the same on common whether the issuer calls it permanent, perpetual, or preferred capital. The letter sets the grade; the word says how hard it bears down.
Any instrument that scores a three on Downside behavior, a forced liquidation or covenant acceleration right against the Bitcoin, is graded D regardless of its other scores. The difference between C and D is the difference between a claim that drains common equity and a claim that can seize the collateral. Those do not belong on the same number line.
Illustrative structures, not company grades. Live grades publish per company, verified first.
The rubric is easiest to read through the instrument shapes it is built to separate. The scores below describe structures, not any company's current instrument, since the Moneyness dimension moves with a live stock price.
| Structure | P | M | D | C | Grade |
|---|---|---|---|---|---|
| Deep in the money zero coupon convert | 0 | 0 | 1 | 0 | A · Minimal |
| Bitcoin denominated convert, active staircase, zero coupon | 0 | 1 | 0 | 0 | A · Minimal |
| Out of the money zero coupon convert, distant maturity | 1 | 2 | 1 | 0 | B · Moderate |
| Fiat face convert settled in a fixed Bitcoin quantity, low coupon | 1 | 2 | 1 | 1 | B · Moderate |
| Out of the money convert maturing inside the stress window | 1 | 2 | 2 | 1 | B · Moderate |
| Perpetual convertible preferred, cumulative, deep out of the money, share payable | 2 | 3 | 2 | 1 | C · Elevated |
| Perpetual preferred, rate managed near par, non cumulative, cash only | 2 | 2 | 1 | 3 | C · Elevated |
| Perpetual preferred, high fixed cumulative, cash only | 3 | 3 | 2 | 3 | C · Elevated |
| Collateral backed Bitcoin loan with liquidation rights | 1 | 1 | 3 | 2 | D · Critical |
Every figure traces to a primary source, an SEC or exchange filing, a regulatory notification, or a company document, and carries a VERIFIED or EST flag. Estimated figures are labeled until a filing confirms them. Footnotes report deployed capital and announcements report raised capital, and the two are never conflated. Snapshots state their date, their BTC price, and their share count basis. Live displays use live prices. Historical comparisons that require a constant price say so explicitly and use one stated normalization price, never on live panels.
Historical BTC prices are daily closes from CoinGecko, the ratified source. Snapshots produced under the current convention, in force since 2026-07-21, store the close of the snapshot's own date, read as the following day's print at 00:00 UTC. Rows produced before that convention carry their original basis, which is stated on the row, until the normalization pass reaches them.
That convention has a name. #A-45 is the identifier for it, and it means exactly one thing: the BTC price for a date D is the CoinGecko print at 00:00 UTC on D plus 1. Where a chart or a card says a figure is struck “at its own snapshot print (#A-45)”, that is the rule being cited, and nothing else on this site carries that id.
#A-45 governs the Bitcoin price and only the Bitcoin price. The equity daily-close convention, which decides what a stock price on a given date means, is a separate rule and does not yet carry an identifier of its own; it will be defined here in one sentence at the point its id is installed, and until then no surface should cite #A-45 for an equity figure.
No CEBE figure depends on any issuer-defined metric. Every published value is recomputed from raw filed inputs, holdings, shares, claims at face, cash, and price. Independently verified at code level, August 2026. Issuer-reported metrics appear on this site only as attributed contrast, named as the issuer's figure and computed from nothing.
When an issuer publishes its own per share or multiple metric with its formula shown, this site reconciles it against verified filing data and publishes the reconciliation, agreement and divergence alike. Issuers that show their work make the whole sector more measurable, and the reconciliations are how convergence gets documented instead of asserted.
The computed dataset behind the tracker is published as a free JSON feed under the rules on this page. Endpoint, field dictionary, and license are documented on the public API page.
CEBE is not an enterprise value metric, and cannot price a takeover. It is not a prediction, and no scenario built on top of it inherits the measurement's authority. It is not a verdict on whether any premium is worth paying, only the denominator that makes the question precise. And it is not finished. The rubric publishes its priors so they can be calibrated against observed conversions, the moneyness bands are open for challenge, and every rule on this page is stated in public so that being wrong is cheap to discover and expensive to hide.
The standard is the point. Companies will keep shipping metrics, and the metrics will keep disagreeing, and that is fine, provided somewhere keeps the books in one consistent set of rules. This page is the rules.
Rulings that change a published construction are logged here, dated, with the effect on served figures stated.
2026-09-07. Preferred basis unified. A non-USD preferred face enters every surface at the snapshot's filed cross, never re-marked at the live cross. The company page and tracker re-marked Strategy's EUR-denominated STRE at the live EUR cross while the API, the comparison table, the charts and the drawdown pages carried preferred_usd as filed, so one company and one date had two CEBE figures on one site. Effect on served figures: Strategy's company page CEBE moves by about 50 sats and its break-even by about $20 to agree with the other surfaces; 155,895 sats at the 2026-08-30 snapshot print is the figure.
2026-09-07. Amplification definition corrected on the company page. The page footnote stated 1 / (1 minus Senior Claims %) and the code used the floored share. The ratified definition of 2026-07-25 is (1 minus Indexed Claims %) over (1 minus Senior Claims %), unfloored, from config/compute-amplification.js. Effect on served figures: Capital B's company page moves from 1.32x to 0.98x; issuers with no BTC-indexed claim are unaffected.
2026-09-07. Break-even has one definition, and the 2026-07-16 entry below installed the wrong one. Break-even solves the fiat leg only: net fiat senior claims (USD) over Bitcoin held net of BTC-denominated claims, and it does not exist where the fiat claims net to cash or the indexed claim covers the treasury. The form that entry installed, Net Senior Claims (USD) over Total BTC, is circular for a BTC-indexed issuer because Net Senior Claims USD is itself a function of the price being solved for, and it coincides with the served form only when the static claim is zero. Effect on served figures: the comparison table, the API and the company page already served the correct form; the drawdown template and the modeler engine are corrected, and the glossary sentence is rewritten.
2026-09-07. STRK's liquidation preference is not a ratchet. Under the July 2025 amendment it is the greatest of the $100 stated amount, the last reported sale price and a ten-day average, so it floats daily above a $100 floor and can fall. The Net Effect's "upward ratchet" and the glossary's "$1,000 per share" are corrected to the certificate's terms. Effect on served figures: none; preferred_usd already carries the accrued preference as filed.
2026-09-01. Claims Grade rubric amended: Downside 3 gains a second limb, and the company rollup rule is stated. The amendment. Downside 3 is now either (a) trigger risk, meaning collateral backed with forced-liquidation rights, margin maintenance, or covenant acceleration that can force Bitcoin sales; or (b) conversion at a floating discount to market with no floor, meaning the number of shares the claim becomes is a function of a future price without limit. Both are unbounded transfers from common equity, of coins in the first case and of shares in the second, and the rubric treats them identically. Fixed-price conversion, whatever its moneyness, is unaffected, and the D-cap applies to both as before. This does not contradict the standing rule that conversion resets and ratchets score 2 as a treadmill; it carves the unbounded case out of that entry. A discrete reset lands on a new fixed price and the share count is knowable the moment it lands, while a floating discount never lands, so the share count is unknowable while the instrument lives. Bounded repricing stays at 2 and unbounded repricing moves to 3. The rollup, a clarification of standing method rather than a change. The company letter is a face-weighted arithmetic mean of the four-dimension sums, rounded to nearest and then banded, with the D-cap as the categorical override at or above 50 percent of face and a D flag below it. It does not weigh letters, so no instrument can carry a company by holding the worst letter on a minority of face; it contributes its sum in proportion to its face and nothing more. The company letter function has no D branch at all, which makes D reachable only through the cap rather than merely usually reached that way. Effect on served figures: none. Every one of the 29 graded instruments was checked against limb (b) and none moves, because every convertible on the instruments tab carries a fixed numeric conversion price and the remaining thirteen instruments are not convertible at all. One limit of that check, stated because it is easy to overstate. It keys on the conversion mechanics as recorded on the instruments tab, so it establishes that no published grade has a recorded floating conversion, not that none has an actual one. A tab can carry a fixed number for an instrument whose operative conversion floats. Closing that gap needs a document-level read per convertible rather than a field scan, and it has not been done.
2026-08-20. Official-site links added. Each company page carries a header link to that issuer's canonical corporate root, and three editorial mentions are linked: the first Strive reference in Why We Tolerate Contradiction, both body references in The Same Companies, and TrueNorth in the management paragraph of the Strive deep dive. Digital Credit is a term of art rather than an entity and is deliberately left unlinked. Three issuers carry no link: the roots recorded for Smarter Web Co, H100 Group and Twenty One Capital have no public DNS record as at this date, and the question of which root is canonical for them is routed to each company rather than guessed, so the link is withheld until it can be stated. Effect on served figures: none. No input, formula, or published number changes.
2026-08-05. The claims perimeter is three fiat terms, and the change moves published figures. Net Senior Claims is debt plus preferred less cash, with preferred carried at accrued liquidation preference. Declared but unpaid preferred dividends no longer enter as a separate item. The reasoning is stated in full beside the Net Senior Claims definition above: the accrued preference is the carrying value, a declared dividend is the claim converting to cash rather than a claim added on top of one already counted, and carrying both would count the same obligation twice. The disclosed dividends-payable figure is retained and still served, as a non-netting disclosure field. Effect on served figures: Strategy only. Senior Claims % moves from 34.00 to 33.71 and CEBE from 143,392 to 144,023 sats, at the 2026-08-02 snapshot basis. No other tracked issuer reported the field, so no other figure moves. The separate-item treatment was live from 2026-07-30 to 2026-08-05 and the correction is logged at cebetracker.io/corrections. Two consequences are worth stating. Today no tracked series carries a suspended dividend, so accrued equals face across the fleet and the two constructions agree everywhere except in the arrears case, which is scoped separately and will be carried through the preference itself. And the Return Attribution specification moves to revision v1.2 for the same reason, which re-arms its reproduction gate, because the layer calls the reader this ruling changed.
2026-08-02. Methodology page batch, five additions, no computed construction changed. One, an independence credential added to data standards: no CEBE figure depends on any issuer-defined metric, and every published value is recomputed from raw filed inputs. The basis is the criterion-7 sweep of the served set and the repository, run 2026-08-02, verdict clean. That sweep confirmed the claims engine reads raw fields only, that every mNAV on a served surface is CEBE mNAV computed from a stock price against a computed residual rather than ingested, that no stored issuer-KPI column is read anywhere in the tree, and that issuer metrics appear only as attributed contrast. The credential expires against new code. A new surface displaying a CEBE figure, or a new source column carrying an issuer-reported metric, falls outside the sweep and needs its own check. Two, the cash netting rule is now stated beside the Net Senior Claims definition, with its state-consistency basis, where previously the answer lived in the principles section and not where a reader checking the formula would meet it. Three, the numéraire is stated as a convention with its rationale rather than as a uniquely valid frame, and the presence of dollar renderings across the site is stated with it. Four, the claims-scope note: accrued preferred dividends tracked at the instrument level, accrued interest on the convertible stack excluded and immaterial at coupons of 0.00 to 2.25 percent, deferred tax liabilities and CAMT excluded with the reasoning deferred to the refusals discussion. Five, the historical BTC price source is stated in data standards, CoinGecko daily closes, with the convention in force since 2026-07-21 storing the snapshot date's close as the following day's 00:00 UTC print. Rows produced before that convention carry their original basis, stated on the row, until the normalization pass reaches them. Effect on served figures: none. All five state existing construction, and none of them changes an input, a formula, or a published number.
2026-08-02. Two rulings land as page text, and the retirement page gains the section the second one points to. Ruling 2.1, the share count constructions. Three constructions are named and held apart: basic shares outstanding, the register fact and CEBE's denominator; market convention fully diluted, the all dilutives count carried in the comparison column; and the moneyness partitioned construction an issuer published in July 2026, which converts only in the money instruments and deducts the rest at notional, a coherent construction and a different object from the second. The comparison column stays on the all dilutives convention deliberately, because it is the harsher of the two and a comparison a measurement builds against itself should not be tuned in the measurement's favor. A reader who prefers a partitioned count finds the gap narrower than shown, never wider. Ruling 2.2, retirement conditionality. The preferred-basis section now points to the crossover treatment on the retirement page for what a retirement gain is conditional on, and that treatment ships with this batch as a new crossover section on the retirement page. It states that the discount is contractual at execution but conditional on the discount standing then, and that the ranking against a common repurchase turns on a threshold that recomputes with the BTC price and the claims stack. Effect on served figures: none. Both rulings describe construction; no denominator, formula, or published number changes, and the comparison column is unchanged in what it computes.
2026-07-28. Adjusted Claims % specification amended to v1.1, ratified 2026-07-28, following an implementation review by Adam Livingston. Seven clarifications, none of which changes a computed construction. One, the intro's description of ACP corrected from "measurement construction" to "valuation construction," matching what section 1 states. Two, the denominator pinned as basic shares plus contractual conversion shares for any claim the adjustment carries as converted, the symmetry rule, with no treasury-method dilution and no modeled share count, and the rule generalized to continuous weighting. Three, the weight convention pinned: conversion_weight runs in the probability-of-equity direction, standing_weight is the fraction remaining senior, the two sum to one, and the four-band schedule published 2026-06-14 on the Claims Grade page is expressed as standing weight. Four, three inherited properties stated explicitly, that conversion value to the claimholder includes accrued amounts and current ratchet state as of the mark date, that ACP inherits CEBE's full netting perimeter including cash and issuer-held instruments, and that the zero floor is inherited. Five, the quote convention stated, every mark a close or a last trade with its source named, and marks older than the weekly cycle rendering flagged STALE. Six, break-even under ACP scoped to held-constant marks and weights, with the dynamic fixed-point solve out of scope for v1. Seven, the review credited on the spec page beside the clarifications it produced. Effect on served figures: none. The 2026-07-26 ratification stamps are unchanged and continue to annotate only what they covered; the v1.1 clarifications render as separate dated notes. The specification lives at cebetracker.io/claims/spec/.
2026-07-28. Section 3.3's external-review credit corrected to Valentin Kosanovic; the surname was misattributed at initial publication. The credited review and the mark rule are unchanged.
2026-07-24. Smarter Web Co (TSWCF) senior claims corrected, and the static BTC claim now has one reader. Smarter Web repaid its convertible loan note in full on 23 July, and the settlement snapshot recorded the retired BTC-denominated claim as an explicit zero. The mechanism, established from the source, was a legitimate zero swallowed by a greater-than-zero guard, not an unrecognized instrument status. Senior Claims % is computed from each snapshot's own btc_claim_static value, resolved in one place, the computeDrag routine in config/compute-drag.js. Before the fix, that routine parsed the column and kept the value only when it was greater than zero, otherwise falling back to a hardcoded override. When Smarter Web's 23 July settlement snapshot reported a filing-true btc_claim_static of zero, the greater-than-zero test read that zero as though the column were empty and substituted the override of 177.89 BTC, so a claim the company had already settled kept being counted. The repaid instrument status had no part in this path. The claims engine reads the snapshot value, never an instrument status field, so there was no status gate present and nothing for the repaid value to slip past. The fix replaced the greater-than-zero test with a check that the column is present and non-empty, a single resolver named resolveBtcClaimStatic in config/compute-drag.js, which returns whatever value the cell holds, including zero, and falls back to the override only when the cell is blank or unparseable. Effect on served figures, at the snapshot price of $66,077: Senior Claims % overstated at 19.5% against a correct 12.9%, CEBE understated at 584 sats against a correct 632, and CEBE mNAV overstated. The error was conservative against the framework's own metric, and it was confined to Smarter Web, since no other company recorded a zero in that column. The fallback exists for old snapshots predating the column and still applies when the column is blank. It no longer applies when a filing reports zero. That single resolver now serves the readers that had each carried their own copy of the test: the public API, the tracker, company pages, credit, the stack, the compass, engine state, and the admin tools. The same pass found that the compass page carried its own copy of the claims formula that had never been migrated to the 2026-07-10 zero-floor ruling below, and was applying the floor to the fiat bucket rather than the total. The compass now uses the shared formula. Effect on the compass: Capital B (ALCPB) claims figures there move into line with every other page.
2026-07-16. Superseded 2026-09-07: the definition this entry installed is the wrong form; see the 2026-09-07 break-even entry above. Break-Even BTC Price glossary definition corrected. The glossary previously defined Break-Even through a market-parity construction, the BTC price where CEBE equals stock price over BTC price. That was wrong. Break-Even is and has always been computed as Net Senior Claims (USD) / Total BTC, the price where Claims % reaches 100% and CEBE reaches zero, a balance-sheet level independent of the stock price. Effect on served figures: none. The tracker, credit page, company pages, and modeler engine already computed break-even canonically; the error was confined to the glossary definition, which is rewritten to canon.
2026-07-10. Zero floor moved from the fiat bucket to the total. An adversarial audit of the site on 2026-07-09 (finding 3) showed the published formula and the serving code disagreed for issuers holding both BTC-indexed claims and surplus cash. The code floored the fiat term at zero before adding BTC-indexed claims, which discarded the cash surplus. The ratified construction converts all claims to BTC at the snapshot price, nets cash against the total, and applies the zero floor to that total, as the formula above now states. Effect on served figures: Capital B (ALCPB) Claims % moves from 33.77% to roughly 29.4%, with CEBE and CEBE mNAV recomputing accordingly. No other company changes, since no other name holds surplus cash against BTC-indexed claims.
CEBE Framework by @chcbearsfan. cebetracker.io.