Canonical CEBE carries every claim at contractual value; the Adjusted Claims layer reprices the same stack at observed market prices and is a valuation output. It never replaces the measurement, because marking claims to market makes a distressed stack flatter the common, which is the circularity the measurement layer exists to refuse.

Adjusted Claims, the layer · Adjusted Claims %, the ratio · MABE, the per-share output · conversion weight, the section 3.1 parameter under test

Study: Adjusted Claims % section 3.1 continuous conversion weight, candidate fitting.
Ruling authority: Bobby Tierney. Date: 2026-07-28.
Status: CLOSED. All rulings received. Panel run. Results in the companion results document.

This log is the companion record to Calibration Study 1. Its text is dated and is never rewritten: corrections are recorded beside the originals rather than replacing them, because a pre-registration log that edits its own history is worth less than none.

1. The tolerance

Stated and ratified in the governing thread on 2026-07-28, before any fit output existed there, and transmitted to this lane late by transport failure. Verbatim:

Primary pass RMSE less than or equal to 5 fair-value-to-face points across the full panel; maximum single-observation deviation less than or equal to 12 points; on the repurchase transaction evidence, absolute error less than or equal to 6 points. All three must clear, on held-out error.

The magnitudes are stated conventions, not derived quantities. Their function is to be fixed before results exist.

Provenance limit, recorded rather than smoothed over. This lane cannot independently verify the timestamp of the original statement, because it was made in a different thread. What this lane attests to is narrower and is verifiable here: the tolerance arrived before any candidate had been scored against any observed mark, and the harness was mechanically unable to produce a score until it did. Closing the chain requires both transcripts. The ruling authority has stated that the governing thread's transcript carries the earlier timestamp, and that statement is recorded as made.

RECORDED 2026-08-05 | The chain closed. Both threads' records were reconciled and the sequence was re-verified against the governing thread's transcript on 2026-08-05: the tolerance was stated blind on 2026-07-28 before any fit output existed in any lane, and the late arrival here was a transport failure rather than a sequencing failure. The 2026-07-28 provenance limit above stands as written, because it was an accurate statement of what this lane could attest to on that date. The results document carries the verified statement in its section 1.

2. Rulings

RefQuestionRuling
R1Price mapMap A, the ratified spec 3.1 blend. Map C deferred to a v2 study.
R2Section 8's price space languageStands as written. Two sided price test. The one sided restatement was offered and declined.
R3Tolerance and metricAs section 1 above. Three tests, all must clear, on held-out error.
R4Fitted versus prior scoringLeave one date out held-out only. The tolerance binds on held-out error for all candidates. AICc may be reported as context; nothing binds on any information criterion.
R5Prior deferenceBoth tests run and report separately. The absolute tolerance is the only launch gate. Beat-the-prior guides v2 only.
R6Candidate 3 readingReading A only for the price fit. Reading B is a claims-line construction, not an instrument value, and cannot price a convert. Documented as a formal finding, not fitted.
R7Candidate 1Both N(d1) and N(d2), reported separately.
R8Rotation rowsQuarantine from the primary panel, dual run as a labelled sensitivity. Hypothesis routes to the MSTR anchor lane for filing-level re-verification. Filed data is never edited on a hypothesis.
R9Band boundariesLeft closed, right open throughout, matching the page's explicit "at or above 130". Convention documented, exact hits on 0.70 and 1.00 flagged.
R10DispositionFail: the no-pass clause publishes verbatim with the structural diagnosis and the continuous layer stays unlaunched. Pass: the gate opens for that candidate; launch remains a separate ratification, the gate is necessary not sufficient. Multiple passes: fewest parameters wins, ties broken by held-out RMSE.
R11Prior marginA challenger earns the label "outperforms the published schedule" at 10 percent or better held-out RMSE improvement. No automatic replacement. Succession is a v1.2 ratification decision informed by the v2 study. All Map A rankings carry the stated bias label.
R12Evidence classesTolerance binds on the 41 disclosed marks. The traded print binds at 6 points absolute and reports as the out-of-sample check with its n=1 and both-legs-approximate caveats stated. Both classes publish.
R-VOLVolatility sourceSource it. Dashboard prints where they exist, option-chain IV interpolated to each convert's remaining term where they do not, method documented, sources per date. Do not fit sigma (it destroys the zero-parameter standing and voids the cap). Do not assume a flat vol. Partial coverage is disclosed as partial rather than padded.
R13Data laneBTC-held and gross-claims series at the ten dates approved. Tolerance's bite reports in Adjusted Claims percentage points as context; the binding tolerance stays in points of face.

3. Corrections to the record

Card option order shifted between rounds and two stamps captured the wrong rulings. The authoritative text of 2026-07-28 governs and the corrections are recorded here rather than overwritten, because a pre-registration log that edits its own history is worth less than none.

RefMis-stamped asGoverning ruling
R4"information criterion penalty"Leave one date out held-out only. AICc is context only and binds nothing.
R5"beat-the-prior test"Both tests run; the absolute tolerance governs as the launch gate. Beat-the-prior guides v2.

Two further items were revised in sequence during the session, both by the ruling authority and both before any scoring:

Process note. Interactive option cards are retired for this workstream. Rulings are requested as numbered plain text with options and a one-line recommendation, and returned as text. Single yes or no confirmations may still use a widget.

4. The floor, and what it settles in advance

The per-mark oracle floor under Map A is 14.07 points of RMSE. That is the error of a hypothetical weight function allowed a different unconstrained w for every mark with full knowledge of the answer. It binds every function in [0,1] equally and exceeds the ratified 5 point threshold by a factor of nearly three.

oracle floor 14.07 points of face ratified gate 5.00 points of face

Consequence, stated in the log because it was known before the panel ran: no candidate can pass under Map A, with or without the volatility series. The volatility series affects relative rankings and v2 groundwork only, never the gate outcome. Candidates 1, 1b, 4a and 4b being pending does not leave the disposition open.

The floor was disclosed to the ruling authority before the tolerance was transmitted here. It is candidate-invariant, so it cannot advantage or disadvantage any candidate and cannot be used to reverse engineer a tolerance that selects a winner. That was the disclosure test applied throughout the setup phase.

5. Outcome

No candidate cleared. The no-pass clause fired. The gate holds and the continuous conversion weight of spec 3.1 stays unlaunched. The structural diagnosis publishes with it, per R10.

Four items route to v1.2: the time-value finding, the binary degeneracy, the two-object finding on the published schedule versus spec 3.1, and the direction of the error.

Companion record

The results this log pre-registers publish at Calibration Study 1: Adjusted Claims Conversion Weight. The gate both documents answer is Specification v1 section 8.

CEBE framework, cebetracker.io. Part of the calibration record. Valuation layer. This is a pre-registration record and not investment advice.