Family: Cal-num, third in order, after Calibration 1 and Calibration 2.
Calibration 1 and Calibration 2 both failed. This is the third attempt, and what makes it different is that the rulebook now names a single construction with its parameters already fixed, rather than a family of candidates to be fitted at scoring time.
A search over 3,578 construction variants selected that specification. Everything the search touched is historical data, and historical data cannot both choose a specification and vouch for it. So the specification was written down, its parameters frozen, and a set of filed values that no part of the search has ever read was left sealed. That sealed set is the test.
This page is the rulebook. It carries no verdict, because the verdict does not exist yet.
Terms expanded on first use below: Adjusted Claims Percent (ACP), implied volatility (IV), root mean squared error (RMSE).
Research extension. This page is a study, not a measurement. The core standard is CEBE, Senior Claims %, CEBE mNAV and Claims Grade, each published with its formula and date. Studies here test priors against outcomes and can be retired; retired studies stay on the site with their correction notes. Nothing on this page enters a tracker figure.
The specification cleared admission on the historical panel. That sentence is easy to misread, so here is the plain version.
The historical panel qualified the specification for scoring. It is not a result. Admission was computed on the same 41 filed marks the search consumed when it chose this construction, so it shows that the specification is well formed and prices that panel within tolerance on a live option leg. It shows nothing about data the specification has never seen. No out-of-sample result exists.
Publishing a registration that carries no result is not an omission. It is what a pre-registration looks like from outside: the criteria are fixed and public first, and the verdict arrives afterwards or the exercise is worthless.
Model value to face, in points of face, for one convertible series at one measurement date:
The bond floor discounts coupons and principal to the investor put date, semi-annually, at the US Treasury par yield for the nearest published CMT tenor plus a credit spread. The conversion option is a European call on conversion parity struck at par, expiring at that same investor put date, so both legs price the instrument to one terminal date.
The conversion weight w occupies the N(d1) slot of the option and that slot only. N(d2) is left as Black-Scholes computes it, which is what keeps the sourced volatility inside the price and leaves the option leg holding time value. Calibration 2's ruled coupling placed the weight in both slots, and those are the only two terms through which a European call's volatility enters, so that construction removed volatility from the price entirely.
The weight responds to standardised log-moneyness, not to raw moneyness:
Raw moneyness carries neither volatility nor tenor. A weight standing in for an exercise probability has to respond to both, and σ√T states them exactly.
Volatility is the sourced 30-day implied volatility at each mark, scaled to the option's tenor by the at-the-money term-structure ratio measured on the 2026-08-14 option chain. Only the ratio travels; the level at every mark stays pinned to that mark's own sourced figure.
| Parameter | Value | Role |
|---|---|---|
| A | 0.877938 | slope of the weight in standardised log-moneyness |
| B | 0.816968 | intercept of the weight |
| SPREAD_ADD | 8.803434 | percentage points added to the published B band spread |
Fitted once, by least squares in price space, on all 41 marks of the Calibration 1 historical panel. They do not move. The confirmation run applies them as given and does not refit for scoring.
Because the widening is fitted, the specification is exactly insensitive to the B band's level and uses only its variation through time.
The scored protocol is leave-one-date-out. Each measurement date is withheld in turn, the specification is refit on the remaining dates, and the withheld date is scored against that refit.
Those three numbers carry forward from Calibration 1 and Calibration 2 unchanged. This registration does not move them.
Admission is the scored protocol plus the degeneracy constraint of section 5, and nothing else. Every other protocol is computed, published, and gates nothing.
| Protocol | Role | Historical panel | Against 5.0 / 12.0 |
|---|---|---|---|
| Leave-one-date-out | admission, scored | 4.711 / 9.60 | PASS / PASS |
| Live option leg | admission, structural | 1.732 / 1.882 / 0.210 | PASS |
| Leave-one-series-out | disclosed diagnostic | 4.830 / 10.14 | PASS / PASS |
| Expanding-window forward | disclosed diagnostic | 4.746 / 11.77 | PASS / PASS |
| Leave-two-dates-out | disclosed diagnostic | 4.826 / 14.94 | PASS / FAIL |
| Half-sample split | disclosed, no tolerance | 7.308 / 13.59 | not applicable |
This structure was chosen with every one of those numbers already in hand, including the failure. Saying so is the point. Three things follow, stated rather than left to be inferred.
An earlier draft of this registration contradicted itself. It made admission the worst of four protocols, under which the leave-two-dates-out failure at 14.94 meant the rulebook rejected its own candidate on the data used to select it. That contradiction was found and raised before the seal was touched.
A strict worst-of-four gate admits nothing sound. Across the 3,578 variants, six clear such a gate on both tolerances and all six fail the degeneracy constraint. Zero clear it while keeping a live option leg. A gate passable only by a construction whose option leg has stopped being an option is not a quality bar.
The operative test is therefore the sealed set, not this panel. Every number in the table above was computed on data the search consumed. That panel selected this specification and cannot also vouch for it.
The leave-two-dates-out failure is named here as the known risk and is not hidden behind a relaxed threshold. Its status is fixed in both directions before the seal breaks: it cannot be restored to gating if it passes, and its 12.0 threshold cannot be relaxed if it fails.
Clearing the numbers is necessary and not sufficient. The specification must also keep an option leg that behaves like an option.
This matters because a convertible construction can stop pricing optionality by two routes that look nothing alike and are identical in effect. Calibration 2 reached it through the coupling, by placing the weight in both N slots. This calibration's search reached the same place from the opposite direction: given a free multiplier on the sourced volatility, the optimiser drove it to whatever ceiling it was given, and at a high enough level the option leg saturates onto conversion parity and volatility stops reaching the price. Those constructions score well and price nothing.
| Measured case | Volatility-price sensitivity | Mean σ√T | Mean N(d2) |
|---|---|---|---|
| Calibration 2 coupling, volatility absent by construction | 0.000 | 1.39 | not applicable |
| A saturated search leader, volatility multiplier 2.56 | 1.148 | 3.56 | 0.075 |
| This specification | 1.732 | 1.882 | 0.210 |
| Plain Black-Scholes at the sourced volatility | 2.660 | 1.39 | 0.325 |
Registered floors: volatility-price sensitivity at or above 1.00 points of face, mean σ√T at or below 2.00, mean N(d2) at or above 0.05. A specification clearing the scored protocol on a construction that fails any of these is reported distinctly from a pass, and is not admitted.
Of the three parameters, the credit widening is much the loosest. Across the nine leave-one-date-out folds its standard deviation is 0.919 points, against 0.071 and 0.051 for the other two.
Derived as the nine-fold mean plus and minus three fold standard deviations, rounded outward. Twenty-five independent refits span 7.084 to 11.144 and sit strictly inside the band. In total-spread terms this asserts, in advance, that this issuer's convertibles price wider than the CCC index across the whole plausible range.
If a diagnostic refit on the confirmation panel falls outside that band, it is a failure of this specification and publishes as one. It is not a data surprise and not an occasion to widen the band. The direction is registered too, so an out-of-range value can be read: the widening moves inversely with the sourced volatility, so a value above the band is the signature of an understated conditioning volatility, while a value below it has no ready explanation and is the more serious case.
Stated here because a registration that publishes only its strengths is not a registration.
The confirmation set holds a tenth measurement date of filed convertible fair values that no exploratory run has read. Its state is committed and auditable, and its provenance is recorded in full, including the part that is less than perfect: the 2029 repurchase print inside it has already been scored twice, under Calibration 1 and Calibration 2, so it is carried as a demoted secondary observation and is not a virgin cell. The Q2 2026 fair values are untouched.
When the seal breaks, the confirmation run applies the frozen specification without refitting, scores the scored protocol at 5.0 and 12.0, publishes the disclosed diagnostics against their stated tolerances, applies the degeneracy floors, and tests the widening against its registered range.
The result publishes either way, pass or fail, following the Calibration 1 and Calibration 2 precedent. If the specification fails, that finding publishes and the section 8 gate holds. No threshold, protocol role or parameter on this page may be moved after the confirmation numbers are seen.
Full registered text, including every date rule, tenor convention and mechanism citation: docs/specs/cal-3-registration-2026-08-19.md in the CEBE Tracker repository, ratified text at blob 0b5e95f23249f54d158c57eb7c58b7e2dacf2d69.
One frozen specification against ten dates of filed convertible fair values, including a sealed quarter the specification had never seen. The tolerance was stated before the specification existed and held unchanged through the run.
Valuation layer, gate result. Canonical CEBE is untouched by this study in either direction. The object under test is the pricing construction behind the continuous conversion weight of Adjusted Claims % Specification v1. Admission opens the section 8 gate. Launch of any conversion-weighted surface remains a separate ratification and has not occurred.
Carried forward unchanged from Calibration 1, stated and ratified blind on 2026-07-28 before any candidate had been scored against any observed mark, and inherited by this registration verbatim: primary pass is RMSE at or below 5 fair-value-to-face points across the full panel under leave-one-date-out scoring; maximum single-observation deviation at or below 12 points; on the repurchase evidence, absolute error at or below 6 points. The registration added the mechanism floors, fixed before the seal: volatility-price sensitivity at or above 1.00 points of face, mean σ√T at or below 2.00, mean N(d2) at or above 0.05. The floors exist to prove the option leg is alive; a specification that clears the error cells with a dead option term is a debt model wearing a convertible label, and it does not admit.
The confirmation set was sealed before the specification was frozen: the tenth measurement date, the Q2 2026 filed convertible fair values from accession 0001050446-26-000044, Note 6, committed untouched and byte-verified at run time. No exploratory run read it. One print inside it, the 2029 series repurchase of 2026-05-19, had been consumed by Calibrations 1 and 2 and was therefore demoted before sealing to a secondary observation with its own disclosed cell, unable to gate in either direction.
The specification prices each convertible series as a bond floor plus an option leg. The bond floor discounts the note's cash flows to the investor put date, semi-annual, at the constant-maturity Treasury rate for that tenor plus a fitted spread. The option leg is a European option on conversion parity struck at par to the same date, priced on sourced implied volatility, with the fitted weight applied in the N(d1) slot only and standardized log-moneyness in the z term. Three parameters were fitted on the historical panel and frozen in the registration on 2026-08-19: A 0.877938, B 0.816968, SPREAD_ADD 8.803434. Nothing was refit at scoring.
Volatility is sourced, never fitted and never flat: 30-day implied volatility at each mark, scaled by the at-the-money term-structure ratio measured on the 2026-08-14 option chain, per the R-VOL sourcing rules. Two input extensions were committed with provenance at the run: IV30 values of 95.16 and 67.77 for the two 2026 dates, quoted verbatim from the same primary evidence file as every existing R-VOL row, and MSTR closes of 86.93 at 2026-06-30, anchored to the verified snapshot, and 164.63 at 2026-07-24, the close the 10-Q itself states. The 2026-05-19 close used for the demoted print carries no in-repo anchor at publication and is flagged as such until independently verified.
Gate check came first in the run. The registration file at HEAD hashed identical to the ratified blob, the sealed files hashed identical to their committed state, and the holdout diff was empty after the run as before it. The run reproduces byte-identically.
| Cell | Figure | Basis | Threshold | Result |
|---|---|---|---|---|
| Leave-one-date-out RMSE, full panel | 4.747 | In-sample. Nine of the ten dates were consumed by the exploratory search. | at or below 5.0 | PASS |
| Leave-one-date-out max deviation | 9.56 (MSTRFV-006) | In-sample, same panel. | at or below 12.0 | PASS |
| Mechanism floors, refit on 47 marks | 1.669 / 1.872 / 0.194 | In-sample, refit on the full panel. | at or above 1.00 / at or below 2.00 / at or above 0.05 | PASS |
| Mechanism floors, frozen, six confirmation marks | 1.169 / 1.800 / 0.079 | Sealed. Frozen parameters on the six confirmation marks. | same | PASS |
All figures in this part are from the confirmation run of 2026-08-22. PASS marks a cell that cleared its threshold on marks the exploratory search had already consumed. PASS marks a cell scored on the sealed confirmation marks, the only data the specification had never seen. A column of in-sample passes is one panel read several ways, not several independent confirmations, and the same reading applies to every in-sample chip in section 4.
ADMITTED. The section 8 gate opens. Launch remains a separate ratification.
The fitted spread survived its own registered test. The ten-date refit of SPREAD_ADD lands at 6.9166, inside the registered range of 6.1 to 11.7. The fold withholding 2024-06-30 refits at 5.664, below the band, and is disclosed beside the in-range result.
Every disclosed diagnostic publishes beside its threshold, the failed cells at the same prominence as the passes. All figures are from the run of 2026-08-22.
| Diagnostic | Figures | Basis | Threshold | Status |
|---|---|---|---|---|
| Leave-one-series-out | 4.885 / 10.71 | In-sample, non-gating. | 5.0 / 12.0 | PASS PASS |
| Forward protocol | 4.808 / 11.77 | In-sample, non-gating. | 5.0 / 12.0 | PASS PASS |
| Leave-two-dates-out | 4.826 / 13.36 | In-sample, non-gating. | 5.0 / 12.0 | PASS FAIL |
| Half-panel split | 7.363 / 13.57 | In-sample, non-gating. | none registered | Disclosed |
| Confirmation marks alone | RMSE 5.077, max 8.73, mean signed error negative 3.516 | Sealed confirmation marks, not a registered test. | not a registered test | Disclosed |
| Demoted 2029 repurchase print | model 78.99 against 92.00, absolute error 13.01 | Consumed by Calibrations 1 and 2, demoted before sealing. | 6.0 | FAIL non-gating by demotion |
Three of these deserve plain language.
The leave-two-dates-out failure is the registration's named known risk, reproduced at scoring smaller than its exploratory value of 14.94 and fixed in advance as non-gating in both directions. It could not have blocked admission, and a better figure could not have substituted for a gating cell. It stands in the record as the specification's known fragility under heavier data removal.
The confirmation marks alone read above the panel tolerance, 5.077 against 5.0, with a mean signed error of negative 3.5 points. On the newest quarter the model prices the notes cheap. Confirmation-alone was never a registered test, the binding statistic is the full-panel leave-one-date-out figure, but the slice is stated here rather than left for a reader to compute, and it is the first line the forward record watches.
The demoted print failed its cell by more than double the tolerance. It was demoted before sealing because Calibrations 1 and 2 had already consumed it, which is why it could not gate; the miss publishes at full prominence anyway. Its price input additionally carries the provenance flag noted in section 2.
One resolution-order note on the mechanism diagnostics. The registered N(d2) figure of 0.210 was produced with the option-leg diagnostic resolved at the specification's fixed spread term. The recompute at the full fitted rate reads 0.252 on the historical marks and 0.094 on the confirmation marks, both as of the 2026-08-22 run. Both constructions clear the 0.05 floor and no verdict turns on the difference; the registered figure stands as the figure of record and the recompute publishes beside it.
Admission means one thing. The pricing construction behind the conversion weight has now priced ten dates of filed fair values, one of them sealed and unseen, within a tolerance fixed before results existed. The gate that Specification v1 section 8 holds in front of the continuous conversion weight is open.
Admission does not launch anything. No conversion-weighted figure appears on any tracker surface as a result of this study. Launch is a separate ratification, taken deliberately and on its own schedule, with this study's full disclosed record in view. The measurement layer is untouched in either direction. CEBE carries every claim at contractual value regardless of what any valuation-layer study finds.
A reader should meet the study's hardest sentence here, in the decision, not in a diagnostics table. Nine of the ten dates in the gating 4.747 were consumed by the exploratory search that produced the specification; the only data the specification had never seen, the sealed quarter, scored 5.077 against the 5.0 tolerance with a mean signed error of negative 3.5 points. The skeptic's one-line reading is that the sealed quarter missed tolerance and admission rests on in-sample dates. The registered answer is that the gating statistic was always the full-panel leave-one-date-out figure, fixed before any result existed, and the sealed quarter's role was to be scored inside that panel, which it was. Both sentences are true, both publish, and the admission retirement rule registered alongside this page exists precisely because the second sentence needs teeth. If the cheap signature persists, the admission comes down by rule, not by discretion.
This is the first admission in the calibration record. It follows two published no-pass results under the same disclosure standard: Calibration 1, where no candidate cleared the tolerance and the structural finding published, and Calibration D1, where the Deep ITM delivery prior failed its interval and the failure serves at full prominence. The record's credibility is the three results together, scored under frozen rules, published at one prominence.
The confirmation-slice signature, cheap pricing on the newest quarter, is the early-warning line. The forward cohort's Friday capture cadence accumulates the volatility record that scores future quarters as they file; if the cheap signature persists or widens across quarters, that is drift, and it will be visible in a disclosed cell before it is visible anywhere else. The leave-two-dates-out fragility stands as the known bound on data-removal robustness. The 2026-05-19 close's provenance flag retires when independently anchored, by dated note.
Registration: docs/specs/cal-3-registration-2026-08-19.md at blob 0b5e95f2. Runner, inputs, outputs, and receipt: study/cal3/confirmation/ on main, run reproduces byte-identically from the committed inputs. Sealed set: study/cal3/holdout/, diff against the sealing commit empty before and after the run. Confirmation fair values: accession 0001050446-26-000044, Note 6, re-verified at run time. Scoring commit of record: merged to main as #C-110.
Calibration record by @chcbearsfan | cebetracker.io
Registered text. Installed verbatim from the ratified draft of 2026-08-22, with the ratification date filled at the status line.
Status line: Proposed 2026-08-22, following the ADMITTED verdict of the same date. Ratified: 2026-08-22. This revision adds a retirement condition to the admission. It changes no tolerance, no parameter, and no verdict already of record.
The admission granted 2026-08-22 retires automatically if either condition holds on two consecutive filed quarters, beginning with the Q3 2026 10-Q:
Retirement is mechanical. If triggered, the section 8 gate closes, any launched conversion-weighted surface comes down within one bake cycle, and the retirement publishes as a dated note on this page at the prominence of the admission itself. Reinstatement requires a new confirmation study under a new registration; the retired admission does not revive.
Each quarter's filed fair values are scored against the frozen specification when the filing is read, and the result publishes as a dated scorecard row on this page whether it trips, approaches, or clears the conditions. Quiet quarters publish too. The scorecard is the forward record the admission stands on.
The RMSE bound reuses the registered panel tolerance of 5.0 unchanged, so the retirement threshold is the same line the study was admitted against, applied to the only data that is always out of sample, the newest quarter. The signed-error bound of negative 5.0 is symmetric with that tolerance and sits safely wide of the known starting state, the 2026-08-22 confirmation slice's negative 3.516, so the rule is not born triggered. Both are stated conventions, fixed now while the Q3 answer is unknowable, which is their entire function.
It is not a launch decision. Launch remains a separate ratification regardless of this rule's existence. It is not a re-scoring of the admission, whose verdict stands as of its date. It is not amendable after a trip: a triggered retirement executes as written, and any argument about the rule's wisdom belongs to the next registration, not to the trip.
Registered before results existed for the quarters it governs. Calibration record by @chcbearsfan | cebetracker.io
Scaffolding added 2026-08-24 with the launch. It carries one dated row per scored quarter and is empty until the Q3 2026 10-Q files. The bounds in the column headers are the retirement conditions of Revision 1, not new tolerances.
Each quarter's filed convertible fair values are scored against the frozen specification with no refit, on the same construction as the 2026-08-22 run's confirmation-slice diagnostic. The result publishes here whether it trips, approaches, or clears. Quiet quarters publish too, because a scorecard that only speaks when something breaks cannot be read as a record.
| Quarter | Filing read | Marksn | Confirmation-slice RMSEretires above 5.0 points of face | Mean signed errorretires below negative 5.0 points | Consecutive breachesretires at 2 | Status |
|---|---|---|---|---|---|---|
| Q2 2026 | 2026-08-22 | 6 | 5.077 | negative 3.516 | not scored under this rule | Baseline, not a scored row |
| Q3 2026 | not yet filed | not yet filed | not yet filed | not yet filed | not yet filed | not yet filed |
The Q2 2026 row is the confirmation slice of the admission run, carried as the starting state Revision 1 names when it explains why the signed-error bound sits at negative 5.0. It is a baseline and not a scored quarter: the scored series begins with the Q3 2026 10-Q, which is the first filing this rule governs.
Two consecutive breaching quarters retire the admission. A breach is either condition, so a quarter that clears RMSE and fails the signed-error bound counts as a breach, and a single breach followed by a clear quarter resets the count. On retirement the section 8 gate closes, every conversion-weighted surface comes down within one bake cycle, and the retirement publishes here at the prominence of the admission itself.
The surface currently at risk under this rule is the conversion weight launched 2026-08-24. It is the only one.
Ruled and recorded here because it is a statement about what this study admitted rather than about how a page renders. This admission's scope is the population it tested. Calibration 3 priced MSTR filed convertible fair values, with parameters fitted on MSTR marks, on MSTR-sourced volatility. Running the same construction over another issuer's convertible would not be an application of this admission, it would be an extension of it.
Accordingly the launched weight is sourced only: a claim carries it when the claim's own volatility input is sourced, and publishes as unweighted otherwise. An issuer joins on sourced volatility plus a dated note, which is the case where the construction is untested in its inputs alone. Whether the construction generalizes across issuers at all is untested in kind, and that would take a new registration rather than a ruling, scored on its own sealed set like this one. Registered as R-23. No verdict, tolerance or parameter of record moves.
The section 8 gate of Specification v1 opened 2026-08-22 on the ADMITTED verdict recorded above. Launch of the continuous conversion weight was ratified as its own decision on 2026-08-24, per the registration's clause that an open gate is necessary and not sufficient. Effective on deploy.
The launched surface is the conversion weight, carrying weighted claim values for the six MSTR convertible senior notes at a 2026-08-14 mark on the frozen parameters with no refit. Nothing from Calibration D1 launched with it; that study's no-pass clause stands and its delivery priors remain display only. The forward scorecard above was added in the same build.
The launch build surfaced a construction question this document does not answer and did not need to until a figure shipped. Specification v1 section 3.1 defines a convertible claim's adjusted value as a convex combination of standing value and conversion value at weight w. This study validated a different object, a convertible bond price in which w occupies one slot beside a bond floor and a discounted strike term. At the 2026-08-14 mark the two disagree by 16.627 points of face across the six notes, and the disagreement runs one way, with the specification's blend carrying the stack lighter than the model.
Recorded here because this rule scores the model while the launched page publishes the blend. The gap is published at full prominence on the launched surface and referred for a specification ruling. No verdict, tolerance, or parameter of record moves.
CEBE framework, cebetracker.io. Part of the calibration record. Preceded by Calibration 1 and Calibration 2, both of which failed, and running beside Calibration D1 on the denominator side. The gate this registration answers, and which the admission of 2026-08-22 opens, is Specification v1 section 8.