TruthSetsFree · Standing Reference

Data and Method

The full working data behind our numerical pieces, with every figure classified, every source named, and every unresolved disagreement printed rather than tidied away.

Maintained by Robby Miller · ParleyBot Intelligence · Last revised 16 October 2026 · Contains a published correction · Revision log at the foot of this page

Where an article on this site rests on a dataset rather than on reporting, the dataset lives here in full. The purpose is not decoration. It is to let a reader or a fact-checker re-derive our numbers, find the point at which they diverge from ours, and see which choices we made where the sources disagree.

Nothing on this page is an argument. Interpretation belongs in the articles; this page holds only the figures and how they were obtained.

Articles resting on this data

The line Congress saved and nobody filled — 9 August 2026. Draws on Datasets 01, 02 and 03: Tomahawk procurement across four wars, the production capacity gap, and the collapse in buying between FY2019 and FY2025.

Scoring Record — standing page, updated daily. Rests on Dataset 04. The continuity baseline that page has promised its readers is computed here for the rows this desk holds, and its limits are stated in both places.

Where an article's argument depends on a figure recorded here, that dependency is stated in the article's method note and the figure is listed above. If a number below is later corrected, any article resting on it is corrected too, and both changes appear in the revision log.

How every figure is classified

Budget quantities are not one kind of number. Mixing them silently is the most common way a procurement series manufactures a trend that does not exist. Every figure below carries one of these labels.

Requested — the quantity in a President's Budget submission. What the administration asked for.

Appropriated — the quantity Congress funded in law.

Enacted — an appropriated figure as later recorded in a subsequent budget document's prior-year column.

Delivered — units physically handed over, which can fall in a different year from the money.

Contracted — units on a signed production contract, which may combine several fiscal years and several customers, including foreign ones.

Standing rule: where a fiscal year's quantity is available both as a request in its own budget book and as a prior-year figure in a later book, we use the later book. The request is what was asked for; the later column is what happened. This is not pedantry — see the FY2021 entry below, where the two differ by 33 missiles inside a single document.

Screening rules applied to counted datasets

1. One entry per episode, not per utterance or per report. Repetitions within a single news cycle collapse to one line.

2. Retrospective narration is not a new event. Describing something already logged does not create a second entry.

3. Conditionality required where the dataset counts threats: a consequence must be attached to something the other party does.

4. Re-issue counts only on a change of register or the attachment of a fresh deadline.

5. Same-methodology denominators. Never divide one analyst's numerator by another's denominator. Where this is impossible, the ratio is reported as an order of magnitude only.

6. Classify before comparing. Requested, appropriated, enacted, delivered and contracted figures are never compared across classes.

7. Replenishment is tagged to the war that consumed the rounds, not merely to the year the money appeared. Where the two differ, both are shown.

8. Contract quantities are not national procurement quantities. Contracts routinely bundle multiple services and foreign customers.

Dataset 01 — Tomahawk procurement, FY1998 to FY2027

Quantities are new-build all-up rounds on the Navy's weapons procurement line item 2101. Modification and recapitalisation funding, which upgrades missiles already held, is a separate line and is excluded throughout.

FY	Qty	Class	Note
1998	0	Appropriated
1999	624	Appropriated	Largest single-year buy in the series
2000	0	Appropriated
2001	0	Appropriated
2002	25	Appropriated
2003	350	Appropriated
2004	322	Appropriated
2005	298	Appropriated
2006	408	Appropriated
2007	355	Appropriated
2008	496	Appropriated
2009	207	Appropriated
2010	196	Appropriated	Exactly at sustaining rate
2011	417	Appropriated
2012	196	Appropriated	Exactly at sustaining rate
2013	196	Appropriated	Exactly at sustaining rate
2014	206	Appropriated
2015	243	Appropriated	Request was 100; Congress raised it
2016	149	Appropriated
2017	196 ?	Appropriated	Disputed — see below
2018	100	Appropriated
2019	0	Appropriated	No missiles procured
2020	90	Appropriated
2021	122	Enacted	Request was 155
2022	70	Enacted	Request was 60
2023	26	Enacted
2024	30	Enacted
2025	22	Enacted	20 recorded as delivered — different class
2026	57	Requested	War begins in this fiscal year
2027	785	Requested	58 discretionary, 727 by reconciliation

Internal consistency check — passed

The Congressional Research Service states that the Navy spent 5.87 billion dollars on 4,984 Tomahawk missiles from FY1998 through FY2018. The quantities extracted above for those twenty-one years sum to exactly 4,984. The series is internally consistent with the total published by the same source.

The sustaining rate

196 missiles per year is the figure given in 2015 by Representative Randy Forbes, then chairman of the House Armed Services Seapower and Projection Forces Subcommittee, as the quantity required to sustain the production line. It is independently corroborated by the procurement pattern itself: the Navy bought exactly 196 in four separate fiscal years.

Unresolved disagreement — FY2017

Two sources give different quantities for the same year while agreeing exactly on the money.

196 missiles — Congressional Research Service, Table B-2, sourced to Department of Defense P-1 procurement budget requests.

0 missiles — commercial budget aggregator drawing on line item 2101.

Both record 297.5 million dollars appropriated. Zero new-build rounds for that sum is implausible at prevailing unit costs, so we carry 196 and mark it disputed. This has not yet been checked against the primary exhibit and should not be treated as settled. A checker wanting to resolve it should pull the P-40 justification exhibit for line item 2101 in a budget book of FY2019 or later and read the FY2017 prior-year column.

A caution about the corroborating source

The commercial aggregator used to extend the series past FY2021 reproduces the Congressional Research Service dollar figures exactly on every overlapping year, and the quantities on seven of eight. But its page metadata is corrupted — it carries a programme description belonging to a different missile and misattributes the budget account. We have used it only where its figures could be cross-checked, and never as a sole source.

Staleness note, 16 October 2026. Datasets 01, 02 and 03 were last revised on 9 August and have still not been re-checked. The war they describe has continued daily throughout that period, and the expenditure figures in Dataset 02 in particular are now more than two months out of date. Nothing below has been shown to be wrong; it has simply not been re-examined, and a reader should treat every 2026 figure in those three datasets as current to 9 August rather than to today.

Dataset 02 — Rounds fired, by operation

These figures come from operational reporting, not from budget documents. They therefore sit in a different evidentiary class from Dataset 01, and any ratio between the two is reported as an order of magnitude only, never as a precise quotient. Rule 5 cannot be satisfied here and we do not pretend otherwise.

Operation	Fired	Basis
Desert Storm, 1991	288	Think-tank compilation of reported strike data. A congressional research account gives 290, launched from 12 submarines; a third cites over 300 Block II variants including 27 of the submunition type. Range 288 to 300.
Desert Fox, Dec 1998	325	Same compilation. A 70-hour campaign. An outlying report of 50 is unsupported elsewhere and is discarded.
Allied Force, 1999	218	Serbia and Montenegro.
Iraqi Freedom, 2003	802	Think-tank compilation. Richard Myers, as chairman of the Joint Chiefs, gave over 725 at the time. Both printed; the gap is likely invasion-phase against full-campaign scope.
Odyssey Dawn, 2011	112	US launches. Also reported as at least 161 and as 196 for US and UK combined.
Epic Fury, 2026	850+	First four weeks, 28 February to 27 March, per a national newspaper investigation citing people familiar with the matter. Later raised — see below.

Epic Fury is the largest Tomahawk campaign on record

More than 850 rounds in four weeks exceeds Iraqi Freedom's 802 across an entire invasion, which held the previous record. That comparison is drawn by the same think-tank analysis supplying the historical figures above, so numerator and denominator share a methodology and Rule 5 is satisfied for this particular claim.

Boundary warning on the 2026 figure

Three figures are in circulation and they do not measure the same thing.

More than 850 in the first four weeks, 28 February to 27 March. Newspaper investigation, sourced to people familiar with the matter. War-bounded and dated.

More than 1,000 by the ceasefire. Originally a financial newspaper figure, subsequently relayed in a think-tank assessment. Attribution note: this figure is often credited to the think tank, which was quoting the newspaper. We have made that error ourselves in an earlier edition and correct it here.

Close to 1,000 across strikes on Iran, operations in Yemen and the Red Sea, Nigeria and other conflicts since June 2025 — an eighteen-month, multi-theatre aggregate.

The third cannot be attributed to this war and is excluded under Rule 7. The first two are compatible: 850 in the opening month, passing 1,000 by the ceasefire. We use 850 where a four-week rate is the point and the later figure where cumulative depletion is.

Dataset 03 — Production rates, lead times and replenishment

Withdrawn figure. This dataset previously stated a maximum production capacity of about 2,330 missiles a year, built from three prime-contractor lines at 600 each plus a second manufacturer at 530, and derived from it a claim that actual buying ran at four per cent of capacity. No such figure appears in any source. It was not a misreading of a real number; it was not supported at all, and it should never have been published. It has been removed here and the article resting on it has been corrected. See the revision log.

Measure	Value	Basis and class
Recent annual production	under 200	Think-tank assessment, May 2026. Cause given as small past orders. Actual.
US-specific annual rate	~60–100	Wire service gives about 60 for the United States; a defence outlet gives 90 to 100. Actual, contested range.
Procurement average FY15–FY26	86	Same think-tank assessment, most orders from the Navy. Our Dataset 01 gives about 92 across that span; the gap is consistent with Navy-only against all-service. Actual.
Target production capacity	1,000+	Framework agreements signed 4 February 2026, up to seven years. Goal, not capability. Must never be used as a current-capacity denominator.
Production lead time	34 months	Department of Defense delivery projections.
FY2027 buy first deliveries	March 2030	Same projections.
Full pre-war replenishment	late 2030	Think-tank estimate.
Foreign orders on the same line	~800	Japan, Australia, Netherlands. Absent from US procurement documents.

The two competing sustainment figures

Roughly 90 a year is commonly given as the minimum needed to keep the line running. 196 traces to a 2015 argument by the chairman of the responsible House subcommittee, made in support of a funding increase — an advocacy figure, not a budget-document rate. It is corroborated only in the weak sense that the Navy bought exactly 196 in four separate years, which is as consistent with contract lot sizing as with a technical floor. Both are printed; neither is treated as settled.

Inventory estimates

Pre-war stockpile is given as the low 3,000s by one think tank and as 4,000 to 4,150 by a defence outlet. The Department of Defense does not disclose it. These are outside estimates and published depletion percentages depend entirely on which is chosen.

A boundary discrepancy in Dataset 01, unresolved

Two outlets give FY2023 Tomahawk procurement as 55. Dataset 01 gives 26, from the Navy weapons procurement line.

The likeliest explanation is scope: a single FY2023 contract modification for 111 Block V rounds split 50 to the Army, 48 to the Navy and 13 to the Marine Corps, so non-Navy buys were the majority of that lot. A Navy-only series will therefore understate national procurement from roughly 2023 onward, when land-based variants entered service.

We have not resolved this and it is a real limitation on Dataset 01's later years. Any ratio using a Navy-only numerator against an all-programme denominator breaches Rule 6 and is not attempted here.

Dataset 04 — The prediction ledger, and the continuity baseline

The Scoring Record has promised its readers a naive continuity baseline: what a desk that simply predicted continuity every time would have scored on the same panels. That page cannot compute it, because it holds block anchors rather than individual grades. This dataset is where the row-level ledger is kept. It is incomplete and its scope is stated exactly.

Classification used

Continuity — the proposition asks whether an existing state of affairs will carry on.

Change — the proposition asks whether somebody will do, say or publish a new thing.

Institutional — a sub-class of continuity in which the state of affairs is maintained by a body declining to act.

Price — a sub-class of continuity in which the state of affairs is a number remaining on one side of a stated line. Separated from institutional continuity on 3 September, after a board on which the two behaved differently.

Affirmative — a grade resting on a positive published act by the body named in the call.

Capped — a grade resting on the absence of an event rather than on evidence of its absence, and limited accordingly.

Hit — a grade of 6 or more. A grade of 5 is a miss.

Reference class — the kind of event whose occurrence decides a prediction, whichever direction the prediction is written in. Fifteen classes are used, listed below. They were assigned on 16 October to every row already graded, after the grades were known; no grade, likelihood or wording was changed in doing so, and from that date each new prediction is classed before it is priced.

Scope, stated plainly

Rows exist for 95 of the 417 finalised predictions, or about 23 per cent: the 16 graded from 3 to 6 September, and the 79 graded from 26 September to 16 October, for which this desk holds the weight, standalone likelihood, falsifier and grade in its own ledger. The predictions graded from 7 to 25 September, and everything graded before 3 September, are recorded on this site only inside their own editions and have not yet been extracted. Until they are, no lifetime baseline can be computed and none is claimed.

Rows graded 3 to 6 September

Graded	Panel	Proposition	Wt	SA	Kind	Grade
3 Sep	27 Aug	The keynote gives nothing away	30	n/p	Continuity, institutional	6 Hit
3 Sep	27 Aug	September pricing stays below sixty	26	n/p	Continuity, price	2 Miss
3 Sep	27 Aug	The British officials stay	24	n/p	Continuity, institutional	7 Hit
3 Sep	27 Aug	The revenue split does not become a text	20	n/p	Continuity, institutional	7 Hit
4 Sep	28 Aug	The corridor does not open	26	50	Continuity, institutional	7 Hit, capped
4 Sep	28 Aug	No goods through Erez	22	45	Continuity, institutional	7 Hit, capped
4 Sep	28 Aug	Diesel holds above 5.60	24	65	Continuity, price	5 Miss
4 Sep	28 Aug	The Russian ban does not lapse	28	80	Continuity, institutional	8 Hit, affirmative
5 Sep	29 Aug	The two-year holds above 4.25	26	70	Continuity, price	8 Hit, affirmative
5 Sep	29 Aug	No American-Iranian talks	24	80	Continuity, institutional	6 Hit, capped
5 Sep	29 Aug	No named West Bank operation	24	60	Continuity, institutional	7 Hit, capped
5 Sep	29 Aug	The branches stay open	26	45	Continuity, institutional	8 Hit, affirmative
6 Sep	30 Aug	The December date holds	28	60	Continuity, institutional	8 Hit, affirmative
6 Sep	30 Aug	The wage is not revised	26	80	Continuity, institutional	6 Hit, capped
6 Sep	30 Aug	No strike inside Iran	24	70	Continuity, institutional	5 Miss
6 Sep	30 Aug	No Taiwan exercise announced	22	85	Continuity, institutional	6 Hit, capped

Wt is the panel weight, SA the standalone likelihood. Where the standalone is marked n/p it was published in the edition but is not recorded in the summary this desk holds for that board, and the row is incomplete pending extraction from the edition itself.

Rows graded 26 September to 16 October

Graded	Panel	Proposition	Wt	SA	Kind	Grade
26 Sep	19 Sep	The delegation appears	27	60	Change, diplomatic contact	8 Hit
26 Sep	20 Sep	No American–Iranian meeting	27	78	Continuity, diplomatic contact	3 Miss
26 Sep	22 Sep	The aviation measures land on their date	24	70	Change, trade or sanctions measure	8 Hit
27 Sep	20 Sep	Washington stays out of the southern war	26	72	Continuity, strike announced	7 Hit
27 Sep	20 Sep	Outside the region: the authorisation lapses unannounced	25	62	Continuity, scheduled process	6 Hit
27 Sep	20 Sep	Outside the region: the long end stays under five	22	64	Continuity, market threshold	2 Miss
28 Sep	21 Sep	Washington does not answer the conditions	30	60	Continuity, diplomatic contact	3 Miss
28 Sep	21 Sep	Outside the region: no fourth straw poll	22	62	Continuity, scheduled process	6 Hit
28 Sep	22 Sep	The fund stays a document	30	72	Continuity, on record statement	6 Hit
28 Sep	22 Sep	No moratorium is agreed	26	85	Continuity, agreement concluded	6 Hit
29 Sep	22 Sep	Outside the region: no coalition in Schwerin	20	88	Continuity, agreement concluded	7 Hit
29 Sep	23 Sep	The second meeting is publicly confirmed to have happened	30	55	Change, diplomatic contact	4 Miss
29 Sep	23 Sep	Tehran does not lower its price	26	82	Continuity, policy reversal	6 Hit
29 Sep	23 Sep	Outside the region: no moratorium for Ukraine either	20	88	Continuity, agreement concluded	7 Hit
30 Sep	23 Sep	Beirut stays out of it for another week	24	75	Continuity, strike announced	6 Hit
30 Sep	24 Sep	The hundred-dollar floor holds	30	65	Continuity, market threshold	7 Hit
30 Sep	24 Sep	The line stays unconfirmed for another week	26	65	Continuity, on record statement	6 Hit
1 Oct	24 Sep	The bombs are not formally notified	24	75	Continuity, trade or sanctions measure	6 Hit
1 Oct	24 Sep	Outside the region: the curbs stay voluntary	20	70	Continuity, trade or sanctions measure	6 Hit
1 Oct	25 Sep	The strait is not declared open	30	85	Continuity, operational status	6 Hit
1 Oct	25 Sep	No impeachment vote on Araghchi	26	85	Continuity, legislative vote	6 Hit
1 Oct	25 Sep	The Palestinian ballot takes candidates	24	85	Change, scheduled process	8 Hit
1 Oct	25 Sep	Outside the region: no loan for Caracas yet	20	78	Continuity, institutional decision	6 Hit
3 Oct	26 Sep	No American strike on Iran before the week is out	30	78	Continuity, strike announced	6 Hit
3 Oct	26 Sep	No inspectors inside yet	26	92	Continuity, operational status	6 Hit
3 Oct	26 Sep	The pact stays defensive	24	80	Continuity, strike announced	6 Hit
3 Oct	26 Sep	Outside the region: no new tariff on China	20	90	Continuity, trade or sanctions measure	6 Hit
4 Oct	27 Sep	The Palestinian vote keeps its date	30	80	Continuity, scheduled process	7 Hit
4 Oct	27 Sep	No Iranian attack on an American warship	26	80	Continuity, strike announced	6 Hit
4 Oct	27 Sep	No Israeli–Syrian meeting confirmed	24	78	Continuity, diplomatic contact	6 Hit
4 Oct	27 Sep	Outside the region: the ten-year holds above five	20	70	Continuity, market threshold	7 Hit
5 Oct	28 Sep	The Ramallah revocation stands	30	78	Continuity, policy reversal	6 Hit
5 Oct	28 Sep	The Abu Dhabi meeting stays unconfirmed	26	65	Continuity, diplomatic contact	3 Miss
5 Oct	28 Sep	The blockade stays on	24	90	Continuity, operational status	7 Hit
5 Oct	28 Sep	Outside the region: Kyiv does not apologise to Seoul	20	65	Continuity, diplomatic contact	6 Hit
6 Oct	29 Sep	Washington marks the exit from Iraq	30	65	Change, institutional decision	8 Hit
6 Oct	29 Sep	Abu Dhabi does not deny the warning	26	85	Continuity, on record statement	6 Hit
6 Oct	29 Sep	Riyadh keeps its seat in Abu Dhabi unconfirmed	24	88	Continuity, diplomatic contact	7 Hit
6 Oct	29 Sep	Outside the region: the Security Council holds its fourth straw poll	20	80	Change, scheduled process	8 Hit
7 Oct	30 Sep	Tehran answers on the record	30	55	Change, diplomatic contact	8 Hit
7 Oct	30 Sep	Tehran claims another ship	26	70	Change, attack claimed	3 Miss
7 Oct	30 Sep	No date for the end of the war	24	85	Continuity, on record statement	6 Hit
7 Oct	30 Sep	Outside the region: Spain's indefinite-lease decree falls	20	75	Continuity, legislative vote	7 Hit
8 Oct	1 Oct	No charge yet at Fairford	30	75	Continuity, court or prosecution	7 Hit
8 Oct	1 Oct	London calls in Tehran's envoy	26	55	Change, diplomatic contact	4 Miss
8 Oct	1 Oct	The alliance files its list	24	60	Change, scheduled process	3 Miss
8 Oct	1 Oct	Outside the region: Milwaukee ends without a joint text	20	50	Continuity, agreement concluded	4 Miss
9 Oct	2 Oct	Abu Dhabi names no sponsor	30	75	Continuity, on record statement	7 Hit
9 Oct	2 Oct	Islamabad and Ankara do not name Medina as the trigger	26	80	Continuity, on record statement	3 Miss
9 Oct	2 Oct	Ra'am is back on the ballot	24	80	Change, court or prosecution	8 Hit
9 Oct	2 Oct	Outside the region: Berlin and Paris hold their diesel	20	50	Continuity, institutional decision	5 Miss
10 Oct	3 Oct	Diesel at the pump eases	30	75	Change, market threshold	8 Hit
10 Oct	3 Oct	Tehran answers Manchester	26	65	Change, on record statement	3 Miss
10 Oct	3 Oct	No date for the Rome round	24	60	Continuity, scheduled process	6 Hit
10 Oct	3 Oct	Outside the region: Brazil goes to a second round	20	85	Continuity, scheduled process	7 Hit
11 Oct	4 Oct	No American strike on Iran this week	30	50	Continuity, strike announced	7 Hit
11 Oct	4 Oct	Three carriers stay off the record	26	75	Continuity, on record statement	6 Hit
11 Oct	4 Oct	No American strike on the Houthis this week	24	50	Continuity, strike announced	6 Hit
11 Oct	4 Oct	Outside the region: Riga takes its time	20	88	Continuity, agreement concluded	6 Hit
12 Oct	5 Oct	The Taiz road stays cut	30	55	Continuity, operational status	7 Hit
12 Oct	5 Oct	No confidence vote on an oil minister	26	88	Continuity, legislative vote	7 Hit
12 Oct	5 Oct	Regev keeps her ministry	24	90	Continuity, office holder change	7 Hit
12 Oct	5 Oct	Outside the region: Okinawa gets new limits	20	60	Change, institutional decision	8 Hit
13 Oct	6 Oct	The promise stays a promise this week	30	50	Continuity, operational status	6 Hit
13 Oct	6 Oct	Ankara does not ratify this week	26	85	Continuity, legislative vote	7 Hit
13 Oct	6 Oct	The Palestinian nomination deadline holds	24	78	Continuity, scheduled process	6 Hit
13 Oct	6 Oct	Outside the region: the warrants stand	20	88	Continuity, court or prosecution	2 Miss
14 Oct	7 Oct	Washington does not go back to zero this week	30	55	Continuity, on record statement	6 Hit
14 Oct	7 Oct	No one sits across the table this week	26	75	Continuity, diplomatic contact	7 Hit
14 Oct	7 Oct	London answers Jerusalem in kind	24	55	Change, diplomatic contact	4 Miss
14 Oct	7 Oct	Outside the region: Nairobi holds at one	20	70	Continuity, count threshold	7 Hit
15 Oct	8 Oct	The seven stay at their desks	30	80	Continuity, policy reversal	6 Hit
15 Oct	8 Oct	Doha names nobody	26	70	Continuity, on record statement	6 Hit
15 Oct	8 Oct	Damascus keeps its troops at home this week	24	50	Continuity, operational status	6 Hit
15 Oct	8 Oct	Outside the region: the torture counts are filed	20	75	Change, court or prosecution	8 Hit
16 Oct	9 Oct	Tehran sends its answer	30	65	Change, diplomatic contact	3 Miss
16 Oct	9 Oct	The blockade outlasts the pledge's first week	26	92	Continuity, operational status	7 Hit
16 Oct	9 Oct	Beirut keeps the general	24	85	Continuity, court or prosecution	6 Hit
16 Oct	9 Oct	Outside the region: Seoul's envoy stays recalled	20	88	Continuity, diplomatic contact	7 Hit

The panel is the date of the edition that published the prediction; the graded date is its close. For these rows this desk's ledger records the grade but not whether a hit rested on an absence or on a positive act, so the Kind column gives the change-or-continuity flag and the reference class instead of the capped and affirmative labels.

The baseline, so far as it can be computed

On the 16 rows graded 3 to 6 September, every prediction was a continuity proposition, so a desk that predicted continuity every time would have produced an identical board: thirteen hits and three misses. That finding stands as published on 6 September.

On the 79 rows graded from 26 September to 16 October, 62 predictions were continuity propositions and 17 were change propositions. On the 62, the continuity-always desk makes the same call and gets the same result: 54 hits. On the 17 it takes the other side, so it hits exactly where this desk missed. This desk hit 10 of the 17; the continuity-always desk would have hit seven. Across the 79, this desk has 64 hits and the continuity-always desk 61.

Across all 95 rows held, this desk has 77 hits and the continuity-always desk 74.

The comparison is made in hits, not in the 0–10 mean, because the opposite of a change proposition was never written or graded and so has no grade to average.

The honest reading: a margin of three hits in 95, all of it earned on 17 change propositions, is too small to show that the analysis adds anything a continuity-always desk would not. Nor does it show that it adds nothing. The change propositions are where the difference would show, and there are not yet enough of them to tell.

Two limits on that finding. First, the rows are not a random sample of the record: they are an early-September week and the most recent three weeks, with the weeks between not yet extracted, so they cannot be extrapolated to the lifetime record. Second, a continuity-always desk still requires somebody to choose which continuity to price. Selecting the propositions is work the baseline does not capture, so what is measured here is answer-selection rather than question-selection.

Reference classes

The fifteen classes, with the number of graded rows from 26 September to 16 October in each and how many came true. A class with fewer than ten graded rows is too thin to read.

Class	What it covers	Graded	Came true
Diplomatic contact	a meeting, talks, delegation, reply delivered, demand answered, envoy summoned, expelled, recalled or returned, apology	14	7
On record statement	a named body states a specified fact on the record: names or denies a culprit, confirms a fact, sets a date, discloses a plan	10	8
Strike announced	a named state's strike, attack or missile launch is announced or reported by its own side, the target or a named official body	7	7
Attack claimed	an armed group or state claims an attack, or a monitoring body reports attacks against a stated floor	1	0
Market threshold	a published price, yield or series closes or settles on one side of a stated line	4	3
Legislative vote	a parliament or assembly holds, passes or fails a vote, or validates a decree	4	4
Court or prosecution	a court rules, or a prosecutor or police body charges, files or opens a case	5	4
Trade or sanctions measure	a tariff, export control, sanction, licence or statute is imposed, revoked, notified or takes effect	4	4
Agreement concluded	a truce, ceasefire, moratorium, coalition agreement, joint text or accepted offer is announced	5	4
Institutional decision	a body adopts a new decision, approval, mission, loan or restriction that is not a reversal of its own standing one	4	3
Policy reversal	a government withdraws, softens or reverses a standing decision, refusal or term of its own	3	3
Operational status	a blockade, route, airport, flow, facility, deployment, inspection or production level starts, stops or resumes	7	7
Scheduled process	a dated procedure (election, filing, deadline, straw poll, round of talks) occurs, holds or slips	9	8
Office holder change	a named office-holder resigns, is removed, dismissed or killed	1	1
Count threshold	an official count (cases, casualties, displaced) crosses a stated number	1	1

What completing this dataset requires

One row per finalised prediction, carrying: edition, run number, date written, proposition as published, panel weight, standalone likelihood, named falsifier, series or body relied on, close date, grade, grading date, hit flag, the change-or-continuity classification and the reference class. The change-or-continuity field is the whole point of the baseline. The rows for predictions graded from 7 to 25 September will be extracted next, then the earlier ones; they exist only inside published editions and must be extracted from them rather than reconstructed from memory or from the Scoring Record's block anchors. This page is revised at each monthly review, and whenever forty more rows are held, and each revision is logged below.

Where to look next

Primary, quantities. Navy weapons procurement justification books, exhibit P-40, line item 2101. Published annually. Read the prior-year column of a later book, not the request column of the year in question.

Primary, programme history. Selected Acquisition Reports for the Tactical Tomahawk programme, which record contract lots, foreign military sales content and supplemental replenishment.

Secondary, consolidated. Congressional Research Service report R45996 on precision-guided munitions, appendices A and B, covering FY1998 to FY2021 by service and by programme.

Gaps we have not closed. Army procurement lines for land-based Tomahawk, which sit outside the Navy books entirely; foreign military sales quantities, which consume the same production capacity and appear in neither.

Revision log

16 October 2026 — fifth revision. Dataset 04 extended from 16 rows to 95: the 79 predictions graded from 26 September to 16 October added from this desk's ledger, each with its change-or-continuity flag and a reference class. A hit defined on this page as a grade of 6 or more. The continuity baseline recomputed on the rows held: this desk 77 hits, a continuity-always desk 74, a margin too small to read either way. Fifteen reference classes added to the classification, with graded rows and hits in each; the classes were assigned after grading, and no grade, likelihood or wording was changed. Staleness note on Datasets 01 to 03 re-dated; they have still not been re-checked. Rows for predictions graded from 7 to 25 September are not yet extracted.

6 September 2026 — fourth revision. Dataset 04 created: the row-level prediction ledger with change-or-continuity classification, covering the sixteen predictions graded from 3 September onward, being about six per cent of the finalised record. A partial continuity baseline computed and published, showing that over this sample the desk's score and a naive continuity-always score are identical, with both limits on that finding stated. Staleness note added to Datasets 01 to 03, which have not been re-checked since 9 August. Article index extended to the Scoring Record. A duplicated section heading removed.

8 August 2026 — created. Dataset 01 and Dataset 02 established. FY2017 flagged as disputed. FY2023 and FY2024 added from a commercial aggregator pending primary confirmation.

9 August 2026 — third revision, correction. Dataset 03 withdrawn and rebuilt. A stated maximum production capacity of about 2,330 a year, and the derived claim that procurement ran at four per cent of capacity, were unsupported by any source and have been removed. The article resting on them carries a correction notice. Replaced with sourced production rates, the 34-month lead time and the March 2030 first-delivery projection. The 196 sustainment figure downgraded from benchmark to contested advocacy figure alongside the more commonly cited 90. A boundary discrepancy in Dataset 01's FY2023 entry recorded as unresolved.

9 August 2026 — second revision. Article index added. First article resting on this data published.

8 August 2026 — first revision. Dataset 02 corrected against a think-tank compilation of strike data: Desert Fox moved from contested to 325, the outlying figure of 50 discarded; Iraqi Freedom given as 802 alongside the 725 attributed to the chairman of the Joint Chiefs; Desert Storm given as 288 with the range to 300 stated. The 2026 figure re-dated to a four-week window. Attribution of the later expenditure figure corrected from the think tank to the newspaper it was quoting, an error this desk had made in print. Dataset 03 added on production capacity, sustaining rate and actual procurement, with inventory estimates.

Corrections to this page are logged here rather than made silently. If you find an error, the fastest route to a fix is to say which figure and which source.

The approach, the six coverage domains and our scoring record — graded daily and reviewed each month — are set out on the About page.

No financial advice is expressed or implied.