QJM: scope and evidence.

Professional wargaming and comparison of authored courses of action in educational and analytical settings, under declared assumptions. Analysts remain responsible for interpreting the result and its applicability.

Excluded: operational casualty or outcome prediction, calibrated probabilities of real-world success, and claims that a historical example establishes general fitness for a new engagement.

This card covers ARGUS QJM engine version 0.1.0: qjm-npw-1977, based on the 1979 edition of Trevor N. Dupuy’s Numbers, Predictions and War, and the separate qjm-manual lineage from the DMSI 1986 manual.

The identifier’s “1977” is a lineage name, not this card’s publication date. These QJM reference models are distinct from ARGUS MDO and ACE. Their evidence and results must not be blended across engines or versions.

Calculations depend on declared force composition, personnel, weapon effectiveness values, mobility, combat effectiveness and environmental or posture factors. The source modes use different factor chains. Preserve the source citation, assumptions and any derived or substituted values with each comparison.

Published worked-example inputs, analyst-authored inputs and synthetic demonstration inputs have different evidentiary roles. Reproducing a published example does not validate a new scenario’s inputs.

NPW verification compares the reduced calculation chain with published Austerlitz and Waterloo worked examples (1979 edition, printed pp. 151–156). The full factor chain and user scenarios have a narrower, structural evidence basis. Manual-mode evidence checks the implementation against its source and reference port; it does not establish predictive performance.

Software correctness, agreement with a published calculation, historical benchmarking and suitability for an intended use are separate questions. A passing software check or a formula-level VALIDATED label is not a claim of calibrated operational accuracy. Consult the ledger’s specific engine, version, dataset, split and claim guidance before citing an evaluation.

These QJM calculations are deterministic for declared inputs; a single ratio does not quantify real-world uncertainty. Where another ARGUS workflow adds parameter variation or a stochastic ensemble, its spread is a sensitivity envelope under those assumptions. It is not a calibrated predictive interval or a bound on an adaptive opponent.

• Incomplete, substituted or ambiguous inputs can change the interpretation. Read each output’s basis, citations and warnings.
• Source contradictions and transcription uncertainty require review; CHECK and SME_REQUIRED are not cleared evidence.
• Missing source support can produce BLOCKED_SOURCE and a null value. A missing result is not zero.
• The strict QJM casualty module is blocked. Appendix B output-only cases do not supply the inputs needed to reconstruct a complete engagement.
• Historical scope and model structure limit transfer to new conditions. Independent critiques examine dimensional consistency, combat-power ratios and ground gain.

• Trevor N. Dupuy, Numbers, Predictions and War: Using History to Evaluate Combat Factors and Predict the Outcome of Battles, Bobbs-Merrill, 1979 edition; worked examples, printed pp. 151–156; full-chain equation 12.
• DMSI, QJM Manual of Rules and Procedures, 1986; equations C-29–C-38 and tables 2–20. This is a separate source lineage.
• Jerome Adrian Clark, Dimensional Analysis of the Quantified Judgement Model, Naval Postgraduate School, June 1989. Original thesis (PDF).
• Joseph F. Ciano, The Quantified Judgment Model and Historic Ground Combat, Naval Postgraduate School, September 1988. Original thesis record.

5 October 2026 — first public card, consolidating existing QJM implementation documentation and intended-use boundaries. No engine, calibration or evaluation result changed with this publication.

For an evaluation discussion, bring the intended use and ask for the matching model version, input provenance, assumptions and ledger record.