Expectation-Constrained Market Inference | Haruspex Trading Co., Inc.

Authorial note and replication boundary

Authorship of the methodological structure, implementation framing, and research posture remains with Enrique Gutierrez. This public edition discloses ECMI's governing thesis, defined research objects, validation posture, aggregate findings, failure analyses, and the decisions through which the methodology evolved. It is intended to permit serious evaluation of the research without providing the production recipe required to reproduce Haruspex systems.

The restricted implementation boundary is deliberately narrow. Haruspex does not disclose the production methods used to measure and structure sentiment, generate expectation or forecast points, construct trend levels, or parameterize the models and thresholds that govern those operations. Expectation lifecycle, fulfillment and non-fulfillment logic, market-location semantics, authority hierarchy, validation design, aggregate findings, risk principles, and execution-governance concepts are described more openly because they explain the contribution without exposing the mechanisms that create the underlying state.

Disclosure boundary. Public findings establish what was observed and what decisions followed. Controlled diligence may provide detailed results, benchmarks, chronological partitions, statistical tests, and selected audit artifacts under NDA. Production sentiment measurement, forecast-point generation, trend-level construction, feature definitions, weights, parameters, thresholds, prompts, and exact mappings remain restricted.

Identifiers follow ECMI-{kind}.{YYYY.MM} for paper and analysis editions and ECMI-C.{slug}.{YYYYMMDD} for candidate freezes. This public page is ECMI-P.2026.08.1. Kind letters: P public paper, R internal manuscript, A analysis artifact, C candidate freeze. A git hash beside an ID is a pin, not a recipe.

Not investment advice. Nothing in this document is an offer, a solicitation, or a guarantee of profitability. Internally observed results are not independently reproduced unless explicitly labeled otherwise. Gross historical outcomes are not described as profitable performance.

Acknowledgment of Strategic and Historical Contribution

David Hoverman provided long-term strategic participation, market-domain dialogue, and continuity of engagement throughout the SynQ, JARVIS, and PHALANX architecture processes. While the methodological structure, authorship, and implementation framing of this paper remain the responsibility of the author, David's sustained involvement over multiple years contributed valuable contextual perspective, challenge, and durable support to the evolution of the underlying research program.

Claimed contribution

Haruspex claims methodological priority over the integrated use of persistent discourse-derived expectation primitives, dual fulfillment and non-fulfillment state construction, joint local and macro behavioral market coordinates, lifecycle-conditioned conviction, trend-first evidence authority, and execution eligibility within a single constraint-preserving inference chain.

Individual components, including sentiment analysis, regime detection, support and resistance modeling, analogue analysis, and execution gating, have established precedents. The claimed contribution lies in their integration around persistent expectation lifecycle and in the use of both fulfilled and unfulfilled behavioral expectations to construct the state through which subsequent evidence is interpreted.

The contribution is not the independent invention of sentiment analysis, regime classification, support and resistance, analogue comparison, or execution gating. It is the construction of a single stateful inference process in which behavioral expectations persist, market acceptance and refusal both modify state, joint location conditions comparability, and no directional conclusion becomes executable merely because it can be calculated.

Evidence availability. This public paper reports the methodology, evaluation structure, selected aggregate findings, and research decisions that define ECMI. Qualified parties may request controlled review under NDA of detailed benchmarks, chronological partitions, statistical analyses, cost sensitivities, ablations, and timestamped evaluation artifacts. The production methods used to measure sentiment, generate forecast points, and construct trend levels remain restricted.

Established component	ECMI integration
Sentiment analysis	Becomes an anticipatory field that must earn evidential status through realization. [2]
Forecasting	Produces persistent expectations with lifecycle rather than disposable point predictions.
Support and resistance	Includes trailing pressure derived from unresolved expectations.
Regime analysis	Separates current local acceptance from macro unresolved pressure. [3]
Analogue comparison	Conditions similarity on compatible joint state and lifecycle.
Directional scoring	Produces bounded conviction subordinate to admissibility.
Risk management	Remains separate from conviction magnitude. [6]
Execution	Requires a governed ruling, including abstention.

Abstract

Sentiment forecasting often collapses too early. Discourse is reduced to polarity or a scalar score; a predicted event is discarded after it succeeds or fails; and market structure is reconstructed from realized price without preserving the expectation field that preceded acceptance or refusal. Those practices erase the lifecycle through which behavioral evidence acquires, loses, or changes meaning.

Expectation-Constrained Market Inference (ECMI) materializes resonant behavioral states as persistent, adjudicable expectation primitives. Fulfilled expectations contribute to current local regime structure. Unfulfilled expectations remain as trailing macro constraint topology. Their intersection is a joint behavioral market location. Directional evidence is interpreted only inside that location and its lifecycle maturity. Local acceptance holds primary directional authority. Fit metrics remain subordinate diagnostics. Conviction, risk allocation, execution eligibility, and size remain separate objects. Abstention is a valid terminal conclusion.

Internal evaluations have quantified lifecycle-dependent outcome differences and identified failure modes in authority assignment. This public edition reports the evaluation structure, selected aggregate findings, and resulting research decisions. Detailed fold-level results, benchmark tables, statistical analyses, and audit artifacts may be reviewed by qualified parties through controlled diligence, while the production mechanisms that generate sentiment state, forecast points, and trend levels remain restricted.

ECMI is therefore presented not as a theoretical architecture awaiting first contact with data, but as an evolving research methodology whose current form reflects both observed successes and components that failed to justify promotion.

The market problem

Three recurring deficiencies prevent ordinary sentiment pipelines from becoming inference. First, sentiment is often reduced to polarity, intensity, or a scalar forecast. Second, a predicted event is commonly discarded after it succeeds or fails, so delayed response, absorption, and unresolved pressure disappear. Third, market structure is usually reconstructed from realized price without preserving the expectation field that preceded acceptance or refusal. These practices erase the lifecycle through which behavioral evidence acquires, loses, or changes meaning.

The research therefore had to observe what discourse existed, when it existed, which instrument or topic it concerned, whether comparable states historically resonated with later market behavior, which expectations were formed, which were fulfilled, which remained unresolved, and where current market behavior sits relative to those histories. ECMI was developed to preserve precisely what those approaches erase: the dated behavioral expectation, the conditions under which it was formed, the manner in which the market accepted or refused it, and the state that remained after resolution. The methodology begins from the proposition that this history is itself market information.

Behavioral thesis

When discourse-derived behavioral conditions are preserved under strict temporal semantics, materialized as persistent expectations, and adjudicated against subsequent market realization, the resulting fulfillment and non-fulfillment histories contain state information that cannot be represented adequately by sentiment polarity or price structure alone.

1. Content exposure 2. Perception shift 3. Expectation formation 4. Execution behavior 5. Market realization 6. Expectation resolution 7. Updated perception

ECMI does not assert deterministic individual conformity or structural causality. It asserts that aggregate behavior may produce recurrent, measurable differences under comparable state conditions. Sentiment is an anticipatory field rather than an instruction to trade. [1] A behavioral state becomes evidence only through its relationship with subsequent realization under governed horizons.

Falsifiable thesis. If expectation lifecycle contains no incremental market information, then fulfillment and non-fulfillment histories should not produce stable differences in subsequent regime behavior, friction, path capacity, outcome timing, or execution quality after comparison with simpler price-only and direction-only baselines.

ECMI succeeds as a research program only if those differences remain measurable under chronological evaluation, controlled horizons, null-preserving rules, realistic implementation assumptions, and comparison with simpler alternatives. Architectural coherence is not treated as evidence of predictive validity.

What the research established

The following findings distinguish methodological propositions from internally observed results. Public aggregate evidence should be attached where verification is complete; detailed supporting analyses may remain in the controlled diligence package. A finding that has not survived the stated evaluation standard must remain labeled as an active hypothesis.

1. Behavioral fields acquire evidence through realization

Methodological proposition

Observation: Measured sentiment has no independent entitlement to influence conviction.
Interpretation: Evidential status is conditional on time-respecting relationships with subsequent market behavior.
Decision: Resonance remains a gate. Raw sentiment cannot authorize conviction.
Consequence: Later stages consume only states that have been tested against realization under governed horizons.

2. Expectations must persist

Methodological proposition

Observation: Transient scores delete timing, delayed response, absorption, rejection, and unresolved pressure.
Interpretation: Those quantities are state, not residual error.
Decision: Expectations are stored as typed, time-bounded, adjudicable primitives.
Consequence: Fulfillment and non-fulfillment become first-class histories.

3. Non-fulfillment is market information

Internally observed finding

Observation: The market's refusal to fulfill a measurable expectation identifies pressure that did not convert under the governing conditions.
Interpretation: Deleting that refusal as a failed forecast erases constraint.
Decision: Unfulfilled expectations remain in state and construct trailing macro topology.
Consequence: SRSS encodes unresolved behavioral pressure rather than a discarded prediction.
Evidence: Internally observed. Infinity-stone / timestone objects appear in-repo from 2019-05-28 (synq-net 02503efb); SRSS feed APIs from 2024-02-14. Class-level friction and path-capacity tables remain in controlled diligence.

4. Local and macro structure arise from different histories

Methodological proposition

Observation: Fulfilled expectations update current local regime. Unfulfilled expectations update trailing constraint topology.
Interpretation: The objects have different temporal semantics and cannot receive identical authority.
Decision: Trend levels own immediate direction. SRSS constrains feasible expression.
Consequence: Joint location becomes interpretable rather than a single blended level.

5. Market location is behavioral as well as geometric

Methodological proposition

Observation: Current meaning depends on local acceptance, unresolved macro pressure, boundary proximity, lifecycle maturity, and volatility state.
Interpretation: Direction alone is underdetermined.
Decision: Coordinate projection is retained as a joint address of the form t_k|m_j. No numbered coordinate is published.
Consequence: Analogues are eligible only under compatible state, not similar price path alone.

6. Signal magnitude is subordinate to lifecycle and location

Internally observed finding

Observation: The same signed magnitude can imply emergence, continuation, fatigue, boundary negotiation, reversal pressure, or inadmissible speculation.
Interpretation: Sequence position changes the meaning of size.
Decision: Lifecycle Regime Interpretation is retained. Continuation, reversal, and fatigue are evaluated separately.
Evidence: Internally observed on BTC native DCM series 41_signals, complete 18h geometry cohort n=13,055 (2025-01-01 to 2026-08-15 UTC): emergence 1,931; validated 3,917; mature 5,279; reset 1,928. Artifact ECMI-A.DCM-SLICE.2026.08. Class-mean expectancy is not published here because supporting cells are location-addressed.

7. Direction requires an authority hierarchy

Operational decision

Observation: Trailing SRSS influence can overpower current local acceptance and overproduce counter-regime theses.
Interpretation: Macro pressure is constraint, not a substitute for present acceptance.
Decision: Trend-first authority: local trend levels form direction; SRSS is restricted to feasibility, friction, target, invalidation, and risk.
Consequence: Fit diagnostics remain subordinate and cannot invent direction.
Evidence: Internally observed negative finding on SRSS-led directional distortion. Trend-level objects in-repo from 2021-02-01 (synq-net 142b6d64); PHALANX trend, SRSS, and coordinate classes from 2026-02-21 (68817f7b). Relative error-reduction table remains in controlled diligence.

ECMI formal method

The ten stages support the discoveries. They are not the narrative. Construction of sentiment state, forecast points, and trend levels remains restricted. Lifecycle, location, authority, and governance are described more openly.

S_t=T₁(D_raw,t) · M_t=T₂(P_raw,t) · R_t=T₃(S_t,M_[t,t+h]) · E_t=T₄(R_t)Discourse structure, market canonicalization, resonance, expectation primitives. Live inference preserves no-lookahead. Sentiment-state construction remains restricted production IP.

φ(e,M_t)∈{0,1} · L_t=T₆(T_t,L_t−1) · C_t=T₇(U_t,C_t−1) · X_t=T₈(M_t,L_t,C_t)Fulfillment partitions history. Local regime and macro topology update from different outcomes. Coordinates project the joint state.

F_t=T₉(X_t,M_t) · D_t=T₁₀(X_t,C_t,F_t,R_t,Q_t), |D_t|≤1Fit is subordinate. Conviction is bounded and may be null. Q is admissibility state.

I. Discourse structuring

Purpose: convert raw discourse into a time-ordered, source-aware, instrument-attributed state. Temporal lineage, instrument attribution, source diversity, and null preservation are public method. The source-specific feature schema, normalization, weighting, attribution, and sentiment-state construction remain restricted production IP. Invariant: lineage and nulls survive. Null: malformed or undated content cannot become S_t. Downstream: resonance has an observational field.

II. Market canonicalization

Purpose: place exchange facts on the same clock. Venue and timestamp alignment, observation-closure rules, missing-data treatment, and future-data exclusion are ordinary quantitative hygiene and are disclosed as method. Invariant: no lookahead. Null: missing prints remain missing. Downstream: realization is comparable to discourse state.

III. Resonance measurement

Purpose: test prior behavioral state against subsequent realization at governed horizon h. Public method includes candidate horizon families, whether the relationship is leading, lagging, contemporaneous, or non-responsive, the evaluation outcome definition, and baseline comparison. Exact resonance formula, feature interaction, weighting, horizon-selection machinery, and thresholds remain restricted. Invariant: horizon is not selected after the outcome. Null: insufficient sample or degenerate variance yields no resonance claim. Downstream: only resonant states may become primitives.

IV. Expectation primitives

Purpose: store E_t = {e₁, …, e_n} as persistent objects with timestamps, validity windows, and fulfilled, unfulfilled, expired, absorbed, or unresolved status. Exact forecast-point generation and reconstructive expectation schema remain restricted. Invariant: objects remain adjudicable. Null: a score is not an expectation. Downstream: fulfillment can be asked as a predicate.

V. Fulfillment and tipping

Purpose: adjudicate φ(e, M_t) under governed fulfillment predicates that may be price-, time-, or state-conditioned. Retrospective fulfillment is prohibited. Exact tolerances remain restricted. Invariant: unfulfilled objects are retained. Null: an event mark is not execution entitlement. Downstream: local and macro updates receive different inputs.

VI. Trend levels

Purpose: encode current local acceptance as an ordinal local regime with boundary and transition semantics. Exact level-generation method, state boundaries, transition thresholds, and persistence logic remain restricted production IP. Invariant: local regime holds primary directional authority. Null: transitional or boundary-ambiguous acceptance cannot force a thesis. Downstream: direction has a present object.

VII. Macro constraints

Purpose: propagate unresolved expectation into SRSS as friction, rejection, latent pressure, or path limitation. Constraints do not independently create direction. Exact construction remains restricted. Invariant: topology constrains; it does not create direction. Null: residual pressure is not a reversal mandate. Downstream: remaining path and invalidation become visible.

VIII. Market coordinates

Purpose: project joint local and macro location as an ordinal address with boundary proximity, transition state, dwell, and maturity. Coordinate dictionary, exact state boundaries, transition rules, and mapping function remain restricted. Invariant: no published numbered coordinate. Null: a coordinate is not a lookup trade. Downstream: analogue eligibility can be gated on compatible state.

Analogue eligibility

Status: Active hypothesis unless a quantified incremental result is later disclosed. Superficially similar paths are ineligible without compatible regime, constraint, lifecycle, horizon, and volatility state.

The distance function may remain protected. Public evaluation names the simpler comparison (direction-only or path-only grouping) and whether joint location changed class behavior. On BTC 41_signals, the 18h geometry cohort (n=13,055, ECMI-A.DCM-SLICE.2026.08) is lifecycle- and location-conditioned; numbered coordinate labels are not published. Status: internally observed.

A_t = {i : d(X_t, X_i) ≤ ε ∧ G_i = G_t}Conceptual relation. Production construction of distance, tolerance, and state gating remains restricted.

IX. Fit metrics

Purpose: diagnose path quality inside location through categories such as path conformity, volatility expectation, persistence, conflict, anomaly, and transition morphology. Exact feature definitions and model prompts remain restricted where reconstructive. Invariant: fit cannot override regime or invent direction. Null: missing diagnostics remain missing. Downstream: conviction may be reduced, not manufactured.

X. Bounded directional conviction

Direction, bounded conviction, trade eligibility, risk budget, and executable units remain distinct. Public method includes bounded range, sign semantics, null behavior, trailing standardization, conflict and staleness as concepts, conviction bands, and the relationship to execution eligibility. Exact aggregation and coefficients remain restricted. A high-conviction state may remain ineligible when extended, boundary-adjacent, illiquid, or uneconomic after implementation cost. Trailing standardization uses only prior observations; degenerate variance yields null.

Z_t = (D_t − μ_t,W) / σ_t,WStrictly trailing window W. Not interchangeable with lifecycle maturity.

Findings and evidence

The findings below report the public evidential status of ECMI's major claims. Haruspex discloses the research question, test design, sample scope, baseline class, aggregate conclusion, limitation, and resulting decision when those fields have been verified for release. Detailed tables, chronological partitions, statistical tests, and selected audit artifacts may be made available to qualified parties under NDA. Production state-generation mechanisms remain restricted.

Classification	Meaning
Methodological proposition	A formal ECMI principle with a defined empirical test.
Internally observed finding	Quantified in Haruspex research but not independently reproduced.
Validated internal result	Survived a defined chronological or held-out internal evaluation.
Negative result	Failed to demonstrate incremental value or degraded the evaluated outcome.
Operational decision	A system or governance change produced by evidence.
Active hypothesis	Testable but not promoted.
Restricted production mechanism	Construction details omitted because they expose the state-generation process.

Findings ledger

Claim	Evidence status	Public aggregate	Baseline	Decision	Diligence availability
Sentiment requires realization	Methodological proposition supported by internal evaluation	Observational field in-repo from 2018-03-05 (`synq` `8dfdfb91`); discourse scrapers from 2019-01-28. Dedicated resonance n is not in current git artifacts.	Raw sentiment polarity or intensity without realization gating	Resonance gate retained	Full horizon analysis and statistical tests
Non-fulfillment retains information	Internally observed	Infinity stones from 2019-05-28; Stones runtime 2020-01-21; SRSS APIs 2024-02-14. Class friction tables in diligence.	Delete or ignore unfulfilled expectations	Unresolved expectations retained in SRSS	Class distributions and implementation-specific analysis
Joint location changes interpretation	Internally observed	BTC 18h cohort n=13,055 (ECMI-A.DCM-SLICE.2026.08). Numbered addresses unpublished. Location-conditioned TP/SL rates differ from the 46.5% / 53.5% resolved baseline.	Direction-only, regime-only, or path-only grouping	Coordinate projection retained	Complete state-conditioned analogue ablation
Lifecycle changes expectancy or risk	Internally observed	Same cohort: emergence 1,931; validated 3,917; mature 5,279; reset 1,928. Forge DCM lifecycle ver 1: 2026-05-08 (`c644076`).	Treat sequential signals as independent	Lifecycle retained	Complete class-level outcomes
Trend-first authority corrected macro distortion	Internally observed negative finding	Trend levels from 2021-02-01; PHALANX trend/SRSS/coordinate classes 2026-02-21. Relative error table in diligence.	SRSS permitted directional weight	Local trend level granted primary directional authority	Detailed error analysis
Outcome timing differed by trade class	Internally observed; adaptive timeout remains a hypothesis	Among 9,799 resolved within 18h, mean exit 6.9 h (not class-stratified in this edition).	Universal timeout	Adaptive timeout retained for testing	Full duration distributions

Negative result: SRSS-led directional distortion

Trailing macro constraint was observed to overpower current local acceptance and overproduce counter-regime theses. The research decision was not to conceal the failure. Trend-first authority restricted SRSS to feasibility, friction, target, invalidation, and risk. Public error-reduction summary remains pending reconciliation; detailed analysis is available under NDA.

Aggregate historical behavior was not assumed uniform across chronological partitions. The public evaluation window for the native DCM BTC series cited here is 2025-01-01 to 2026-08-15 UTC (ECMI-A.DCM-CENSUS.2026.08 / ECMI-A.DCM-SLICE.2026.08). Fold-level P&L after full cost treatment is not published as a performance claim. Subsequent promotion requires a preregistered, execution-linked evaluation. Simulation alone is not sufficient until direction, entry, allocation, exits, overlap, risk rules, and the actionable decision artifact are defined consistently.

Execution-linked candidate study

Public aggregate is the internal BTC native DCM series 41_signals (artifact ECMI-A.DCM-CENSUS.2026.08; census 2026-08-24). Window 2025-01-01 to 2026-08-15 UTC. n=14,047 signals. Evaluable 18h TP/SL geometry n=13,055. Resolved within 18h n=9,799: TP-first 46.5%, SL-first 53.5%, mean exit 6.9 h among resolved (ECMI-A.DCM-SLICE.2026.08). Horizon is a fixed 18h clock, not a production timeout recipe. Fees, slippage, funding, latency, overlap, and capital allocation are not complete in this summary. Gross expectancy per action is not published as a performance claim.

Candidate identity: ECMI-C.native-dcm.20260518. Live trading required a DCM signal source from 2026-05-01 (synq-phalanx 66f7d3a0). Detailed outcome ledgers remain available under NDA. Production execution logic remains restricted.

Public falsifiability matrix

ECMI claims are stated so they can fail. Architectural coherence is not a substitute for the tests below.

Claim	Observable consequence	Baseline	Failure condition
Resonance adds information	Realization-gated behavioral states separate subsequent outcomes better than raw sentiment	Raw polarity or intensity	No stable incremental separation
Non-fulfillment retains information	Unresolved expectations improve friction, rejection, or path-capacity classification	Delete unresolved objects	No incremental information beyond price state
Joint location matters	Coordinate-conditioned groupings improve separation or stability	Direction-only, regime-only, or shape-only	No improvement or unstable chronological behavior
Lifecycle matters	Sequence classes differ in expectancy, excursion, risk, or timing	Treat signals as independent	Materially indistinguishable distributions
Trend-first authority helps	SRSS-led counter-regime errors decline without indiscriminate action suppression	Macro allowed directional authority	Apparent improvement explained only by fewer actions
Adaptive timeout helps	State-conditioned timeout improves net outcome after winner truncation	Universal timeout	No net improvement or excessive missed winners
Abstention adds value	Selective ineligibility improves risk-adjusted outcome	Continuous signal emission	No improvement or unacceptable opportunity loss

Evidence and controlled diligence

ECMI is supported by internal research artifacts spanning behavioral-state evaluation, expectation resolution, lifecycle analysis, coordinate-conditioned comparison, fit-metric ablation, and execution-linked testing. The public paper reports the methodology, evaluation logic, evidence status, selected aggregate findings, negative results, and research decisions necessary to make the claims falsifiable without disclosing the production processes that generate Haruspex state.

Qualified parties may request controlled access under NDA to selected supporting materials, including detailed benchmark results, chronological partitions, statistical analyses, cost sensitivities, ablations, and timestamped evaluation artifacts. Access is determined by diligence scope and does not include unrestricted disclosure of sentiment measurement, forecast-point generation, trend-level construction, production features, parameterization, model prompts, coordinate mappings, or execution logic.

Evidence artifact	Public paper	Controlled diligence	Restricted
Behavioral resonance study	Design, scope, aggregate result, status	Full horizon and significance analysis	Sentiment and resonance construction
Expectation resolution study	Counts, outcome classes, aggregate conclusion	Event-level resolution data	Forecast-point generation and predicates
Lifecycle analysis	Public state definitions and aggregate comparison	Complete stratified results	Exact transition thresholds
Coordinate-conditioning ablation	Baselines and aggregate conclusion	Complete state comparison	Coordinate mapping and distance function
Trend-first comparison	Failure mode, relative result, decision	Full version comparison	Production trend-level generation
Native DCM execution study	Verified sample and bounded aggregate summary	Complete outcome ledger and folds	Exact model and production execution logic
Cost sensitivity	Cost classes and public range	Exact assumptions and tables	Venue-specific production policy

What the evidence changed

1. Sentiment cannot authorize action directly

Raw sentiment remains upstream of resonance, expectation, and realization.

2. Unfulfilled expectations remain in state

Unresolved pressure becomes macro topology rather than discarded prediction error.

3. Local acceptance owns immediate direction

SRSS cannot manufacture direction from trailing mean-reversion pressure.

4. Lifecycle modifies the meaning of magnitude

Emergence, validated persistence, and maturity are evaluated separately.

5. Conviction does not equal position size

Risk allocation remains a separate downstream layer.

6. Simulation cannot complete an undefined trade policy

Direction, entry, allocation, exits, overlap, risk rules, and actionable decision state must be specified before simulation can establish execution viability.

7. Freeze before further tuning

Following the reviewed execution-linked DCM path, the native candidate was frozen against further discretionary tuning under identifier ECMI-C.native-dcm.20260518. Subsequent promotion requires an execution-linked objective, a preregistered evaluation protocol, chronological out-of-sample testing, and an untouched holdout. The identifier establishes which candidate was evaluated and prevents retrospective version substitution. Git pin for DCM-required live routing: synq-phalanx 66f7d3a0 (2026-05-01). A hash is not the model.

Lifecycle and Trend Mapping

Lifecycle is not an interpretive metaphor. It partitions signal emissions by sequence maturity so that continuation, reversal, fatigue, and timeout behavior can be evaluated separately. Magnitude without that position is underdetermined. Counts below are from the BTC 41_signals complete 18h geometry cohort (n=13,055; ECMI-A.DCM-SLICE.2026.08; generated 2026-08-24). Exact state-transition thresholds remain restricted. Reset (n=1,928) is retained in state and is not collapsed into the three public classes.

Lifecycle class	Public definition	Sample	Cohort note	Status
Emergence	Early directional formation	1,931	Class-mean P&L not published without location cells	Internally observed
Validated persistence	Repeated accepted direction	3,917	Class-mean P&L not published without location cells	Internally observed
Mature persistence	Late sequence with possible fatigue	5,279	Class-mean P&L not published without location cells	Internally observed

Trend Mapping is the named correction after SRSS-led counter-regime error: local trend levels retain directional authority; model-assisted fit remains subordinate.

Dir_local ≻ Fit_model ≻ Constraint_macroRepresentative trend-first order. Production construction of the operators remains restricted.

Adaptive timeout is an active hypothesis unless timing evidence is later disclosed. A thesis that fails to behave as expected within its state-appropriate window should lose authority even if its original direction remains intact.

Conviction, risk, and execution

Evidence → expectation primitive Resolution fulfilled → local regime · unfulfilled → macro constraint Joint location → lifecycle and fit review → bounded conviction Risk allocation is a later layer, not a rescaling of D_t Ruling Eligible · Conditional · Probe · Manage only · Abstain

These rulings are governance outputs, not performance claims. Abstention is affirmative when the chain is incomplete, conflicted, stale, obstructed, or uneconomic after implementation cost.

Primary direction: local trend levels. Cannot be replaced by a model score or by unresolved macro pressure alone.
Constraint: SRSS. Cannot create direction.
Diagnosis: fit metrics. Cannot override regime.
Synthesis: bounded conviction. Cannot collapse into size or order entitlement.
Capital: risk allocation. Cannot repair a poor location.
Action: execution ruling. Cannot manufacture confirmation.

Entry eligibility classes, risk-budget separation, invalidation principles, cost and liquidity checks, timeout governance, abstention logic, and portfolio-exposure checks are public method. Exact sizing coefficients, exact entry or exit triggers, venue-specific order placement, and production caps remain restricted.

RiskBudget_t = B · κ(|D_t|) · q_loc · q_life · q_fit · q_liq · q_volRepresentative governing form, not a production function. Exact coefficients are production-specific; the public relationship is that conviction can influence risk only after location, lifecycle, fit, liquidity, volatility, and portfolio constraints remain admissible.

Size_t = min(Size_cap, RiskBudget_t / LossDistance_t)Representative risk-normalized sizing. Portfolio construction is not claimed by this edition.

Execute_t = 𝟙[Q_t=valid] · 𝟙[Edge_net>0] · 𝟙[Risk_t≤Budget] · 𝟙[State_t∈Eligible]Conceptual admissibility gate.

Research provenance

Duration of the program is not proof of advantage. Dates below are git first-commits or first in-repo appearances of the named object, plus the 2026-08-24 BTC census and slice artifacts. The eight-year span is 2018-03-05 to this edition.

Date or period	Research problem	Artifact or milestone	Finding or failure	Decision produced
2018-03-05	Sentiment treated as instruction	SynQ application (`synq` `8dfdfb91`)	Conversion loop required temporal discipline	Thesis retained as governing frame
2019-01-28 / 2019-01-30	Incomparable clocks	Discourse scrapers and SynQ net API	Observation without a shared clock is incomparable	Structured observational field retained
2019-05-28	Discarded forecasts	Infinity stones / timestones in `synq-net`	Transient scores delete timing and unresolved pressure	Persistent adjudicable expectations retained
2020-01-21	Failed predictions deleted	Stones runtime (`synq_stones` `4e68315`)	Refusal itself contains constraint	Dual information channels retained
2021-02-01	Direction without location	Trend levels on symbol pair (`synq-net` `142b6d64`)	Joint address required for comparability	Trend levels retained; SRSS and coordinates follow in later artifacts
2023-03-01 (as_of)	Unbounded directional scores	Archived DCM `v1_baseline` series start (n=28,045 through 2026-07-18)	Score magnitude is not entitlement	Bounded conviction distinct from action
2026-02-14 / 2026-02-21	Magnitude without sequence; local vs macro authority	PHALANX repo; trend, SRSS, coordinate classes (`68817f7b`)	Sequence position changes meaning of size; macro can overpower local acceptance	Lifecycle and trend-first authority retained
2026-03-23 / 2026-05-08	Simulation of an undefined trade	Forge sim bed; DCM lifecycle ver 1	Simulation cannot complete an unspecified policy	Ruling classes and freeze discipline retained
2026-05-01 / 2026-05-18	Discretionary retuning	DCM-required live routing; manuscript ECMI-R.2026.05.1	Candidate identity must not be substituted after review	ECMI-C.native-dcm.20260518
2026-08-24	Public cohort pin	BTC census and 18h slice artifacts	Sample and window can be stated without the production recipe	ECMI-A.DCM-CENSUS.2026.08 and ECMI-A.DCM-SLICE.2026.08

Validation and promotion

No version advances because it is elegant. Promotion requires survival of backfill, out-of-sample review, symmetry where applicable, null handling, and operational sanity checks. Guardrails include trailing-only statistics, fixed horizons, null preservation, regime segmentation, and signal/trade separation.

Version hashes, holdout boundaries, and preregistered evaluation records do not disclose production mechanics and should be public wherever they exist. This edition pins candidate ECMI-C.native-dcm.20260518, evaluation window 2025-01-01 to 2026-08-15 UTC, and a fixed 18h geometry horizon for the cited BTC cohort. Detailed event data and complete fold-level results may remain in controlled diligence. Costs in the public summary are explicitly incomplete.

Gate	Public requirement
Frozen candidate	ECMI-C.native-dcm.20260518 (git pin `66f7d3a0`, 2026-05-01)
Defined objective	Directional, trade-level, or portfolio-level target
Chronological evaluation	2025-01-01 to 2026-08-15 UTC for the cited BTC series
Untouched holdout	Declared before evaluation; 18h clock used for the cited geometry cohort
Baselines	Simple and native alternatives defined
Costs	Included or explicitly incomplete
Null handling	Missing and inadmissible states preserved
Negative result	Disclosed rather than absorbed through retuning
Promotion decision	Advance, retain, reject, or continue study

Limitations and open questions

ECMI does not claim causal proof, stationarity, or guaranteed profitability. Material limitations include non-stationarity, regime discontinuity, source drift, adversarial narrative manipulation, liquidity shocks, venue anomalies, exogenous events, model and prompt drift, multiple-testing risk, selection bias, cost sensitivity, execution latency, portfolio overlap, limited evidence in rare state classes, and the absence of independent public replication.

The public evidence remains internally produced and has not yet been independently replicated. Controlled review under NDA can increase confidence in procedure and reported results, but it is not equivalent to public reproducibility or independent academic validation. The paper therefore distinguishes internal validation from external confirmation.

Active questions: whether coordinate conditioning adds stable incremental value beyond simpler regime classification; whether non-fulfillment topology improves target, invalidation, or path-capacity estimation; whether Lifecycle improves expectancy or primarily reduces risk; which fit metrics add information after local regime and coordinate state; whether adaptive timeout improves net results after winner truncation; whether abstention improves risk-adjusted performance without unacceptable opportunity loss.

Claimed contribution and conclusion

ECMI does not claim that discourse mechanically determines price or that every behavioral state is actionable. It establishes a governed method for preserving expectations, adjudicating how markets fulfill or refuse them, constructing state from both outcomes, and conditioning conviction upon the location and maturity in which evidence arises. Internal research has shown that these distinctions are measurable and that they alter downstream interpretation. The surviving framework therefore represents not a single prediction model, but an accumulated research discipline for deciding when behavioral evidence may become risk and when it must remain observation.

What was discovered is that fulfillment and non-fulfillment both construct state. Internal research quantified differences associated with expectation resolution, lifecycle maturity, authority assignment, and outcome timing. This edition reports selected verified aggregates and the decisions they produced; detailed supporting analyses are available through controlled diligence where appropriate. Public numeric aggregates that have not been reconciled remain labeled pending reconciliation rather than estimated. What changed is authority, retention of unresolved expectation, and separation of conviction from size. What survived is the constraint-preserving chain. What remains unproven includes analogue incremental value, timeout net benefit, and independent replication. What Haruspex claims is the integrated contribution stated at the opening of this paper.

ECMI's contribution is not that discourse invariably predicts price. It is the construction of a persistent expectation lifecycle through which behavioral evidence can be tested, accepted, refused, retained, located, and governed. The methodology's current form reflects empirical distinctions that survived internal review and failure modes that required correction, including SRSS-led directional distortion. Its purpose is to determine when behavioral information has earned the authority to influence risk and when the correct conclusion remains observation, management of existing exposure, or abstention.

Compact operating definitions

Term	Meaning in this paper
ECMI	Expectation-Constrained Market Inference: the unified chain from sentiment landscape through bounded conviction and execution eligibility.
SynQ	Observation and structuring layer for discourse and market intake.
JARVIS	Interpretation and state construction: lifecycle, trend levels, SRSS, coordinates.
PHALANX	Governance: fit, bounded conviction, sizing eligibility, execution rulings.
Trend levels	Current micro or local regime with primary directional authority.
SRSS	Trailing macro sentiment support and resistance topology from unresolved expectation.
Trend Mapping	Trend-first refinement; fit remains subordinate.
Lifecycle Regime Interpretation	Sequence-aware reading of emergence, continuation, conflict, fatigue, and reversal pressure.
Bounded conviction	Signed, limited synthesis of admissible evidence; distinct from profit probability and from trade entitlement.
Abstain	Affirmative governance output when the chain does not authorize new exposure.

References

Cited literature supplies conceptual lineage. It does not validate ECMI. Implementation-adjacent data sources are observational inputs, not reconstructive references.

Antweiler, W., and Frank, M. Z. (2004). Is all that talk just noise? The information content of internet stock message boards. Journal of Finance, 59(3), 1259–1290.
Tetlock, P. C. (2007). Giving content to investor sentiment: The role of media in the stock market. Journal of Finance, 62(3), 1139–1168.
Hamilton, J. D. (1989). A new approach to the economic analysis of nonstationary time series and the business cycle. Econometrica, 57(2), 357–384.
Lo, A. W. (2004). The Adaptive Markets Hypothesis. Journal of Portfolio Management, 30(5), 15–29.
Shiller, R. J. (2003). From efficient markets theory to behavioral finance. Journal of Economic Perspectives, 17(1), 83–104.
Almgren, R., and Chriss, N. (2000). Optimal execution of portfolio transactions. Journal of Risk, 3(2), 5–39.
Haruspex Trading Co., Inc. Consolidated ECMI methodological paper, 18 May 2026, and subsequent internal SynQ, JARVIS, and PHALANX documentation. Operational parameterization omitted.

Internal evidence corpus

Sanitized identifiers. File locations and production parameterization are omitted. Numbered coordinate dictionaries are not listed.

Artifact	Date	Version	Evidence status	Availability
Behavioral / discourse observational field	2018-03-05 / 2019-01-28	ECMI-A.SYNQ.2018.03	Provenance; resonance n not in git artifacts	Available in controlled diligence
Expectation-resolution lineage (infinity stones)	2019-05-28	ECMI-A.NET.2019.05	Internally observed	Available in controlled diligence
Lifecycle regime study (BTC 18h cohort)	2026-08-24	ECMI-A.DCM-SLICE.2026.08	Internally observed	Public class counts; cell P&L in diligence
Trend-first comparison	2021-02-01 / 2026-02-21	ECMI-A.PHX.2026.02	Internally observed negative finding	Available in controlled diligence
Native DCM census and geometry study	2026-08-24	ECMI-A.DCM-CENSUS.2026.08	Internally observed; costs incomplete	Public window and n; ledger in diligence
Outcome-timing (resolved-within-18h mean)	2026-08-24	ECMI-A.DCM-SLICE.2026.08	Internally observed; adaptive timeout still a hypothesis	Mean 6.9 h public; class distributions in diligence
Current validation protocol	2026-05-18	ECMI-R.2026.05.1 / ECMI-C.native-dcm.20260518	Operational standard	This paper plus controlled diligence