ARIMA_SAX_Hybrid_Forecaster/RESEARCH_CHARTER.md

# RESEARCH CHARTER

| field | value |
|---|---|
| Document | RESEARCH_CHARTER.md |
| Version | 1.0 (2026-08-28) |
| Status | **CONTROLLED** |
| Project | ARIMA_SAX_Hybrid_Forecaster |
| Authoritative workspace | `D:\TradingTerminal\MetaTrader 5\MQL5\Shared Projects\ARIMA_SAX_Hybrid_Forecaster` |
| Remote | `https://forge.mql5.io/chiki2bum2/ARIMA_SAX_Hybrid_Forecaster.git` |

This charter is the **permanent statement of purpose** for this research
repository. It is the highest-level governance document. It does not change
with experiments, results, or implementation details. It may change only
through a documented, versioned charter amendment as defined in §6.

---

## 1. Permanent project purpose

> Determine **empirically** whether information available at time `t` contains
> **out-of-sample predictive information** about subsequent asset-price
> movement, and whether such information has **economic value after realistic
> transaction costs**.

The purpose is deliberately framed as a question that can be answered
positively, negatively, or not at all. It is not framed as a mandate to
produce a profitable system.

---

## 2. Governing principles

The following principles govern every research family, protocol, experiment,
and result in this repository:

1. **This is a research project, not a profitability-optimization project.**
   The objective is to determine whether information at time `t` predicts
   subsequent price movement and whether any such information survives
   realistic transaction costs.

2. **Positive, negative, null, and inconclusive results are all valid.**
   A negative or null result is a legitimate scientific outcome and must never
   be treated as a failure of the research or as something to be corrected.

3. **Models are instruments, not the research objective.**
   Models, statistical methods, and pipelines are tools used to answer the
   research question. They are selected and constrained to preserve the
   scientific design — never as the subject being optimized.

4. **Conclusions must be out-of-sample.**
   Any claim about predictive information or economic value must rest on
   evidence that was genuinely out-of-sample with respect to model fitting,
   feature construction, and any selection or tuning carried out prior to
   evaluation.

5. **Lookahead is prohibited.**
   No feature, label, or origin may use information that would not have been
   available at time `t`. No-leakage is a hard, non-negotiable invariant, and
   any violation invalidates the affected result.

6. **Economic claims require explicit cost analysis.**
   No economic value may be asserted without an explicit, pre-registered
   treatment of realistic transaction costs applied to gross results.

7. **Scientific scope changes require a new protocol.**
   The research question, hypothesis, target, features, horizons, models,
   costs, statistical or economic design may be changed only as a deliberate,
   documented, scientific-scope change carried by a new or amended protocol —
   never implicitly by results.

8. **Previous negative results must not be "rescued" through model shopping.**
   Closed negative results are immutable and respected. They may not be
   resurrected, reopened, or reinterpreted by selecting models, parameters,
   features, horizons, or costs that happen to produce a preferred outcome.

---

## 3. Classifications of repository content

Every writing artifact in this repository is allocated to one of three classes.

### IMMUTABLE
Frozen and byte-preserved. Never rewritten, regenerated, reopened, or amended,
except by an explicit, separate, owner-authorized scientific step that also
produces new, non-overwriting evidence.

- R1-R evidence (`results/R1_R/`)
- R2-H evidence (`results/R2-H/`)
- Frozen MALE5-ML protocols (`results/MALE5-ML/MALE5_ML_PROTOCOL_v0.1.1_AMENDED.md`,
  and the preserved historical `..._v0.1.0` identity)
- Frozen MALE5-ML manifests (e.g. `MALE5_ML_MANIFEST_v0.1.1_AMENDED.json`)
- Reference MALE5 v3.2.0 (commit `d6fb0695a7c2fbeaa03de379557a8b5447915384`)
  and the dependency-finalization evidence.

### MUTABLE
- `RESEARCH_STATE.md` — the compact machine-readable checkpoint. This may be
  updated to reflect completed gates and verified their results. Updating the
  state may not redefine the research question or amend a protocol.

### CONTROLLED
- `RESEARCH_CHARTER.md` — this document.
- `RESEARCH_RULES.md` — the companion rules document.

## 4. Status of the current protocol
Controlled documents may be changed only by a documented, versioned change as
described in §6.

---

## 5. Relationship to protocols and evidence

- The charter defines **why** the research exists.
- Protocol documents define **how** a specific research family is conducted and
  are themselves frozen at their `-frozen` design state.
- Evidence documents record **what was observed** and are immutable.
- The state document records **where the project currently is**.

These documents are the **source of truth for the research**. Conversation
history, prior sessions, and unstructured notes are never authoritative over
them.

---

## 6. Charter change governance

Any change to this charter (or to `RESEARCH_RULES.md`) must:

1. Be proposed, documented, and explicitly authorized before it is applied;
2. Produce a new version of this document with an incrementing version number
   and a dated change entry in the `Changes` table below; and
3. Never quietly rewrite an earlier version — prior wording, where relevant,
   is preserved in the change log.

### Change table

| Version | Date | Scope | Authorized by |
|---------|------|-------|---------------|
| 1.0 | 2026-08-28 | Initial permanent charter establishing the research purpose and governing principles | RESEARCH GOVERNANCE SETUP gate |

---

_End of RESEARCH_CHARTER.md — version 1.0._