Research Note 2 – Appendix

Appendix A — Optuna Dashboard Artefacts: EURUSD 1H Candidate T0012

The following screenshots are cropped Optuna Dashboard artefacts from the EURUSD 1H generic scout study. They document the discovery-stage provenance for candidate T0012 and illustrate how the Workbench links Optuna trial output to downstream SQL, clustering, walk-forward, and execution-engine validation.

These figures are included as process artefacts only. They do not constitute validation of deployable alpha.

Figure A1 — Optuna Study Registration

Optuna study registration for EURUSD 1H generic scout run
Figure A1. Optuna study registration for the EURUSD 1H generic scout run. The study is configured as a maximisation problem and records the discovery process used to surface candidate configurations for later validation.

Figure A2 — Trial-Level Provenance for EURUSD 1H T0012

Optuna trial detail for EURUSD 1H T0012
Figure A2. Trial-level provenance for EURUSD 1H T0012. The dashboard links Optuna trial 12 to the sampled parameter set and persisted user attributes, including run identifier, objective score, profit factor, session window, and test-period metadata.

The visible sampled parameter set includes:

θ12 = {ema_period=28, atr_period=10, rsi_period=10, adx_period=16, max_adx=30.0, entry_stretch_atr=1.3, …}

The associated discovery-stage result vector is summarised as:

R12 = {N=1195, P=860.66, p̄=0.7202, PF=1.1272, WY=1.0804, B=0, M=37}

Figure A3 — Objective History and Preliminary Hyperparameter Importance

Optuna objective history and hyperparameter importance for EURUSD 1H study
Figure A3. Objective-history and preliminary hyperparameter-importance view for the EURUSD 1H discovery run. Penalised trials are visible in the lower objective region, while candidate-quality trials occupy the upper region. The importance ranking suggests that max_adx was the dominant contributor to objective variation in this specific study.

The hyperparameter-importance view is interpreted as an exploratory diagnostic. It estimates which sampled parameters helped explain variation in the Optuna objective score. It does not prove causality or future robustness.

Figure A4 — Parameter Relationship Diagnostics

Optuna parameter relationship and contour diagnostics for EURUSD 1H study
Figure A4. Parameter relationship diagnostics for the EURUSD 1H discovery run. The slice and contour views illustrate how objective values varied across sampled regions of the parameter space, including interaction views such as adx_period versus atr_period.

Apparent structure in these visualisations is not treated as sufficient evidence of robustness. Candidate validity remains dependent on downstream SQL diagnostics, HDBSCAN feature-space clustering, walk-forward validation, embargo controls, veto eligibility review, coded retesting, and demo-live monitoring.