Strategy Study Results

Review a completed Strategy Study

Open Studies -> Study Results, keep Study Level set to Strategy, and select the completed Study Result. Use the six result tabs in order when you need a complete review: Overview, Leaderboard, Parameter Surface, Robustness, Diagnostics, and Artifacts.

The validation strip remains visible across the result. Read its requested and effective data fidelity, execution fidelity, Analysis Depth, execution model, out-of-sample status, reproducibility, and limitations before interpreting performance. A completed Study is not automatically execution evidence.

Overview

Overview confirms what ran. It identifies the Study, saved Strategy version, Dataset bindings, method, status, and attempt totals.

Completed, failed, canceled, pruned, retried, skipped, resumed, deduplicated, and burned trials are different states. Do not use only the completed count to describe how much searching occurred.

For multi-input or multi-asset Studies, review the input and asset bindings. The lineage panel records the saved Strategy, Study definition, execution settings, and retained Result identity.

Before continuing, confirm:

1. the intended Strategy and saved version;
2. the expected Dataset, assets, and date window;
3. requested versus effective fidelity and Analysis Depth;
4. whether the output is discovery or execution evidence;
5. the full attempt accounting; and
6. any visible limitation or missing retained output.

If the identity or assumptions are wrong, return to Strategy Study Design and queue a new Study. The existing Result remains attached to what it actually used.

Leaderboard

Leaderboard compares retained candidates and shows the underlying trial records. For a multi-asset Study, use Asset Filter to view one asset or all assets.

Choose the available rank basis before reading the table:

• Peak emphasizes the strongest observed objective value.
• Plateau favors a region of nearby parameter values with consistently strong results.

These views can select different candidates. Read the objective together with the parameter values, asset or selection identity, trade count, and validation state. A high score with few trades, a failed trial, or one isolated parameter combination needs more checking.

Where available, save a retained candidate as a new Strategy Configuration or send it through the supported validation action. Saving copies its parameter values into a new Configuration; it does not change the original Strategy or prove performance in another period.

Validation may require retained trial trade paths. If the action is disabled, use the Analysis Depth named by the interface in a new Study.

Parameter Surface

Parameter Surface shows how the selected metric changes across the tested parameter space.

• Heatmap shows observed values for the displayed parameter pair.
• Projections show supported higher-dimensional views while holding other parameters at displayed anchors.
• Cross-asset stability compares whether a useful region appears across compatible selected assets.
• Search response summarizes retained convergence, importance, or related method-specific information when available.

An empty cell means no compatible observation is displayed for that combination; it is not a zero result. A smooth chart can also hide sparse coverage, so check the distribution of observed trials.

Prefer a candidate whose neighboring tested values remain acceptable over one supported by a single sharp cell, all else equal. Then check whether the same region survives the relevant assets, time splits, and execution assumptions.

Robustness

Robustness gathers the validation reports produced by the Study method and selected Analysis Depth. These can include optimizer details, multi-objective results, attempt statistics, bootstrap or overfitting checks, cross-asset comparisons, walk-forward folds, and Optuna pruning information.

Only reports actually produced by the Study appear. A missing report is not a passing result.

• A confidence interval applies under that report's resampling assumptions.
• A multiple-testing adjustment covers the declared attempts, not experiments performed outside the Study.
• A walk-forward panel describes the folds that were actually formed and run.
• Pruning information explains which Optuna trials stopped early; pruned trials are not completed comparisons.
• A Pareto frontier shows trade-offs among declared objectives and may not have one universally best point.

If a required check was not retained, return to Strategy Study Design and queue a new Study with the necessary Analysis Depth, data, and method.

Diagnostics

Diagnostics shows phase-level telemetry for the selected Study Result. Use the phase name, status, message, and elapsed time together to understand where processing slowed or failed.

To recover:

1. Find the first failed or blocked phase rather than the last repeated message.
2. Match it to the owning data, dependency, Study setting, execution path, or retained output.
3. Open System & Jobs -> Jobs for job history, cancellation, or retry.
4. Correct the owning input and queue a new Study when the definition changes.

Diagnostics explain processing; they do not replace lineage or validation.

Artifacts

Artifacts lists the files and reports retained for the Study Result. The list can include the result summary, top candidates, trial records, fold reports, pruning reports, and other supported validation outputs.

Read each artifact's role and availability state. A listed path does not mean every optional output was requested or successfully produced. If a Leaderboard or Robustness action needs an absent artifact, rerun the Study with the necessary retention setting.

Keep exported artifacts with the Result's Strategy version, data bindings, execution settings, and lineage. Do not substitute a similarly named file from another Study.

Next

Open Results when you want to inspect an individual retained Run Output in detail.