Studies
Portfolio Study Design
Configure and queue a coordinated Study for an exact saved Portfolio version.
App path
- Studies -> Study Design -> Portfolio
Design a Portfolio Study
Portfolio Study Design tests how an exact saved Portfolio version behaves when its Strategy Configurations share data, capital, admission rules, risk, and exposure limits.
The saved Portfolio defines its members and standing policies. Study Design adds the test window, market bindings, execution assumptions, and any supported values you want to compare.
Select the Portfolio
Open Studies -> Study Design and choose Portfolio under Study Level. Select the saved Portfolio and confirm its displayed version. Create Study from Portfolio Library carries the selected Portfolio and version into this workspace.
Enter a clear Study Name. Presets can save and restore a reusable design setup, but they do not contain completed Studies or Results.
Choose the Study mode
| Field | Meaning |
|---|---|
| Study Type | Runs one coordinated setup or compares supported Portfolio candidates. |
| Optimization Method | Chooses how candidate policy combinations are generated. |
| Execution Fidelity | Chooses the supported coordinated simulation path. |
| Analysis Depth | Chooses the validation and retained output depth. |
| Chart Output | Retains additional data for interactive charts when enabled. |
| Objective and Direction | Define how optimization candidates are ordered. |
| Retain Top | Sets how many leading candidates keep detailed Run Outputs. |
| Candidate Limit | Caps the candidate set that can be generated. |
The current estimate is the authoritative candidate count. Do not infer it from one axis when participation, priority, weights, policies, or Controls also vary.
See Study Types, Optimization, and Walk-Forward for the search methods, Optuna sampler and pruner choices, walk-forward window types, Execution Fidelity, and Analysis Depth. Readiness remains authoritative for the combinations supported by the current Portfolio path.
Bind the market data
Market Inputs lists every participating Strategy Configuration and its required data slots. Select compatible Datasets and assets, then choose the common Study window.
Configurations with several inputs expose those bindings under Strategy Inputs. Signal Sources shows required saved Signal outputs. Every active Configuration needs compatible coverage for the requested coordinated window.
Use the visible Participation control only when exclusion is an intentional candidate choice. Otherwise, missing data for one member blocks the coordinated Study rather than silently removing that member.
Set execution assumptions
Use Execution Assumptions to review the execution model, starting capital, and supported walk-forward settings. These values apply to the coordinated Portfolio test and affect every candidate's meaning.
When walk-forward is enabled, define the supported method, unit, training, validation, step, purge, and embargo values. All participating inputs must support the requested folds.
Configure Portfolio Controls
Portfolio Controls defines supported conditions that can affect admission or risk during the Study. A Control can bind a Signal output, scope it to the Portfolio or one Configuration, and apply a supported predicate, threshold, hysteresis, dwell rule, or stale-data rule.
Controls are explicit Study inputs. If a required Signal, Dataset, asset, or output column is unavailable, resolve the dependency instead of guessing a replacement.
Choose Portfolio Study axes
Study Axes defines the Portfolio values that may vary. Supported axes can include:
- Configuration participation;
- priority orders;
- fixed-weight vectors;
- Configuration risk multipliers;
- Portfolio policy values;
- Control values;
- risk limits and allocation settings supported by the current Portfolio path;
- retained-finalist counterfactuals; and
- retained-candidate stress scenarios.
Hard Portfolio limits remain authoritative. A candidate multiplier cannot override them. Explicit priority orders and weight vectors are the candidates you supplied; Arizmic does not imply that every possible permutation was tested.
Keep the set small enough to understand. A large set of policy combinations can make it difficult to tell which change caused a result.
Validate, estimate, and queue
Review Readiness after any change to the Portfolio, data, Controls, axes, execution settings, or output depth.
- Resolve the local findings.
- Select Estimate Study.
- Review the candidate count, preparation plan, fidelity, and resource estimate.
- Accept the current confirmation.
- Select Queue Study.
Changing the design invalidates the earlier estimate. Create a new estimate before queueing.
The top status area shows active work. System & Jobs -> Jobs owns detailed history, cancellation, retry, and recovery.
Troubleshooting
- No Portfolio is available: save a Portfolio containing at least one saved Strategy Configuration.
- A Configuration has no compatible data: correct its Dataset and asset binding or intentionally exclude it where participation is a supported axis.
- A Control is unresolved: repair its Signal, Dataset, asset, or output reference.
- The candidate count is too large: reduce policy axes, weight vectors, priority orders, participation choices, counterfactuals, or stress scenarios.
- Queue Study is disabled: resolve Readiness findings and create a current estimate.
- The Study fails: inspect the first failed phase through the top status or Jobs, correct the owning input, and queue a new Study.
Next
Open Portfolio Study Results when the coordinated Study finishes.