Skip to main content

Performance Database

HyperionX saves completed Validator, optimizer, live/manual, and standalone performance artifacts so that you can find and review prior results.

The current database is a result browser and archive. It is not a complete experiment-reproducibility system.

Open And Filter Results​

Open the performance/results database from the main application, then filter the available records by fields such as:

  • Result type or status.
  • Symbol.
  • Strategy name and version.
  • Data series or timeframe.
  • Saved date.

Depending on the result type, you can open, rename, remove, refresh, or open its storage folder. Not every saved result exposes the same views.

Save A Result​

Validator results are saved when their runs complete. Optimizer results can be saved with a test name, and completed optimization workflows can also create saved result data. Manual/live result workflows can save standalone trade artifacts.

Use a descriptive name that includes the research purpose rather than only “test 1.” Keep strategy source and presets under separate version control.

Result artifacts are stored below:

  • Documents\HyperionX\PerformanceResults
  • Documents\HyperionX\OptimizerResults

Do not edit active result files while HyperionX is writing them.

What Is Preserved​

The stored database record can include:

  • Result name, type, and status.
  • Strategy identifiers, name, and strategy version where available.
  • Symbol, account, and connection identifiers where applicable.
  • Data-series and bar-type information.
  • Historical start and end dates.
  • Headline performance metrics.
  • A path to the saved artifacts.
  • Save/update timestamps.

Standalone performance storage can include:

  • A result summary.
  • Trade data in JSON.
  • Trade data in CSV.
  • Audit summary and run-manifest artifacts.

The trade export can include commission, leverage, margin, liquidation-related fields, fill-model identifiers, and ambiguous-fill information when those values exist in the source result.

What Is Not A Reproducibility Guarantee​

Persistence depends on both the originating workflow and its version. The signed 1.1.25 release record adds Validator reproducibility metadata; this does not change older saved files or establish identical fields for every Optimizer result. See the documentation version map before assuming the installed package contains those changes.

Result artifactWhat to expect
Validator runManifest.json from the 1.1.25 changesRequested and loaded ranges, range mode, timezone, historical fill model, bar timestamp mode, commission model and serialized commission parameters, slippage ticks, session-close settings, continuous-futures merge mode, and comparison warnings.
Validator summary export from those changesRange/timezone, commission, slippage, fill-model, and continuous-merge settings alongside results. Inspect the actual columns for that build.
Validator strategy preset from those changesCommission, slippage, and session-close execution settings accompany the strategy preset. Keep the preset with the result.
Optional Validator decision diagnosticsdiagnostics.csv when the strategy supplies diagnostic records; absence does not mean the strategy had no decisions.
Older results and other result typesMay lack the newer metadata; inspect the actual manifest, preset, and export.

The manifest is still not a complete reproducibility guarantee. Preserve these separately when the artifacts do not contain them:

  • Strategy source or source-control revision, full parameter values, and optimizer ranges.
  • Selected optimizer and fitness configuration.
  • Custom commission state that is not represented by serialized parameters.
  • Starting account value and any remaining quantity, leverage, margin, or session assumptions.
  • Application/build version.
  • Complete order history, chart state, intrabar equity path, or risk-level views.

Validator slippage ticks apply to historical Market and StopMarket fills, not historical Limit fills. This setting is not an ordinary Optimizer control and does not model queue position or liquidity.

When a standalone result is reopened from stored trades, Chart, Orders, and RiskLevels views are not reconstructed. Older records and different result types can contain less information than newer ones.

Keep A Reproducible Research Record​

For an important run, save these items together:

  1. Strategy source or a source-control commit identifier.
  2. Strategy preset and exact parameter values.
  3. Instrument, provider, bar type, interval, and date ranges.
  4. Historical fill model.
  5. Commission module and its configuration.
  6. Starting capital, position-sizing, leverage, and margin assumptions.
  7. Optimizer ranges, optimizer type, fitness metric, and Keep best value.
  8. HyperionX version from Help > About.
  9. Saved result identifier and exported trade data.
  10. Notes identifying in-sample, out-of-sample, and later evaluation periods.

Store that research record outside the application database as well. A result that can be reopened is not necessarily a run that can be reproduced exactly.

Review A Result​

  1. Open the saved record.
  2. Confirm the strategy, symbol, timeframe, and date range.
  3. Review headline metrics and trade count. Keep closed-trade max drawdown separate from intrabar-equity max drawdown where the result provides both; the latter requires the equity path and cannot be inferred from closed trades alone.
  4. Inspect long/short and per-instrument contribution where available.
  5. Inspect the trade list and commission values.
  6. Check the selected fill-model information when recorded.
  7. Compare the result with the external research record.
  8. Re-run the original preset when exact reproduction is required.

If a stored result is missing Chart, Orders, or RiskLevels, that is a current persistence limitation rather than evidence that those views existed in the saved artifact.

Export​

Available export actions depend on the originating workflow. Current examples include:

  • Validator summary CSV.
  • Optimizer PDF.
  • Saved trade CSV and JSON.
  • Chart-dataset CSV where supported.
  • Research Lab CSV and JSON.

Exports can contain account labels, strategy details, instrument data, local paths, or proprietary research. Inspect each file before sharing it. Research Lab exports are plaintext even when the saved Research Lab dataset is encrypted.

Remove Or Archive Results​

Before removing a database entry:

  1. Open its storage folder.
  2. Confirm whether another record refers to the same files.
  3. Copy required exports and research notes to the approved archive.
  4. Remove only the intended result.

Deleting files directly can leave an incomplete database record. Removing an application record can be irreversible.