RTX5 DesktopBacktesting
Review the proposed desktop backtesting workflow for RX5. Tested builds, licensed fixtures, supported execution models and export formats are not yet published.
Backtesting is currently proposed as a desktop-based workflow.
Proposed Backtesting Workflow
Prepare Data
Prepare licensed historical data with source, coverage, resolution and checksum. Confirm the actual import format and validate missing or duplicate records.
Configure Test
Declare spread, fees, slippage, partial-fill assumptions and any seed. Which models and seed controls are implemented must be confirmed for the desktop build.
Run Backtest
Once a compatible SDK/desktop build is available, run the reviewed strategy against the licensed dataset and retain its configuration and diagnostics.
Validate Results
Review equity curve, drawdown, trade log, fill prices and execution statistics. Compare with walk-forward periods.
Export
Retain configuration, data identity, trade log and results together. Actual export formats and a tested example bundle remain unpublished.
Prepare a Desktop Test You Can Reproduce
Keep the strategy, data and execution assumptions together. A result is only comparable when another run uses the same inputs.
Swipe horizontally to view all columns.
| Input | What to record | RTX5 publication status |
|---|---|---|
| Desktop and strategy | Operating system, desktop build, SDK/compiler version, source revision and strategy parameters. | Tested desktop build and SDK version: not yet published. |
| Historical data | Provider and usage rights; instrument, tick or bar resolution, date range, time zone and dataset checksum. | Licensed sample dataset and import specification: not yet published. |
| Data quality | Missing intervals, duplicate or out-of-order records, session boundaries and any filtering or adjustments. | Data validation behavior: to be confirmed for the tested build. |
| Account and instrument | Starting balance, account currency, contract size, price increment, volume increment and trading session. | Supported instrument and account models: to be confirmed. |
| Run identity | Run identifier, random seed where used, model settings and the complete output bundle. | Export schema and seed behavior: not yet published. |
Model Spread, Fees, Fills and Slippage
Write down the execution model before comparing strategies. The following calculation illustrates the documentation format; it is not an RTX5 test result.
Swipe horizontally to view all columns.
| Assumption | Illustrative value | Effect on the result |
|---|---|---|
| Buy quote | Bid 99.90 / ask 100.00 | A market buy starts from the ask, rather than a mid-price. |
| Sell quote | Bid 101.00 / ask 101.10 | A closing sell starts from the bid. |
| Adverse slippage | 0.02 per side | Buy fill 100.02; sell fill 100.98; gross difference 0.96. |
| Commission | 0.10 on entry and 0.10 on exit | Illustrative net result: 0.96 − 0.20 = 0.76 account-currency units. |
| Fill assumptions | Immediate full fills in this example | Partial fills, rejections, latency, financing and market impact are excluded. A tested model must declare its treatment of each. |
Validate, Walk Forward and Export
Use a repeatable review process before approving a strategy for another environment.
Swipe horizontally to view all columns.
| Check | Method | Record to keep |
|---|---|---|
| Repeatability | Repeat the run with the same source, data, build, model and seed. Compare fills and account totals. | Both run IDs and an explanation of any difference; deterministic guarantees remain to be confirmed. |
| Walk-forward example | Illustrative split: tune on months 1–6, freeze parameters, then evaluate months 7–8. Roll the window forward without using future data in earlier decisions. | Training and evaluation boundaries, frozen parameters and results for every evaluation window. |
| Sensitivity | Repeat with changed spread, commission and slippage assumptions. | Cost settings alongside the resulting trade count, equity curve and drawdown. |
| Export bundle | Keep configuration, data reference, strategy revision, trade log, equity series, metrics and reviewer notes together. | Actual export formats and a licensed example bundle: not yet published. |
| Approval | Review unexpected differences and document the accepted limitations before hosted execution. | Reviewer, decision, reviewed build and any conditions on deployment. |