Strategy Backtest & Parameter Sweep
Published Aug 12, 2026 · Updated Oct 16, 2026
Strategy backtest checks a template’s parameters against historical market data, so you can see how they would have behaved before you create a strategy. The parameter sweep scans many combinations at once to compare returns and robustness. Go to Strategy → Strategy backtest; sign-in is required.
New to it? Run one version end to end with Your First Strategy Backtest. For checking inputs, saving versions, and creating a strategy from parameters, see Backtest Settings and Reproducibility. This page explains what the workbench does and how to read sweep results.
Which templates can be backtested?
When you create a backtest, the “Strategy” dropdown lists only backtest-capable templates. Today these are DCA Strategy, DCA Grid Strategy, Grid Strategy, Future Combo Strategy, Moving Average Breakout, CTA Trend v1, Narrow Stop-Loss, and Passivbot v8.1.0. Other templates (for example the Copy templates and Smart DCA Strategy) do not support backtesting yet; validate them on paper instead.
Besides New backtest on the backtest page, two entry points create an experiment from current parameters:
- Backtest current parameters on the configure and review steps of strategy creation;
- Backtest this strategy on a strategy’s detail page.
How is the workbench organized?
| Concept | What it is |
|---|---|
| Backtest experiment | Fixed to one strategy template (immutable after creation); you can keep adjusting parameters and the market environment |
| Backtest version | Every run creates a new version (v1, v2, …) holding its parameter snapshot and results |
| Parameter explorer | Adjust the market source, symbols, dates, initial equity, and the three parameter groups, then run |
| Parameter sweep | Scans parameter combinations in bulk and diagnoses returns, risk, and robustness |
| Saved backtest parameters | View each version’s snapshot, compare up to 2 versions, or bring a version’s parameters back into the draft |
The experiment list can be searched by name, filtered by strategy, and sorted by created time, latest run, return, max drawdown, Sharpe, or win rate.
The explorer groups parameters into Execution and exposure (leverage sets both the target notional exposure and the order leverage), Strategy parameters, and Risk control (stop loss, take profit, trailing exits, and so on—these feed the backtest and also appear in the sweep list). The Exposure preview shows target gross notional, target gross leverage, and estimated initial margin as a share of initial equity in real time.
Backtests run in your own browser. Before a run, the workbench checks the market data (whether markets exist, candle gaps, and so on). If you see “Incomplete market data”, switch the market source, adjust the dates, or remove the affected symbols from the draft.
How do I run a parameter sweep?
Choose parameters to scan
Open the Parameter sweep tab and pick the numeric parameters to scan. Required strategy parameters, symbols, dates, and initial equity must be filled in first.
Choose the sweep range
Under Sweep range, Neighborhood (around current) suits fine-tuning and Full range suits exploring from scratch. You can also use Set range from current values and then edit min, max, and step by hand. Fast search does a coarse pass first, then refines around the best region.
Start the search
Click Run search. The page shows the estimated number of runs and duration; above the combination cap it samples automatically and tells you how many of the requested combinations it ran.
Read the results
Each combination shows return, drawdown, max leverage, win rate, liquidation, and an Island risk score. Filter by minimum return or maximum drawdown, or Hide liquidated.
Use the parameters
Click Apply best / Apply best combination, or Use parameters on any row and choose Backtest with these parameters or Create a strategy with these parameters. After applying, run a regular backtest to produce a saved version.
| Plan | Parameters per sweep | Combination cap | Concurrency |
|---|---|---|---|
| Free / Plus | Up to 5 | 125 | Serial only |
| Pro / Pro Max | Unlimited | 2000 | 1–8 parallel workers (default 2) |
Higher concurrency uses more of your device’s CPU and memory; if the page stops responding, lower it and try again.
How do I tell whether parameters are robust?
| Signal | Meaning | What to do |
|---|---|---|
| Island risk score | Lower is more robust, graded Robust / Fairly robust / Medium / High risk; 50 and above signals overfitting risk | A high score means the parameters only work in a narrow range, and neighboring values may liquidate or collapse |
| Liquidation / liquidated combinations | Some values or combinations in the range liquidated | Before applying the top-return value, make sure it sits well away from them |
| Failed values / combinations | Some runs failed; the page lists the failure reasons | Fix the configuration or backtest environment and run the search again |
| Search result is out of date | You changed the configuration or environment after the search | You can still apply it, but run a backtest again to confirm |
The highest-return combination is often not the one to use. Check the island risk score and liquidation warnings first, and prefer a slightly lower return where neighboring values behave similarly. Historical backtest results do not guarantee future returns.
How long is sweep history kept?
Search history keeps recent results so you can review or apply them later. Summaries stay in the cloud for at most 7 days (latest 20) and are deleted automatically when they expire. Only per-combination summaries such as return and drawdown are stored—no fills—and the metrics are visible only to you and not server-verified. To keep a result long-term, apply its parameters and run a regular backtest.
If a finished search shows “Cloud save failed”, the local result is still available; click Retry save.
Next steps
Once you have chosen parameters, run them on paper for a while before considering live trading.