Running backtests and interpreting results

When to use this guide

Use this guide whenever you want to evaluate a strategy revision on historical data before (or alongside) live trading.

By the end, you will be able to:

• Configure and submit a backtest job.
• Track backtest status.
• Access and interpret results and statistics.
• Use insights to create new revisions.

Key concepts

• A backtest is a historical simulation of a specific strategy revision over a chosen time window.
• Each backtest is:
  • Tied to a strategy and a revision version.
  • Uniquely identified by a hash of (strategy, revision, arguments) so identical runs are deduplicated.

Prerequisites

• You have at least one strategy with one or more revisions.

You do not need live trading accounts to run a backtest, because simulations operate on historical market data.

Step 1: Choose a strategy and revision

1. Navigate to the Backtests section or a dedicated backtesting entry point.
2. Click New backtest.
3. Select:
   • The strategy you want to test.
   • The revision of that strategy to backtest.

Remember: each revision is immutable, so re‑running a backtest on the same revision and inputs will always simulate the same logic.

Step 2: Configure backtest parameters

You will typically be asked to provide:

• Time range:
  • A start date/time.
  • An end date/time.
  • Internally, these may be converted to high‑precision timestamps, but you select them as normal dates.
• Markets or instruments:
  • Which assets or instruments the backtest should consider (for example, SOL‑PERP on Drift).
  • In many cases this is driven by your strategy revision’s settings.
• Execution model (if exposed):
  • Latency assumptions or execution delay.
  • Slippage or fee models.
• Initial conditions (if exposed):
  • Starting balances or positions.
  • Risk limits or margin settings.

The app may offer preset configurations (for example, “typical live‑like execution”) that translate into detailed arguments under the hood.

Step 3: Review and submit

Before submitting, Structure may show:

• The selected revision.
• The requested time window.
• Any relevant execution assumptions or market selections.

Review and confirm:

1. Double‑check that the revision is correct (for example, v3 vs v2).
2. Confirm the time window and, if relevant, markets or instruments.
3. Click Run backtest.

This creates a backtest job with:

• A unique ID (for referencing and monitoring).
• Initial status set to pending.

Step 4: Track backtest status

From the Backtests list you will see, for each job:

• Strategy and revision.
• Time range.
• Current status:
  • pending – queued and waiting to start.
  • running – actively simulating.
  • completed – finished successfully; results are ready.
  • failed – something went wrong; see error details.
• When it was created and last updated.

You can:

• Click into a backtest to see more detail.
• Refresh or wait for the UI to update as the job progresses.

Step 5: Handle failures

A backtest may fail for reasons such as:

• Invalid or inconsistent arguments (for example, time range outside available data).
• Insufficient historical data for the requested markets or timeframe.
• Exceeding internal resource or time limits.
• Transient infrastructure or data issues.

When a backtest is failed, open its detail page to see:

• A short error message.
• Any structured details the system provides.

Typical next steps:

• Adjust the time window to one known to have data.
• Simplify the configuration (for example, fewer markets) and retry.
• If the error persists and appears internal, contact support with the backtest ID.

Step 6: Access results and statistics

When a backtest reaches completed:

1. Open the backtest detail page.
2. You will typically find:
   • A Result download:
     • A file containing time‑series data for the simulation (for example, positions, P&L, trades).
   • A Statistics download:
     • Pre‑computed summary metrics, such as:
       • Total and annualized return.
       • Volatility.
       • Maximum drawdown.
       • Sharpe‑like ratios.
       • Win/loss rates and trade counts.

The UI may also present a visual summary, such as:

• An equity curve over time.
• Drawdown charts.
• Per‑instrument breakdowns.

Under the hood, both the result and statistics files are stored in durable object storage and securely streamed to you; you do not need to manage those locations directly.

Step 7: Interpret and act on results

Use the backtest outputs to answer questions like:

• “Does this strategy behave as expected in different market regimes?”
• “Are drawdowns acceptable for my risk tolerance?”
• “How sensitive is performance to parameter changes?”

Typical workflows:

• Compare multiple revisions:
  • Run the same time window on revisions v1, v2, and v3.
  • Compare statistics side‑by‑side.
  • Promote the strongest revision to live trading.
• Stress‑test across time periods:
  • Backtest across calm, volatile, and crisis periods.
  • Look for failure modes or instability.
• Calibrate position sizing and risk:
  • Use drawdown and volatility metrics to set leverage and notional size.

When you decide on changes:

1. Return to the Strategies section.
2. Create a new revision incorporating what you learned.
3. Optionally backtest again.
4. When confident, start the revised strategy in live trading.

Next steps

• If you haven’t yet connected accounts and run live, see Creating and running strategies and Connecting trading accounts.
• To understand how backtests fit into the broader mental model, read the Backtests concept page.