Keep an experiment record
Before the first run, write down the following fields. These are a planning checklist, not completed backtest results. Use the same values for the candidate and its baseline.
| Strategy version and portfolio ID | Connect the result to the exact entry, exit and sizing rules |
| Dates and evaluation interval | Prevent a different test window from changing the comparison |
| Starting capital and benchmark | Make the baseline comparable |
| Cost and fill assumptions | Explain what the simulation models and leaves unresolved |
| Unseen period and candidate budget | Record how selection will be assessed before seeing the results |
No matching rows. Clear the filter to see all records.
Write the experiment before the strategy
State what the rule is intended to test. Specify the universe, entries, exits, sizing, interval, initial capital, date range and benchmark. For a broad equity strategy SPY can be appropriate; a single-name options strategy needs its underlying benchmark instead of an automatic SPY comparison.
The first run should answer a specific question. A search over many candidates needs a recorded budget and a held-out period so the winner is not merely the best retrospective fit.
The tool sequence
Inspect the discovered inputs before submitting a call. For a new candidate, validate and save it first; for a saved candidate, start with get_portfolio. Record the portfolio and backtest identifiers returned by the account tools.
| Validate a new candidate | build_portfolio | Canonical configuration and component issues; nothing is saved |
| Save the validated draft | create_portfolio | Use the reviewed configuration and retain the returned chat portfolio ID |
| Read the saved book | get_portfolio | Exact strategies, action sizing, and portfolio identity |
| Submit the historical test | backtest_portfolio | Explicit dates, interval, capital and benchmark |
| Poll the same operation | query_backtest_status | Terminal state and warnings; do not resubmit while running |
| Inspect the curve | query_backtest_history | Historical values for the returned backtest_id |
| Compare candidates | compare_backtests | Identical period and assumptions, with activity and risk |
No matching rows. Clear the filter to see all records.
Connect and inspect the candidate
Connect Claude Code through the current remote MCP OAuth guide. Ask for the complete strategy definition and confirm the exact portfolio ID before calling a backtest tool. Authoring and saving a candidate does not start a deployment.
Draft a daily SPY strategy using explicit entry, sizing and exit rules. Show the complete configuration and proposed backtest assumptions. Wait for my approval before submitting compute. Do not deploy or edit an existing live account.Read the completed historical result
Keep the backtest_id returned by the accepted operation and poll its status. Inspect warnings, drawdown, activity, exposure and the benchmark comparison. An empty history while a job is running is not a failed strategy and is not a reason to submit a duplicate run.
A historical curve models the configured strategy over past data. It is not the curve of a running paper or brokerage portfolio.
Test the frozen rule on an unseen period
Freeze the selected rule before opening the unseen period. Record rejected variants and why the selected configuration advanced. Compare assumptions as well as returns, particularly for options fills and unavailable data.
Claude Code can write a typed Python or TypeScript script when repeated research calls need stable configuration and idempotency keys. The SDK guides show the exact authoring and polling lifecycle.