Turn an AI Analyst question into a backtest you can inspect

Follow an AAPL RSI research question from a proposed setup to a simulated backtest, then inspect buy and hold, the trade ledger, CSV export, and a grounded follow-up.

Review the proposed rules before running. Read the benchmark, simulated account and individual trades together, because they describe different aspects of the result.

Make the question specific enough to test

This request fixes the symbol, dates, daily bars, direction, close entry, and RSI threshold. Settings that are not specified can appear as initially proposed assumptions. You can edit those settings before reviewing the cost and submitting a simulation.

Review, run, and inspect

A proposed test is a starting point. Review and edit setup loads the saved settings, including any earlier edits. The proposal’s assumptions remain visible for reference; the settings you review determine the run. Changing a setting clears the earlier review, so an old estimate cannot authorize a different test.

What the production run showed

The November 9 run covered January 2–June 27, 2025, rather than the requested June 30 end date. Its six trades produced one win and five losses. With $10,000 starting capital and 1% of equity at risk at the stop, the separate account simulation executed all six trades and ended at $9,472: a −5.3% return with 5.7% closed-trade drawdown. Buy and hold returned −17.3% over the available price window. The −38.8% sum of individual trade returns and 35.4% compounded one-position drawdown use different calculations from that sized account simulation; they are not interchangeable account results.

The CSV export contained six trade rows, ten columns, and UTC timestamps. The shared context preview excluded local account simulation values and balances. The follow-up explained the weak, small sample and the limits of comparing trade aggregates directly with buy and hold. Zero exit slippage and no modeled commissions are cost assumptions, not a promise of costless execution. This case used five AI credits for the proposal, one backtest credit for the run, and one AI credit for the follow-up; usage varies by request. Six trades cannot establish a reliable strategy or future performance.

Compare the same kind of return

Buy and hold measures holding the benchmark over its available price window. Simulated account return also reflects sizing, idle cash and the sequence of trades. A sum of individual trade percentages is another calculation; it cannot be read as the account’s return.

The ledger helps explain the headline. Check how many trades occurred, whether a few trades dominate, and how the exits behaved. The CSV is useful for checking those observations, while the report keeps the interpretation attached to the run.

Keep the result within its limits

One symbol and a short historical window answer a narrow question. Few trades, a particular market regime, imperfect coverage, and cost assumptions can all limit the conclusion.

Ask the Analyst to explain what the recorded result supports and what it does not. A backtest does not establish future profitability or suitability for your account. Backtests and paper results are informational and do not guarantee future results.