Help me turn a market idea into a testable trading-strategy specification. This is educational research, not personalized investment advice, and I will not treat the output as a recommendation to buy, sell, or hold. Market and instruments: [MARKET AND INSTRUMENTS] Hypothesis or edge I want to test: [HYPOTHESIS] Timeframe, data period, and data source: [TESTING DATA] Trading constraints: [CONSTRAINTS] Risk limits and assumptions: [RISK PARAMETERS] Produce a strategy research brief with: 1. A falsifiable hypothesis and exact entry, exit, stop, profit-taking, position-sizing, and no-trade rules. Define every term so two people could implement the same rules. 2. A data specification: required fields, session and timezone conventions, corporate-action treatment where relevant, survivorship-bias risk, missing-data handling, and the difference between signal time and executable fill time. 3. A backtest plan in numbered steps, including train/test separation or walk-forward testing, realistic commissions, spread, slippage, borrow cost, funding, and liquidity constraints as applicable. 4. A results table template with number of trades, win rate, average win/loss, expectancy, profit factor, maximum drawdown, exposure, turnover, and out-of-sample results. Show the formulas for expectancy and position size using my assumptions. 5. Failure modes: overfitting, look-ahead bias, regime dependence, concentrated exposure, gap risk, and execution risk. State what result would invalidate the hypothesis. 6. A paper-trading checklist before any capital is used. Do not invent prices, returns, news, historical results, or data access. Ask up to 3 clarifying questions only if a required input is missing. Before answering, check that every rule can be tested without future information and that the position-size math never exceeds my stated risk limit. Flag assumptions I must verify with a regulated professional, broker, tax adviser, or current official source.
Fill in
| Placeholder | What to enter | Example |
|---|---|---|
| [MARKET AND INSTRUMENTS] | Name the asset class, instruments, exchange or venue, and liquidity limits you want to study. | Highly liquid U.S.-listed ETFs only: SPY, QQQ, IWM, and TLT; no options, no leverage, and average daily volume above 5 million shares. |
| [HYPOTHESIS] | State the observed pattern or idea you want to test, without claiming it is proven. | After a large one-day decline, broad index ETFs may mean-revert over the next two to five sessions when the longer-term trend remains positive. |
| [TESTING DATA] | Describe the timeframe, historical period, bar frequency, and source of data you have or plan to use. | Daily OHLCV data from January 2012 through August 2026, adjusted for splits and distributions; I can export data from my brokerage platform. |
| [CONSTRAINTS] | List practical limits such as holding period, order types, account restrictions, and markets you will avoid. | End-of-day signals, next-day market-on-open entries, maximum five trading-day hold, one position at a time, no trades around data gaps. |
| [RISK PARAMETERS] | State maximum risk per trade, portfolio exposure, drawdown limit, capital amount, and cost assumptions. | Research account size $25,000; risk no more than 0.5% of capital per trade; 20% maximum portfolio exposure; assume $0 commissions and 0.03% slippage each side. |
How to use
- Replace every bracketed field with a narrow market, test period, and explicit risk constraint.
- Ask the AI for the research brief, then implement the rules in a backtest using data you can inspect and reproduce.
- Check timestamps, fill assumptions, transaction costs, and out-of-sample results before looking at headline returns.
- Follow up with: “Create pseudocode for these rules and list every field it needs, without adding indicators I did not specify.”
Variations
Backtest audit
Use this when a strategy backtest looks unusually strong and you need to inspect it.
Audit this trading backtest for false confidence. Strategy rules: [RULES]. Market and sample period: [MARKET]. Entry and fill assumptions: [EXECUTION]. Reported results: [RESULTS]. Code or pseudocode: [LOGIC]. Produce a prioritized audit covering look-ahead bias, survivorship bias, corporate actions, data snooping, overlapping trades, position sizing, fees, slippage, liquidity, parameter selection, and out-of-sample testing. For each issue, state the exact test or code change needed. Do not estimate corrected returns or recommend trades. Ask up to 3 questions only if a required input is missing.
Risk sizing plan
Use this when rules exist but position size and loss limits are unclear.
Create a risk-sizing worksheet for a proposed trade strategy. Account value: [ACCOUNT SIZE]. Entry, stop, and instrument details: [TRADE DETAILS]. Maximum loss per trade and portfolio limits: [LIMITS]. Costs, leverage, and gap-risk assumptions: [ASSUMPTIONS]. Show the formulas and worked arithmetic for risk per share or unit, maximum position size, notional exposure, and worst-case loss under my stated assumptions. Include checks for gaps, correlation, liquidity, and broker margin rules. This is educational, not personalized advice. Ask up to 3 questions only if a required input is missing.
Trade journal review
Use this when you want to learn from completed paper or live trades.
Analyze my de-identified trade journal for process quality, not prediction. Strategy rules: [RULES]. Journal entries: [TRADES]. Benchmark or market context: [CONTEXT]. Identify rule adherence, recurring execution errors, unplanned risk, outcome bias, and samples too small for conclusions. Produce a table with trade ID, planned versus actual action, rule status, evidence, and a corrective process change. Separate process failures from normal losses. Do not infer future returns or tell me what to trade. Ask up to 3 questions only if a required input is missing.
Tips
- Write entry and exit rules in terms available at the moment of decision; a rule that relies on a day’s closing value cannot be filled at that same close without an explicit assumption.
- Use a test period that includes different volatility and rate environments, then reserve a later period for out-of-sample evaluation.
- Model costs pessimistically for the instrument and order type; small edges often disappear after spread, slippage, financing, and tax effects.
- Keep strategy research separate from execution: decide position limits and stop conditions before reviewing a specific trade setup.
FAQ
Can AI find a profitable trading strategy?
It can help formalize and test an idea, but it cannot establish that an edge will persist. Historical results are sensitive to data, assumptions, costs, and changing market regimes.
What makes a backtest credible?
Clear rules, reproducible data, realistic fills and costs, no future information, and a genuine out-of-sample test are the minimum. Paper trading adds an execution check.
Should I use AI-generated code to trade automatically?
Do not deploy it without independent testing, broker safeguards, position limits, and monitoring. Automated trading can create losses quickly when assumptions or integrations fail.