Start by framing the problem as a full research pipeline: hypothesis, data handling, strategy rules, backtest engine with realistic costs, performance metrics, and validation. Emphasize that mean reversion on minute bars requires careful treatment of microstructure noise and that validation is critical to avoid overfitting. Conclude with a statistical test to assess significance and discuss limitations.
Pro tip: Optiver values rigorous, production-quality backtesting: mention that you would use event-driven simulation with order book dynamics, and that you'd test the strategy's sensitivity to parameter perturbations and execution latency.
State the mean-reversion hypothesis (e.g., short-term price dislocations revert) and describe data cleaning: handle missing bars, adjust for splits/dividends, and compute returns or z-scores over a lookback window.
Define entry when price deviates by k standard deviations from a moving average, exit on reversion to mean or after a fixed holding period. Include stop-loss, take-profit, and time-based exits to control risk.
Use volatility-based sizing (e.g., target constant risk) and cap exposure. Model transaction costs (commission, fees) and slippage (e.g., fixed bps or volume-dependent) to reflect realistic execution.
Implement an event-driven backtest that processes minute bars sequentially. Calculate PnL, win rate, average trade return, max drawdown, volatility, Sharpe ratio, and turnover.
Split data into in-sample and out-of-sample (hold-out) or use walk-forward analysis. Apply a bootstrap or reality-check procedure (e.g., White's Reality Check) to test if performance is statistically significant and discuss overfitting risks.
AI-generated suggestions, not part of the candidate's original notes. May be inaccurate — verify before relying on them.