Options Strategy Optimisation

The Brief

The client had a working Pine Script strategy and wanted to find the best parameter configurations for trading options. The goal was to identify presets meeting a criteria of metrics and validate them against real options data, with robustness to hold up against real trading assumptions.

What We Did

Phase 1 - Initial Optimisation: We converted the Pine Script strategy to MQL5 to run large-scale parameter sweeps in MetaTrader. Targeting a handful of metrics and sorting by net profit, we explored holding 1DTE options overnight versus closing at market close.

Phase 2 Multi-Market Extension: As we developed a strong foundation to optimise presets, we extended the methodology across multiple three target markets. We accounted for commission impact and identified that commission disproportionately affects high-frequency presets, refining selection accordingly.

Phase 3 Volume Analysis: The strategy had a component that required volume data. However, our data’s volume did not match with TradingView’s volume. Because of this, we substituted volume with fixed values for a clean initial optimisation, then reintroduced and separately optimised the component which required volume using volume data downloaded from TradingView. A percentage of the total data was reserved as out-of-sample to prevent overfitting.

Phase 4 Options Backtests Options data was stored in an S3 bucket and we had to find an innovative method to fetch it. The first challenge was that the data size was too large and downloading it would take too long. The second challenge was that simply fetching the data on-demand would be extremely slow, even with the data downloaded.

Since OHLC data was readily available on the chart, we found that using a second-order Taylor expansion of the entry option price was a reliable and accurate way to approximate exit option prices. Using this technique, we did not need to fetch the full options chain and we were able to test hundreds of thousands of configurations across all presets relatively quickly.

Phase 5 Multi-Metric Optimisation: Across all three markets, we ran optimisations using three optimisation targets:

• Combined metric (net profit + profit factor)
• Profit factor + number of trades
• Custom Sortino ratio

Phase 6 Monte Carlo Analysis: When we got the results we were looking for, we performed a comprehensive Monte Carlo simulation analysis. The report covered equity curves, drawdown analysis, trade statistics, and profit factor distributions across all validated presets.

Results

The optimisation identified profitable presets across three major ETFs and timeframes, each meeting certain performance thresholds and holding up in out-of-sample validation.

The strategy has been forward-tested for over a year by the client and they continued to see profitability using the presets that we’ve idnetified.