Smart Wheel Engine: A Machine Learning-Based Options Strategy Evaluation Engine

Core Information

This article introduces the open-source project Smart Wheel Engine, maintained by MertYakar66 and released on May 24, 2026 (GitHub link: https://github.com/MertYakar66/smart-wheel-engine). Using machine learning techniques, the engine automates the evaluation of the Wheel strategy (combining put selling and covered call writing) through probabilistic modeling, expected profit and loss analysis, and market state filtering, providing investors with data-driven decision support.

Project Objectives

To solve the time-consuming and error-prone problems of manual Wheel strategy evaluation, and help identify attractive trading opportunities.

Project Background

In financial derivatives trading, the complexity of options strategies is a major challenge for investors. The Wheel strategy (a cyclic strategy combining put selling and covered call writing) has gained attention for its stable returns, but manual evaluation of its risk-return ratio, probability distribution, and market adaptability is inefficient and error-prone. Thus, the Smart Wheel Engine was developed.

Wheel Strategy Logic

• Phase 1: Sell Put Options: Collect premiums. If the price is above the strike price at expiration, keep the premium; if below, buy the asset at the strike price to reduce holding costs.
• Phase 2: Covered Call Writing: If assigned to buy the asset, hold the asset while selling call options to collect premiums; if the price is above the strike price, the asset is sold, gaining the strike price plus premium income.

Strategy Advantages and Disadvantages

• Advantages: Continuously collecting premiums reduces holding costs; stable returns in oscillating or moderately rising markets.
• Disadvantages: May buy assets at high prices during unilateral declines; may sell too early and miss gains during unilateral rises.

Probabilistic Modeling

The engine's core is a probabilistic model that estimates the future price probability distribution of the underlying asset. Unlike the Black-Scholes model, which assumes a log-normal distribution, it can capture non-normal characteristics of real markets such as fat tails and volatility clustering. Accurate probability estimation helps judge trading risk exposure.

Expected Profit and Loss Analysis

Based on the probabilistic model, calculate the expected profit and loss (probability-weighted average of all outcomes) to reveal the true expected value of the trade, avoiding being misled by high premium yields while ignoring risks.

Market State Filtering

Identify current market states (bull/bear market, high/low volatility, etc.) and dynamically adjust strategy evaluation criteria. For example, selling put options is riskier in high-volatility rising markets, while the Wheel strategy's premium-collecting feature is more optimal in oscillating markets.

ML Application Advantages

1. Capture Market Microstructure: Neural networks effectively model high-frequency features such as order flow imbalance and bid-ask spread dynamics.
2. Handle High-Dimensional Features: Address multiple factors like underlying prices, volatility surfaces, interest rate term structures, and market sentiment.
3. Non-Parametric Flexibility: Adaptively fit complex patterns in data without preset function forms.

Challenges

• Overfitting Risk: Financial data has low signal-to-noise ratio, so out-of-sample generalization ability is key. Need to balance model complexity and generalization ability.

1. Data Quality and Frequency: Model performance depends on data quality; historical prices, volatility, fundamentals, etc., need cleaning and preprocessing.
2. Model Updates: Markets evolve dynamically; regular retraining or online learning is required to adapt to changes.
3. Risk Management: Black swan events cannot be predicted; strict position management, stop-loss mechanisms, and diversified investments are necessary.
4. Execution Costs: Theoretical returns need to deduct transaction costs like commissions and slippage; high-frequency strategies are more sensitive.

Smart Wheel Engine demonstrates the potential of machine learning to empower traditional options strategies, improving the efficiency of Wheel strategy evaluation through a data-driven framework. However, tools are only auxiliary; successful investment requires market understanding, risk management, and continuous learning. This open-source project is of reference value to quantitative trading enthusiasts and options researchers, and is worth attention and learning.