MesoMiner Overview
Advanced Strategy Mining Service
MesoMiner is a proprietary genetic algorithm-based optimization service that discovers high-performance options trading strategies through multi-objective evolutionary computation. Built on state-of-the-art optimization technology, MesoMiner explores vast strategy parameter spaces to identify Pareto-optimal trading configurations that balance multiple performance objectives simultaneously.
The service leverages the computational power of your FundPro cluster combined with Deltaray's proprietary algorithms to mine for strategies that would be impossible to discover through manual parameter tuning or grid search approaches.
Inspired by Google's groundbreaking AlphaEvolve paper, MesoMiner applies evolutionary computation principles specifically tailored for the options trading domain, combining advanced genetic algorithms with robust statistical techniques to ensure discovered strategies are genuinely predictive rather than artifacts of data mining.
Key Features
- Multi-Objective Optimization: Simultaneously optimize for CAGR, Sharpe ratio, Sortino ratio, and custom objectives
- Advanced Genetic Algorithms: State-of-the-art evolutionary algorithms with adaptive operators and constraint handling
- Massive Parallel Evaluation: Leverages MesoSim infrastructure for high-throughput strategy backtesting
- Checkpoint & Resume: Long-running optimizations with automatic state preservation and recovery
- Pareto-Optimal Solutions: Discover entire frontiers of non-dominated strategies for informed decision-making
- Full Integration: Results seamlessly available in MesoSim Portal, API, and Quantify for analysis
How It Works
MesoMiner employs sophisticated genetic algorithms to evolve populations of trading strategies over multiple generations:
- Population Initialization: Diverse strategy configurations are generated across the parameter space based on the user provided base strategy / starting point
- Fitness Evaluation: Each strategy is backtested using the specified date ranges and the performance is evaluated based on the provided objective(s)
- Selection & Evolution: High-performing strategies are selected for crossover and mutation
- Constraint Satisfaction: Strategies must meet minimum performance thresholds (e.g., minimum trades, Sharpe > 1)
- Convergence: The algorithm continues until discovering optimal strategy frontiers or reaching execution limits
Addressing Data-Mining Bias
MesoMiner incorporates sophisticated statistical techniques to minimize overfitting and ensure discovered strategies have genuine predictive power: These metrics are used during the evaluation phase of every executed trading strategy.
Deflated Sharpe Ratio
- Multiple Testing Correction: Accounts for the number of strategies tested during optimization
- Probability of Backtest Overfitting (PBO): Quantifies the likelihood that a strategy's performance is due to chance
- Adjusted Performance Metrics: Reports metrics that account for selection bias and multiple comparisons
Bias-Corrected & Accelerated (BCa) Bootstrap
- Robust Confidence Intervals: Provides accurate performance bounds using advanced bootstrap techniques
- Bias Correction: Adjusts for systematic biases in performance estimation
- Acceleration Adjustment: Accounts for the rate of change in standard error
These statistical safeguards ensure that MesoMiner's discovered strategies are robust and likely to perform well in live trading, not just historical backtests.
Please note, that MesoMiner is doing In-Sample Optimization.
It's the users' responsibility to validate the resulting backtests on a Out Of Sample period.
Service Model
MesoMiner operates as a managed optimization service to ensure optimal results:
- Request-Based Execution: Submit optimization requests with your specifications
- Expert Configuration: Our team configures the optimization based on your objectives
- Managed Execution: Runs are scheduled during specified time windows on your cluster
- Results Delivery: Discovered strategies are stored directly in your cluster for analysis
Please make sure to study the Service Process & Entitlements document for more details.