Logbook
CLAVUS Updates
Throughout our development process, we’ve maintained a steadfast focus on trading highly liquid assets, primarily major currencies and commodities. This focus has allowed us to refine our strategies and risk management techniques in some of the most dynamic and challenging markets in the world.
Our commitment to continuous improvement and adaptation remains unwavering. Each version of our system builds upon the strengths of its predecessors, incorporating cutting-edge technologies and methodologies to deliver superior risk-adjusted returns. From our initial forays into quantitative investing to our current state-of-the-art platform, Clavus has consistently pushed the boundaries of what’s possible in algorithmic trading and portfolio management.
Version 1.9 February 5th, 2025
This groundbreaking update further refines our approach to portfolio construction and risk management, introducing dynamic subcomponent optimization and marking the public release of our Clavus-signum portfolio.
​
Key Enhancements:​
​
-
Introduction of adaptive evaluation and optimization for individual modules within larger systems
-
Deployment of advanced performance assessment for components using multi-layered metrics
-
Intelligent, dynamic replacement and integration of modules based on ongoing performance outcomes
-
Refined MOEP framework with enhanced constraint management and portfolio adjustment functions
-
Improved GBEF integration with deeper feature engineering and optimized hyperparameter calibration
​
This version marks a significant leap forward in our ability to construct resilient, adaptive portfolios. By evaluating and optimizing at the subcomponent level, we’re unlocking a new dimension of portfolio customization and risk management. Our enhanced MOEP implementation now explores a vast solution space with unprecedented efficiency.
​
This release represents a major advancement in our ability to design resilient, adaptive portfolios. By applying evaluation and optimization at the subcomponent level, we are unlocking a new dimension of portfolio customization and precision risk management. Our enhanced multi-objective evolutionary optimizer now explores the solution space with greater efficiency and consistency than ever before.
​
Our upgraded gradient-boosted ensemble framework now leverages a broader range of engineered features and applies adaptive hyperparameter calibration, resulting in superior out-of-sample prediction accuracy for drawdown detection and mitigation.
​
The dynamic subcomponent evaluation engine marks a fundamental shift in our research and development process. By continuously monitoring and analyzing the performance of individual strategy components, we can isolate and refine alpha sources with extraordinary granularity. This framework employs a multi-factor ranking process that evaluates not only base performance metrics but also higher-order statistical properties to ensure robust and persistent signals.​​
Version 1.8 November 19th, 2024
This landmark release introduces Clavus-Aurum, the first in our flagship A-Series of portfolios, built on advanced optimization frameworks to deliver stronger, more resilient risk-adjusted returns.
​
Key Enhancements:​
​
-
Introduction of multi-objective evolutionary optimizer (MOEP) for portfolio optimization
-
Implementation of advanced gradient-boosted ensemble framework considering return, risk, and correlation
-
Addition of sophisticated black swan resilience testing
-
Enhancement of Scoring+ methodology with dynamic weighting based on market regimes
​
The advanced gradient-boosted ensemble framework now considers a complex interplay of factors including expected return, various risk measures (VaR, CVaR, maximum drawdown), and a novel measure of portfolio fragility. This approach allows us to construct portfolios that are not only high-performing but also structurally robust to market shocks.
​
Our black swan resilience testing now incorporates extreme value theory to generate more realistic stress scenarios. We’ve developed a proprietary “tail dependency score” that quantifies a portfolio’s vulnerability to simultaneous extreme events across multiple assets.
​
The enhanced Scoring+ methodology now employs a dynamic weighting system that adapts to different market regimes. We use a hidden Markov model to identify regime shifts, allowing our scoring system to emphasize different factors as market conditions evolve.
​
This advanced framework represents a new benchmark in quantitative investment management. Initial results show significant improvements in drawdown reduction and risk-adjusted returns compared to our previous generation of portfolios.​​
Version 1.7 August 21st, 2024
This version marks a significant leap in our data granularity, correlation analysis capabilities, and risk management approach, revolutionizing our ability to capture and exploit market microstructure.
​
Key Enhancements:​
​
-
Transition to equity-based calculations for all performance and risk metrics
-
Upgrade of data granularity to 15-minute intervals for all calculations
-
Introduction of advanced multi-dimensional correlation analysis
-
Implementation of sophisticated drawdown recovery efficiency metrics
-
Enhancement of Scoring methodology to Scoring+, incorporating over 30 performance metrics
​
The shift to equity-based calculations represents a fundamental change in our approach to performance measurement and risk management. By basing all metrics on equity rather than balance, we’re achieving a more accurate representation of true portfolio risk. This allows for more precise position sizing and risk allocation, particularly in high-leverage environments.​
​
The transition to 15-minute interval data provides a more nuanced view of market dynamics, allowing for more precise strategy execution and risk management. This granular data is now used to compute a wide array of derived features, enhancing our market microstructure analysis capabilities.
​
Our multi-dimensional correlation analysis now employs advanced techniques to capture higher-order dependencies between assets and strategies. This approach allows us to identify and exploit complex, non-linear relationships that are invisible to traditional correlation measures.
​
The new drawdown recovery efficiency metrics go beyond simple measures like maximum drawdown. We now compute a “recovery surface” for each strategy, modeling the expected time to recovery as a function of drawdown depth and market conditions. This allows for more nuanced risk management and capital allocation decisions.​
​
The evolution to Scoring+ represents a quantum leap in our ability to evaluate and rank trading strategies. The system now incorporates over 30 performance metrics, including novel measures like the “alpha decay half-life” and “regime-conditional Sharpe ratio”. These metrics are combined using a dynamic weighting scheme that adapts to changing market conditions, ensuring that our strategy selection remains optimal across different regimes.
​
We’ve begun preliminary research into advanced genetic algorithms for portfolio optimization, including multi-objective evolutionary optimizer (MOEP). While not yet implemented, this research is laying the groundwork for significant improvements in our portfolio construction process in future releases.
Version 1.5 May 25th, 2024
This release focuses on enhancing our Scoring methodology and introducing more sophisticated risk measures, improving our ability to evaluate and manage complex trading strategies.
​
Key Enhancements:​
​
-
Expansion of Scoring methodology to include daily metrics and performance
-
Introduction of regime-dependent behavior analysis
-
Implementation of advanced tail risk measures
-
Development of an “alpha stability” metric
​
The Scoring methodology has been enhanced to Scoring+, incorporating a wider range of metrics and more sophisticated weighting schemes. This evolution allows for a more nuanced evaluation of strategy performance across various market conditions.​
​
The enhanced Scoring methodology now incorporates daily performance metrics, allowing for a more comprehensive evaluation of strategy behavior across different timeframes. We’ve introduced a “time-scale consistency score” that quantifies how well a strategy’s performance holds up across multiple time horizons.
​
Our regime-dependent behavior analysis uses a hidden Markov model to identify distinct market regimes. We compute conditional performance metrics for each identified regime, allowing us to select strategies that are robust across various market conditions.
​
The implementation of advanced tail risk measures improves our ability to estimate the shape of return distributions, particularly in the tails. This allows for more accurate risk assessment, especially for strategies with non-normal return profiles.​
​
Our new “alpha stability” metric quantifies the consistency of a strategy’s alpha generation over time, using changepoint detection algorithms to identify shifts in alpha production.
Version 1.5 January 13th, 2024
This update focuses on enhancing our risk management capabilities and improving the adaptability of our strategies to changing market conditions.
​
Key Enhancements:​
​
-
Implementation of adaptive risk allocation across subsystems
-
Introduction of anomaly detection for risk management
-
Enhancement of our liquidity management system
-
Upgrade of our backtesting engine to include more realistic transaction cost modeling
​
Our new adaptive risk allocation approach dynamically adjusts risk across subsystems based on their predicted risk-return characteristics and current market conditions. This allows for more efficient capital utilization and improved overall portfolio stability.
​
The introduction of anomaly detection enhances our ability to identify and respond to unusual market conditions or strategy behavior. This system uses machine learning techniques to detect outliers in multi-dimensional feature spaces.
​
We’ve upgraded our liquidity management capabilities by incorporating more detailed market data into our analysis. This allows for more accurate estimation of transaction costs and potential market impact, particularly for larger trades.
​
Our backtesting engine now includes more sophisticated transaction cost modeling, taking into account factors such as bid-ask spread and market impact. This provides a more realistic assessment of strategy performance, especially for higher-frequency trading approaches.