We have designed and implemented a comprehensive set of frameworks that allows us to model the full end-to-end life-cycle of an investment strategy on a single platform, thereby eliminating inefficiencies.
Our systems ingest data from multiple sources and run comprehensive data quality checks to ensure that downstream modules get data of the highest quality.
We ingest data from across various asset classes, time periods ranging from tick data to annual financials and unstructured data.
A complete framework for the creation, management, and delivery of factors to various parts of the research process, that allows the creation of simple, composite, and cross-sectional factors from all kinds of datasets.
The framework is supported by a proprietary multilevel factor management that combines best practices from different data access paradigms for efficient storage and quick access.
Predictive models are at the heart of the investment process. Such models assign a probability of winning to every signal generated by a base investment strategy, and by trading only those signals that have the highest probabilities of winning, one can significantly improve the performance of the base investment strategy.
Our systems enable building predictive models for any given base investment strategy through a process of intelligent factor engineering and model building.
An event driven backtesting workflow ensures that one avoids common backtesting biases and enables time and memory efficiency. Essential backtest function such as the data handler, execution simulator, record keep, performance computation, etc. are available as ready-to-use modules that allows the researcher to focus only on the ideas being backtested without worrying about the other housekeeping tasks.
Realistic market impact models have been built using live trading data over decades to ensure that the simulated trades in the backtest environment are extremely close to real world fills.
Robust optimization methods are critical to building reliable models. Our optimization workflows overcome the limitations of standard optimization methods such as training-test splits and N-fold cross validations as these methods do not work for time-series data.
We utilise a walk forward based approach for optimising strategy parameters, which finds the optimal parameter set for each test period. The optimal parameters for any backtest period are identified by running the optimisation algorithm on the immediate history, just prior to the backtest period being considered. In addition to avoiding look-ahead bias, this approach also ensures that the strategy can dynamically respond to the changing market conditions.
Rigorous stress tests are performed to understand reactions to extreme market conditions. These include historical scenario analysis, data release impact analysis, co-moments analysis, factor risk decomposition, time stability analysis, risk simulation testing, return distribution analysis, amongst others.
A deep understanding of these return drivers helps the research team identify significant pockets and is a critical feedback to investment research, enabling improved iterations of the investment hypothesis.
Risk Monitoring & Management
Risk monitoring and management practices are implemented at multiple levels. These include diversification across multiple dimensions (styles, holding periods, asset classes, geographies), position sizing, unsupervised asset clustering to understand true exposure to asset classes and sectors, tail risk controls, drawdown-based scaling, etc.
Each risk management practice achieves specific desired outcomes, that help manage risks at trade, strategy, and portfolio levels.
A consistent tech-stack across research and production environments ensure that accepted research ideas flow seamlessly into production.
The execution systems can be interfaced with any third-party execution and monitoring systems.
Performance and Risk Monitoring
Detailed performance and risk monitoring dashboards are a critical tool for an investment manager. Our real-time monitoring systems generate customized reports that provide timely, actionable intelligence.
Our decades’ worth of experience allows us to extract meaningful metrics from raw data and provide informative summaries to CXOs.