Our Scientific Investment Process
A Ph.D. Level Algorithmic Trading Factory
- Our team of Ph.D. has created a robust and sophisticated infrastructure to build and monitor portfolios of well-diversified algorithmic trading strategies.
- We created dozens of algorithmic trading strategies based on an innovative mix of econometrics and machine learning.
- All our strategies stem from our proprietary research.
- We backtest our strategies with a variety of approaches to provide a sound scientific validation.
- A minimum of 6 months of successful paper-trading is performed before a strategy is proposed to investors.
Step 1 : Research
- We constantly read the most advanced studies in the field of machine learning and finance to offer the newest technologies.
- We follow a rigorous scientific approach in developing our strategies, and a massive bibliography always backs our results.
- Our models are based on our proprietary research, intentionally kept private for our clients.
- All our models are backtested, and results are reported in a dedicated appendix.
- Our research team comprises a heterogeneous group of Ph.D. graduates in economics, statistics, and engineering.
- We deliver customized research to our clients upon request.
Step 2 : Data Management
- We employ NUMBA and RAPIDS to analyze Tick Data in Python.
- We use C++ with multithreading and the CUDA to analyze BIG-DATA.
- We work with data from several sources: OptionMetrics, CRSP, Compustat, IBES, ICE, …
- Our data are cleaned and reorganized to maximize their informative content.
- Streaming data comes from InteractiveBrokers.
Step 3 : Strategy Building
- Our approach to building trading and investment strategies is based on two steps: first, the building of the system in a traditional framework, and second, the application of machine learning to raise its predictive power.
- We are constantly building many strategies with different economic rationales to achieve a more robust diversification: event-based, option-based, relative values, directionals, high-frequency, …
- We avoid data mining as much as we can.
Step 4 : Backtesting
- We developed a vast backtesting engine, including over 30 different statistical procedures to minimize the occurrence of false discoveries.
- We are constantly updating the procedures included in our backtesting software which we consider one of the primary sources of our competitive advantage.
- We dedicated above 6000 hours of work to this Backtesting software: we believe that the time dedicated to building a robust backtesting engine is never enough.
Step 5 : Deployment
- We run our strategies (C++/Python) with the Interactive Brokers API on dedicated Amazon clouds.
- Our execution strategies are based on a deep study of the market microstructure and leverage the predictive power of machine learning to minimize the price impact and the other transaction costs.
- We use C++, CUDA, and CPU multithreading to perform complex analyses with data streaming.
- We employ web scraping to import data from websites into our trading systems directly.
Step 6 : Risk Management and Monitoring
- Our forward-looking risk management is based on a unique mix of signals extrapolated from options and other derivatives.
- We monitor the performance of our trading strategies to verify that the current results are coherent with the results from our backtesting analysis.
- We monitor the market conditions to select the trading strategies to include in our active portfolios of trading systems.
- Our performances are certified by a third party, public and transparent.
Step 7 : Portfolio Management
- We apply the most advanced portfolio optimization algorithms to our portfolio trading systems.
- We developed various complex models to estimate the possibility of a market crash and assume a proactive approach.
- Our optimization procedures include dedicated features for the analysis of transaction costs.
- We tailor machine learning-based portfolio optimization algorithms for our clients upon request.