Building the Next Generation of Algos
While many buy-side firms trade on a suite of existing algorithms, is it time for a new breed of algo? Kathryn Zhao, Global Head of Electronic Trading at Cantor Fitzgerald, sat down with Markets Media’s GlobalTrading to discuss the roadmap to building a new generation of algos.
Are algos commoditized?
Trading in today’s complex marketplace requires advanced technology solutions that are performant, robust and flexible. Plug-and-play algorithms used to be enough to satisfy the needs of large institutions who engaged in electronic trading; however, today it is a different story. Utilization of algorithmic trading is growing and so are the customization demands of the buy-side. Investments in analytics raise the bar for algo performance, while also varying the nature of a successful trading outcome. Modern, competitive electronic trading offerings are required to deliver customized solutions tailored to each client’s specific trading preferences, and that fits our business model perfectly.
A helpful analogy is that of Precision Medicine, a medical model, where healthcare is customized according to each patient’s specific characteristics. By targeting the context of each patient’s DNA structure and tailoring a medical treatment, doctors can achieve the best healthcare outcome with the fewest side effects. Likewise, the next generation of algos must target each client’s trading DNA and optimize trading performance for each client. This is the exact definition of our Precision Algo Platform.
With growing sophistication of quantitative models and ever increasing electronic solution variety, buy-side firms face a unique challenge: how to choose a product that is straightforward to use and yet easy to customize. We kept these considerations in our focus when we designed our algorithmic trading platform.
Our Precision Algo framework is modular and easily customizable, with speed of turnaround in mind, to help traders reduce their order trading flow and take advantage of price and liquidity conditions as they occur.
Replicate or blaze a new trail?
Our Precision Algo Platform was designed from scratch, with latest technologies, armed by experienced electronic trading professionals, devoted to pure meritocracy and achieving the best results for our clients.
When we began the process of developing a new suite of next generation algos nine months ago, we had a late-mover advantage. The electronic trading technology landscape has experienced dramatic changes over the past 20 years. Innovations in hardware, networking, and software have had an immense impact on current state of the art technology. To truly stand out from the rest of the competition, an electronic trading product must be performant from latency and throughput point of view and resilient to failure without compromising capabilities and quality of execution. We took and will continue to take a craftsman-like approach to building and continually improving our software. Our platform is designed with performance in mind.
For example, we use cutting-edge open-source libraries for inter-process communication and execution framework throughout the entire Client Gateway / Algo Engine / Exchange Gateway environment. We also use these highly efficient data structures for event sourcing to support full determinism in analysis and debugging. We pay a tremendous amount of attention to reliability and failover to eliminate the possibility of an outage and to minimize client impact if an outage does occur. Additionally, we rely on a comprehensive suite of automated testing and simulated execution tools to validate and guarantee the quality of our product.
With regard to quantitative research, we choose to use AI/ML only where there is a clear case for adding value. There are certainly cases when AI/ML- powered models give the best results. However, in multiple cases traditional statistical models still represent an overall better solution.
Clients’ key considerations and when to involve them?
The electronic trading business is highly competitive. In my view, state of the art technology and sophisticated quantitative research enable differentiated product. Execution quality, customization, access to liquidity and system stability are differentiators of a competitive low-touch offering.
Traders demand better performance and predictability of execution results, all without sacrificing the ability to source liquidity, transparency and control of their executions. With the launch of our Precision Algo Platform, we offer our clients a fully customizable algo suite optimized to each client’s trading DNA.
It is critical to engage clients actively at every stage of the process to ensure transparency and understanding of the full capabilities of the product, and to cater for clients’ current and future requirements. Underlying algos, such as VWAP / TWAP / POV, are the building blocks for “algo of algos”. As our Precision Algo Platform is modular and flexible, the majority of customizations are easy to achieve without having to directly alter the code base.
Extended into other asset classes
The flexibility incorporated in a modular algo framework should enable the next generation of algos to be adapted to new asset classes and market structures without a full rebuild. Most of the code base, for example, the Allocation and Order Optimization framework, the Macro-Trader (Scheduler) and the Micro-Trader (analytics-driven, quantitative model-based order placement logic) are common components of any electronic trading algorithm and can be shared across different asset classes.
The essential differences between asset classes are market data and their microstructure characteristics. Specialized quantitative models may need to be developed and calibrated to deliver better performing algorithms. Unique aggregators are also required to deliver a unified and normalized interface to both market data and market access.
As we roll out our next generation, cross-asset electronic trading product, we hope to help shape the industry standard for algo development.
New algorithms use quantitative approach to execution.
The volume grew by 41% in the second quarter of 2019.
The JavaFX-based interface now runs across a greater number of platforms.
The new offering sports 11 liquidity-seeking algorithms with more to come.
GMAG says vacation season is a good time for design, implementation and testing of machine-based trading.