Benjamin Chung Email & Phone Number

http://ssrn.com/abstract=3853181- Trading signals must be binary, communicating timing information *only*, and must not carry magnitude information.- Frequency of correctness matters more than gain/loss asymmetry.- Portfolio diversification is achieved through the diversification of information, not "really" through diversification of assets.- Measurement error is handled naturally, not needing special consideration.- Linear models are particularly vulnerable to catastrophic tail events in ways that binary models aren't.- Unconstrained MVO weights may actually harm capital growth if magnitudes of expected returns approach/exceed that of their deviation, and should have their weights scaled down drastically if they do.- Models and signals of any/all varying lengths/holding-periods are treated identically, and can be assimilated together via a single procedure.- Asset baskets are actually coding blocks, and optimal portfolio construction is just image/JPEG encoding/decoding.- Capital flows are *not* Brownian, but *anti-Brownian*. ie. Brownian motion in reverse.- "Momentum" and "mean-reversion" factors model different phases of the same underlying process.- Financial "information" and trading "signals" finally have the same quantitative meaning as they do in other fields of study that deal with signals and information.

Toronto

• Worked with Blackheath’s Portfolio Manager to create Blackheath AI Global Macro FX strategy (BAI) – a systematic global macro FX strategy with month-long holding period. • Innovated and implemented new improvements to portfolio construction to reduce idiosyncratic risk of USD and global systematic risk. • Researched and backtested new machine learning models, and implemented improved models using tensorflow. • Original models achieved annualized Sharpe of +3.3 pre-Trump-China Trade War. Improved machine learning models achieved annualized Sharpe of +1.8 pre-COVID-19. • Portfolio volatility never exceeding its risk target despite Brexit, Trump U.S./China Trade War, and COVID-19-related global financial market and economic collapse in 2020 H1. • Pursuing new alpha research in the field of stock volatility using innovated Portfolio Theory and thermodynamics, describing stock volatility as a function of its debt and available float.

• Independently researched, developed, and implemented a scalable, purely systematic, long-horizon global macro quantitative FX allocation strategy. • Researched and implemented machine learning models in Java (and eventually Python) to create forecasts in currency markets. • Innovated a new Portfolio Theory upon the fundamentals of Information Theory, resolving many issues with Modern Portfolio Theory and Tactical Asset Allocation (what a trading signal is, low levels of input/training data, model output sensitivity to data, conviction, etc). • Architected, developed, and integrated all systems involved in the systematic trading system. Macro-economic data acquisition, ingestion, sanitation, normalization, monitoring; machine learning model creation, training, validation; portfolio construction and optimization; trade list creation and execution. • Integrated the strategy’s tactical asset allocations with Interactive Brokers trade execution. • Achieved +400% total return with annualized Sharpe of 2.5 over the first 15 months.

Toronto

• Created systems to ensure data correctness for backtests and model ingestion. • Conducted research with Portfolio Managers on macroeconomic forecast models for GTAA's systematic emerging market currencies strategy. • Developed and performed backtesting on new models. • Integrated new factor models into production infrastructure. • Assisted in the daily monitoring and management of portfolio trades, positions, and risk.

Sourceforge

• Created the world's first opensource object-oriented deep learning library (Java) and released it under the GPLv2 in 2007. https://sourceforge.net/projects/jarbm/ • Assissted post-graduate students around the world in its use for their theses.