Research

Take a deep-dive into our latest quantitative research projects, algorithmic trading innovations, and mathematical finance publications. Our work bridges theoretical foundations with practical high-frequency trading applications.

Publications

Low Latency Crypto Market Making Through HJB with Jump Diffusions

Arithmax Research Team
Peer Review Jul 31, 2025

Advanced market making strategy utilizing Hamilton-Jacobi-Bellman equations enhanced with jump diffusion modeling for cryptocurrency markets. Achieves superior risk-adjusted returns through sophisticated mathematical optimization and CUDA-accelerated computation.

Enhanced HJB equations with jump diffusion modeling (λ = 3.2 jumps/day), achieving 1.48 Sharpe ratio in BTC/USD backtests, reducing inventory risk by 62% vs traditional quoting strategies
Implemented finite difference methods on 201×201 grid with 0.89ms latency using CUDA-optimized kernels (136× speedup vs CPU), maintaining 97.4% theoretical PnL efficiency after transaction costs
Developed flow toxicity tracking system, reducing adverse selection by 41% through real-time order imbalance detection (R² = 0.79) vs actual toxic flow events
Market Making HJB Equations Jump Diffusion CUDA Cryptocurrency
Read Paper →
HJB Market Making Visualization

Cross-Regime Performance Analysis of an Algorithmic Strategy for a Diversified Leverage Index Fund Portfolio

Arithmax Research Team
HKUCS Jul 31, 2025

A comprehensive framework for diversified leverage strategies designed to capture enhanced returns through leveraged ETFs while managing downside risk through strategic asset allocation and volatility-adaptive methodology across different market regimes.

Dynamic rebalancing with momentum preservation across leveraged equity, defensive bonds, and commodity diversification
Multi-asset risk parity approach across leveraged instruments with volatility-adaptive allocation methodology
Tariff-resilient sector diversification demonstrating superior risk-adjusted returns with controlled drawdowns during market stress
Mathematical framework addressing critical challenges in leveraged portfolio construction and systematic rebalancing
Portfolio Theory Leverage Risk Parity ETF Strategy Volatility
Read Paper →
Leverage Portfolio Analysis

Accelerating Python-based Algorithmic Trading Code Through GPU Programming

Arithmax Research Team
HKUCS Mar 3, 2025

Proof-of-concept research exploring GPU acceleration optimization for Python-based algorithmic trading systems. Demonstrates significant performance improvements through CuPy and Numba implementations, enabling faster backtesting and potentially real-time strategy execution.

Comprehensive analysis of Python performance bottlenecks in large dataset processing and complex algorithmic calculations
GPU-accelerated swing-high trading strategy implementation using CuPy and Numba frameworks
Significant execution time improvements comparing CPU-bound vs GPU-accelerated implementations
Practical framework for enhancing Python-based algorithmic trading system efficiency
GPU Computing Python CuPy Numba Backtesting
Read Paper →
GPU Trading Acceleration