Enhanced fill probability estimates in institutional algorithmic bond trading using statistical learning algorithms with quantum computers

Battle Count: 75

Directly addresses a core problem in institutional algorithmic trading: fill probability estimation for RFQ responses in corporate bond markets. Uses real production-scale trading data from HSBC. The framework is designed for practical deployment in low-latency trading systems with decoupled offline quantum feature generation. Results show up to 34% improvement in out-of-sample AUC for fill probability prediction. The methodology is specifically tailored for market-making dealers responding to RFQs, with considerations for model risk management and regulatory controls. Highly relevant for quantitative trading practitioners exploring quantum computing as a complementary tool.

Requires access to IBM Quantum hardware (Heron processors with 109+ qubits), Qiskit framework with Estimator Primitive, Matrix Product State simulation, and error mitigation techniques (Pauli Twirling, TREX). The classical-quantum event matching protocol adds complexity. The backtesting framework involves extensive hyperparameter tuning across multiple model types. However, the quantum component is decoupled as an offline process, simplifying online deployment. The overall system requires expertise in both quantum computing and financial ML modeling.

The paper provides detailed quantum circuit parameters (Table 1), hyperparameter search spaces (Table A.3), backtesting protocol settings (Table A.4), and describes the PQFM methodology with mathematical formulations. However, the trading dataset is proprietary HSBC data (1,073,926 RFQs over 294 trading days), and access to IBM Quantum hardware (ibm_torino, ibm_fez, ibm_marrakesh) is required. The event matching protocol and backtesting framework are described in sufficient detail for replication with similar data, but exact reproduction requires the same proprietary dataset and quantum hardware access. No public code repository is mentioned.