Bridging the Reality Gap in Limit Order Book Simulation

Battle Count: 59

This paper is directly and fundamentally relevant to quantitative trading. It addresses the core problem of building realistic limit order book simulators for strategy evaluation, execution cost estimation, and risk management. The three extensions (state projection, latency-aware timing, market impact feedback) target precisely the gaps that cause simulation-to-reality discrepancies in practice. The case studies demonstrate how profitability becomes highly sensitive to execution parameters when impact and latency are properly modeled. The work is immediately actionable for market makers, HFT firms, and brokerages.

The core QR simulation loop is straightforward (Algorithm 1). State projection to (Imb, n) simplifies estimation significantly. The GMM fitting for inter-event times and the sum-of-exponentials kernel approximation add moderate complexity. The impact feedback mechanism requires maintaining a running state variable with O(K) updates per event. The main complexity lies in: (1) proper data preprocessing (volume aggregation for Create and Trade events), (2) calibration of impact parameters via Monte Carlo, (3) handling queue reindexing during price moves, and (4) the race/fill logic for interactive simulation. Overall, a competent quantitative developer could implement this in 2-4 weeks.

Code is available on GitHub (mentioned in footnote 1). All parameters are explicitly stated (tau=50s, beta=1.5, m=0.036, 21 imbalance bins, K=12 exponential components). Data source (Databento MBP-10) is commercial but well-identified. The methodology is presented as a clear recipe with algorithm pseudocode. However, Databento data requires a paid subscription, and some calibration details (binary search over 100,000 paths) are computationally intensive.