New arXiv methods for LOB and IVs
Recent arXiv posts highlighted in social feeds include 'Bridging the Reality Gap in Limit Order Book Simulation' and work on sensitivity analysis for instrumental variables under relaxed assumptions — directly relevant to simulation‑based trading models and causal work in finance. Both papers were flagged as ready‑to‑use resources for advanced quant projects. (x.com) (x.com)
ArXiv submission 2603.24137 introduces a practical, interactive limit‑order‑book simulator for large‑tick assets authored by Patrick Noble, Mathieu Rosenbaum, and Saad Souilmi and posted March 26, 2026. (arxiv.org) The simulator projects full book state onto a low‑dimensional representation built from spread and volume imbalance for robust, data‑driven parameter estimation. (arxiv.org) Event‑timing is explicitly calibrated to market microstructure and the authors report a pronounced mode at exchange round‑trip latency consistent with latency races, while a feedback term that accumulates signed trade flow via a power‑law decay kernel reproduces concave market impact and partial post‑trade reversion. (arxiv.org) The paper presents several stock‑level strategy case studies showing that simulated profitability is highly sensitive to execution parameters and frames a reproducible workflow—“project, estimate, validate, adapt”—for building simulators suitable for execution testing and P&L analysis. (arxiv.org) ArXiv submission 2603.25529 by Pedro Picchetti (posted March 26, 2026) develops a breakdown‑frontier sensitivity analysis for instrumental variables that jointly relaxes monotonicity and independence assumptions using c‑dependence from Masten & Poirier (2018). (arxiv.org) Picchetti derives identified sets for LATE and ATE whose bounds are explicit functions of two sensitivity parameters, proposes consistent sample‑analogue estimators with a bootstrap inference procedure, and reports Monte‑Carlo evidence for the estimators’ finite‑sample performance. (arxiv.org) An empirical application re‑examining Angrist & Evans (1998) on family size and unemployment finds that conclusions about the causal effect are “highly sensitive” to simultaneous violations of independence and monotonicity, illustrating practical stakes for IV work in applied finance and labor economics. (arxiv.org)
