Reproduce “StochLOB: Heavy-Tailed Score Matching for Robust Limit Order Book Trend Prediction” on OpenResearch
Turn your favorite coding agents into research agents that review literature, develop hypotheses, and run experiments.
Universal app for Apple silicon and Intel. Requires macOS 11 or later.
About this paper
Arshia Abolghasemi, Seyed Reza Tavakoli, Parsa Naderi, Nazanin Mirzaei, Behrad Jahani, Saba Yousefzadeh, Tahereh Javaheri, Mohammad Hosein Rohban and 2 more
StochLOB predicts three-class limit-order-book trends by training a classifier with corruptions and an auxiliary score-matching objective. At inference, it uses one deterministic classifier pass. On 3,185 top-decile jump-residual Coinbase BTC windows at horizon 100, its alpha-stable variant reaches macro-F1 0.637, versus 0.567 for matched Gaussian StochLOB and 0.597 for Linear VAR; this is not a general test-set gain.
Summary by alphaXiv.
Tip: change arxiv.org to autoarxiv.org in any arXiv abstract link to open the paper here.