Metaorder Data-Science: How Predictable Order-Flow Can Generate Random Prices in Intraday Markets

Date: Wed 1st April 2026

Abstract

The mixing of sequences of buying and selling parent orders in intraday financial markets exhibits long-range correlations. This is a widely known stylised fact of financial markets. A popular hypothesis for this stylised fact of market-order flow comes from the Lillo–Mike–Farmer (LMF) order-splitting theory. This can help explain both the concave shape of price impact and the emergence of correlations (via the Epps effect). Prices are then an emergent property arising from the interaction of orders. Measured prices are not themselves fundamental quantities in real financial markets. The LMF theory was recently verified on the Tokyo Stock Exchange (TSE). This phenomenology motivates the use of models that directly model limit-order book dynamics. Using Johannesburg Stock Exchange (JSE) data this talk will provide a practical overview of the LMF theory, a brief discussion of the data-science used to verify this on the JSE, and how this motivates both the use of Reaction-Diffusion equations using Random walk methods, as well agent-based models more generally. This will be used to briefly argue for the importance of developing approaches to modelling methods that do not directly use Ito calculus with its implied globally unique calendar time and reliance on continuous semi-Martingale methods, but to rather start with discrete and agent-based models and to explicitly use diffusion limits.  The talk aims to be applied with a more generalist introduction to the ideas and data-science.