Model of cunning agents

TL;DR

This paper introduces a novel agent-based spin model for financial markets, where agents' opinions influence their actions, capturing market dynamics and aligning well with empirical data.

Contribution

It presents a new interpretation of the Potts model for financial markets, modeling agents as cunning with opinion-based interactions affecting their trading actions.

Findings

Model reproduces autocorrelation functions of returns

Captures distribution of times between large losses

Aligns well with empirical market data

Abstract

A numerical agent-based spin model of financial markets, based on the Potts model from statistical mechanics, with a novel interpretation of the spin variable (as regards financial-market models) is presented. In this model, a value of the spin variable is only the agent's opinion concerning current market situation, which he communicates to his nearest neighbors. Instead, the agent's action (i.e., buying, selling, or staying inactive) is connected with a change of the spin variable. Hence, the agents can be considered as cunning in this model. That is, these agents encourage their neighbors to buy stocks if the agents have an opportunity to sell them, and the agents encourage their neighbors to sell stocks if the agents have a reversed opportunity. Predictions of the model are in good agreement with empirical data from various real-life financial markets. The model reproduces the shape of the usual and absolute-value autocorrelation function of returns as well as the distribution of times between superthreshold losses.