A deterministic behaviour for realistic price dynamics

International audience; In recent years, many studies on financial markets have relied on artificial agents, whether for the evaluation of strategies, the study of price dynamics or the efficient execution of orders. The behaviours used in those studies, often Zero-Intelligence Traders, fundamentalists or chartists, are stochastic and therefore non-deterministic, mainly because such agents easily yield a market that continuously fixes prices. We argue here that a rational and fully deterministic behaviour is sufficient both to reproduce the classic stylized facts of the field, but also to ensure that agents with different initial parameters have different opportunities to enrich themselves. To illustrate this purpose, we introduce Deterministic Artificial Traders, or DAT, and we show their performances in several situations. This result illustrates the fact that financial markets are a complex system, as some deterministic behaviours lead to some randomness in the market, both at the macroscopic and microscopic levels