Information Technology Reference
In-Depth Information
trading instruments and rules. The second one is geared at designing a general envi-
ronment with flexible settings and functionalities that can accept heterogeneous agents
populations.
We first list a limited set of these models belonging to these philosophies and that
point out some design questions. - Altreva Adaptive Modeler is a tool for creating
agent-based market simulations for a specific problem: stock price forecasting [4]. The
author, among other questions, describes a problem of a memory limitation during the
framework processing.
- JLM market simulator is another tool that investors can use to create a market
model using their own inputs. The authors (Jacobs Levy Equity Management, Inc.)
conclude that it is not an easy task to build a complex asynchronous simulation with
reasonably realistic properties. The first highlighted source of potential problems is the
diversity of agents' behaviours (for instance, there are only mean-variance investors in
JLM); a second one is the specification of user's trading strategies, that does not require
the user's programming skills [5].
- Ascape is a general agent-based framework, developed at Brookings Institute in the
90's [6], that is actively used in financial market modelling. Its developers discuss de-
sign possibilities to express the same basic modelling ideas in one way and have them
tested in many different environments and configurations.
Most of present artificial market platforms suffer from a lack of flexibility and must be
viewed as softwares rather than APIs, because they are mainly oriented for solving a
given problem and, most of the time, cannot easily be used to explore a wide range of
financial issues. This is due to some structural choices that are made by the developers
during the coding phase.
In this paper we present the ArTificial Open Market API (here-after ATOM, see
http://atom.univ-lille1.fr) and focus our attention on some important issues of agent-
based stock market design we have faced during its development. Among others, we
make a specific point on the ability of this new, generic architecture to overcome some
issues mentioned previously. After a general exposure of our main choices in terms of
software engineering, we propose our solution for several problems : i) how should we
manage information to reprocess real market order-flows (necessity of a unique order
identity?), ii) how should the scheduling system be organized, iii) how can we integrate
a human agent in the simulation process, with his own strategy, (”human in the loop”)
...? In addition, we tackle an important design question in the representation and struc-
ture of the trading agents, from zero-intelligence agents to sophisticated intelligence
ones, like technical traders. We emphasize the importance of calibration and validation
aspects of agent-based market and run several series of validation tests in order to show
the ability of ATOM at replicating real market price motions.
2
The Artificial Trading Open Market API
ATOM is an environment for Agent-Based simulations of stock markets. At the present
moment, it is realized based on the architecture close to the Euronext-NYSE Stock Ex-
change one. Agent-Based artificial stock markets aim at matching orders sent by virtual
traders to fix quotation prices. Price formation is ruled by a negotiation system between
