Table of Links
Abstract, Acknowledgements, and Statements and Declarations
-
Background and Related Work
-
2010 Flash Crash Scenarios and 5.1 Simulating Historical Flash Crash
-
Mini Flash Crash Scenarios and 6.1 Introduction of Spiking Trader (ST)
-
Conclusion and Future Work
3.7 Simulation Dynamics
The whole simulation runs as follows. For each step, each trader collects and processes market information. Internal variables associated with each trader are calculated. According to agent type and values of internal variables, actions are taken by the traders. These actions include limit order submission, market order submission, and order cancellation. The programmed matching engine matches these orders and updates the state of the limit order book. Finally, transactions and limit order book status are published to all traders. The whole simulation procedure is shown in Algorithm 6.
We suggest that the proposed five types of traders reflect a sufficiently realistic and diverse market environment. According to O’Hara (1995), there are three major market-microstructure trader types: uninformed traders, informed traders and market makers. The noise traders in our model correspond to uninformed traders, while market makers in the proposed model obviously correspond to the market makers in literature. The remaining three types of traders represent informed traders in our model. Specifically, fundamental traders utilise exogenous information implied by the fundamental value, while the two types of momentum traders exploit the endogenous technical indicator information. In addition, among the informed traders some perceived trading opportunities are based only on an analysis of short-horizon returns, while others focus on market information revealed by long-term return horizons. This is reflected by the division of momentum traders into long-term and short-term momentum traders. Overall, a sea of different informed and uninformed traders in the proposed model compete with each other, with market makers providing liquidity and ensuring realistic limit order book behaviours. In conclusion, the proposed model with five types of traders represents a complete range of micro-behaviours of real financial markets.
3.7.1 Fundamental Value from Kalman Smoother
The only remaining unknown variable is the fundamental value of the stock. The simulation can proceed only if the fundamental value is known and is exogenously input to the model. One difficulty is the non-observability of the fundamental value. According to the economic literature, the fundamental value of a stock equals the expected value of discounted dividends that the company will pay to the shareholders in the future. However, this methodology requires extremely strong assumptions about the future dynamics of the stock dividends. Furthermore, this approach can never reflect the intra-day change of fundamental value, while the consensus fundamental value can indeed vary during the trading day due to the continuous feed of events and news.
Authors:
(1) Kang Gao, Department of Computing, Imperial College London, London SW7 2AZ, UK and Simudyne Limited, London EC3V 9DS, UK ([email protected]);
(2) Perukrishnen Vytelingum, Simudyne Limited, London EC3V 9DS, UK;
(3) Stephen Weston, Department of Computing, Imperial College London, London SW7 2AZ, UK;
(4) Wayne Luk, Department of Computing, Imperial College London, London SW7 2AZ, UK;
(5) Ce Guo, Department of Computing, Imperial College London, London SW7 2AZ, UK.
This paper is available on arxiv under CC BY-NC-ND 4.0 DEED license.
[6] The algorithm is already implemented in Python package "pykalman".
