Which trading agent is best? Using a threaded parallel simulation of a financial market changes the pecking-order

Abstract

This paper presents novel results, generated from a new simulation model of a contemporary financial market, that cast serious doubt on the previously widely accepted view of the relative performance of various well-known public-domain automated-trading algorithms. Put simply, we show here that if you use a more realistic market simulator, then trading algorithms previously thought to be the best-performing are shown to be not as good as people think they are, and some algorithms previously thought to be poor performers can be seen to do surprisingly well. Automated trading is now entirely commonplace in most of the world's major financial markets: adaptive algorithmic trading systems operate largely autonomously, interacting with other traders (either other automated systems, or humans) via an electronic exchange platform. Various public-domain trading algorithms have been proposed over the past 25 years in a kind of arms-race, where each new trading algorithm was compared to the previous best, thereby establishing a "pecking order", i.e. a partially-ordered dominance hierarchy from best to worst of the various trading algorithms. Many of these algorithms were developed, tested, and evaluated using simple minimal simulations of financial markets that only very weakly approximated the fact that real markets involve many different trading systems operating asynchronously and in parallel. In this paper we use BSE, a long-established public-domain market simulator, to run a set of experiments generating benchmark results from several well-known trading algorithms. BSE incorporates a very simple time-sliced approach to simulating parallelism, which has obvious known weaknesses. We then alter and extend BSE to make it threaded, so that different trader algorithms operate asynchronously and in parallel: we call this simulator Threaded-BSE (TBSE). We then re-run the trader experiments on TBSE and compare the TBSE results to our earlier benchmark results from BSE. Our comparison shows that the dominance hierarchy in our more realistic experiments is different from the one given by the original simple simulator. We conclude that simulated parallelism matters a lot, and that earlier results from simple simulations comparing different trader algorithms are no longer to be entirely trusted. 

 Financial Markets | Market Simulators | Simulation Methods |  Trading Agents | Experimental Economics 

References

1. BSE, 2012. Bristol Stock Exchange. GitHub repository at  https://github.com/davecliff/BristolStockExchange
   
2. Cliff, D., 1997. Minimal-Intelligence Agents for Bargaining Behaviours in Market-Based Environments. HP Labs Tech. Rep. HPL-97-91.  
   
3. Cliff, D., 2019. Exhaustive Testing of Trader-Agents in Realistically Dynamic CDA Markets. In Proceedings ICAART-2019.
   
4. Cliff, D., 2019. Simulation-Based Evaluation of Automated Trading Strategies: A Manifesto for
   Modern Methods. Proc. EMSS-2019. 
   
5. Das, R., Hanson, J., Kephart, J., Tesauro, G., 2001. Agent-Human Interactions in the CDA. Proc. IJCAI2001, pp.1169-1176. 
   
6. De Luca, M., Cliff, D., 2011a. Agent-Human Interactions in the CDA, Redux. Proceedings ICAART-2011.  
   
7. De Luca, M., Cliff, D., 2011b. Human-Agent Auction Interactions: Adaptive-Aggressive Agents
   Dominate. Proceedings IJCAI-2011.
   
8. De Luca, M., 2015. Adaptive Algorithmic Trading Systems. PhD Thesis, University of Bristol, UK.
   
9. Gjerstad, S., Dickhaut, J., 1997. Price Formation in Continuous Double Auctions. Games & Economic Behavior. 22(1):1-29.
   
10. Gode, D., Sunder, S., 1993. Allocative Efficiency of Markets with Zero-Intelligence Traders. JPE,
    101(1):119-137.
    
11. Rollins, M. 2020. Threaded BSE: Experiments with Parallel Asynchronous Algorithmic Trading Systems. Master's Thesis, University of Bristol.
    
12. Rust, J., Miller, J., Palmer, R., 1992. Behavior of Trading Automata in a CDA Market. In Friedman, D., Rust, J. (eds) The Double Auction Market: Theories and Evidence. Addison-Wesley, pp.155-198.
    
13. Smith, V., 1962. An Experimental Study of Competitive Market Behavior. Journal of Political Economy 70(2):111-137.  
    
14. Snashall, D., 2019. An Exhaustive Comparison of Algorithmic Trading Strategies: AA Does Not Dominate. Master's Thesis, Uni. of Bristol.
    
15. Snashall, D., Cliff, D., 2019. Adaptive-Aggressive Traders Don't Dominate. In van den Herik, J., Rocha, A., Steels, L., (eds) Agents & Artificial Intelligence: Papers from ICAART 2019. Springer.
    
16. Sommerville, I., Cliff, D., et al. 2012. Large Scale IT Systems. CACM, 55(7).
    
17. Tesauro, G., Das, R., 2001. High-performance Bidding Agents for the CDA. Proc. 3rd ACM Conf. on ECommerce, pp.206-209.
    
18. Tesauro, G., Bredin, J., 2002. Sequential Strategic Bidding in Auctions using Dynamic Programming. Proceedings AAMAS2002.
    
19. Vach, D., 2015. Comparison of Double Auction Bidding Strategies for Trading Agents. MSc Thesis, Charles University in Prague.
    
20. Vytelingum, P., 2006. The Structure and Behaviour of the Continuous Double Auction. PhD Thesis, University of Southampton.
    
21. Vytelingum, P., Cliff, D. Jennings, N., 2008. Strategic Bidding in Continuous Double Auctions. Artificial Intelligence, 172(14):1700-1729.