Simulation Studies of Automated Trading Algorithms for Financial Exchanges Operating Frequent Batch Auctions

In recent years major financial exchanges have introduced Frequent Batch Auctions (FBAs) as a novel automated auction mechanism for matching buyers and sellers of various types of asset, in contrast to the traditional Continuous Double Auction (CDA) that has been the basis of such exchanges since the 18th Century. FBAs have been proposed to mitigate against the ill-effects of High-Frequency Trading (HFT) systems which trade at super-human speeds. In this paper we report on simulation studies of automated trading in FBA-based financial markets: we have extended a long-established open-source simulation model of a CDA market to also allow FBA-based trading; after that, we adapted existing automated trading algorithms initially designed for CDA-based markets so that they could work usefully in the FBA-based exchange. As far as we know, this is the first paper to be published on the extension of these automated trading algorithms to operate on a FBA-based exchange. By conducting more than 1.7M simulation experiments, we examine the pairwise dominance relationships of six well-known trading algorithms, evaluating each A/B pair of algorithms across a range of different ratios of A:B. Our research hypothesis was that the profitability of the minimally-simple SHVR trading algorithm, ‘a tongue-in-cheek model of contemporary high-frequency trading’, would decline significantly due to FBAs being designed to curb the advantage of HFTs. The results of our simulation studies reveal that, surprisingly, a minor modification of SHVR was able to maintain profitability in short batch intervals. In fact, we show here that in FBAs SHVR dominates both AA and GDX, two well-known trading algorithms that have previously been proven to out-perform human traders. Further to this, we show here that the dominance hierarchy of CDA trading algorithms first established in a paper published at EMSS2020 was disrupted by the switch to FBAs: surprisingly, the algorithm GVWY rose to the top of the hierarchy, demonstrating an unexpected effectiveness in the FBA. Python source-code for the simulation experiments reported here is being made available on GitHub, for other researchers to use.