Algos are typically structured as
-(a) Execution strategies (TWAP, VWAP etc),
-(b) Proprietary (prop) trading strategies (Statistical Arbitrage, Mean Reversion, Pairs Trading)
-(c) HFT algos (ultra-low latency, FPGA based solutions, etc)
There has been an increase in ‘packaged strategies’ for DMA clients such as TWAP, VWAP to facilitate trade execution (ie more flow through ‘vanilla’ algos), and also an interest in more competitive prop trading strategies (seeking enhanced backtesting and strategies based on more proven, verifiable mathematical models). The focus on HFT strategies has largely dissipated due to expensive infrastructure, diminishing spreads and lower returns.
There has also been increased competition to provide more efficient trading platforms (Chi-X Australia, ATPs, lit/dark pools etc) and a move to provide common infrastructure to facilitate sell side competition. This includes tick as a service, co-location solutions, common algo testing and infrastructure.
In particular as the strategies themselves become more competitive, and regulations become more stringent, and punitive in that fines are widely being applied for market misconduct of algorithmic strategies, we are specifically seeing a move to much more robust use of:
– historical data models for backtesting (both for PnL model verification and strategic stress testing to ensure models fully comply under all circumstances).
– more dynamic simulation models to facilitate random market moves, shocks, and verification for strategy conflicts – primarily within a participant, but also between participants. The testing infrastructure sought should provide multiple trading venues, not just one market venue.
This is leading to more competitive strategies and more efficient venues that can provide more cost-effective trade execution as spreads narrow, but regulatory compliance increases.
The trading desk flow is becoming more automated as orders are directed to multiple venues, and there is now a strong requirement to ensure compliance on order flow across all venues to avoid financial penalties yet ensure competitive trading strategies. The increased compliance and surveillance focus is highly apparent.
What comes next
A full service testing infrastructure for algorithmic trading across multiple venues with multiple participants having access to run their tests both alone and concurrently with other participants. Historical focus in the past has been on participants to test their own strategies, but it is also clear that potential strategy conflicts between participants must also be fully tested before reaching production. This is only possible through the use of a shared testing infrastructure, but to be complete should also provide multiple market venues (e.g. HKEX, TSE, SGX, ASX etc) so that strategies can be fully tested, with the presence of other participants also concurrently testing on the same platform. The sensible way to do this is to provide a backtesting platform that offers (a) historical market replay, (b) a price adjusted market replay so that historical orders are price adjusted to be more realistic of the current testing environment, for example when strategies might interfere and shift the current bid/ask, and (c) fully simulated testing complete with time scheduled market shocks (eg to move a security or sector by a significant percent in a defined time period … eg 5 msecs or 5 minutes). By capturing all orders and trades, participants, the exchange and the regulator are able to analyse the specific behaviour of all participants’ strategies to verify a shift in order burst rates, increased spoofing or radical positions shifts in response to such market shifts. The improved quality of market behaviour is apparent as such testing can only improve the behaviour of all strategies in the production market.