[ I N D E P T H | F U T U R E S A L G O T R A D I N G ]
modified client parameters and customisations . A classic example of customisation is where a client wants to follow a particular trading pattern but then switch urgency or strategy based on upon predefined triggers .
JP Morgan rebuilt its algo platform around five years ago to provide the buy-side with more choice about the parameters they can set on their side for algorithms , and there are further customisations that the bank ’ s electronic traders can configure on behalf of clients . Ward adds this has allowed his team to have a “ richer ” dialogue with clients and demand is clearly there .
“ There has always been demand for customised algos , even 10 years ago there was a lot of demand ,” he says . “ We just didn ’ t have a scalable way back then to adapt an algo to what a client really wanted . The reason for that is when a client wants something different , we needed developers to code that and then release it for implementation in the client ’ s platform , which takes a lot of time .”
Reinforcement learning The Aqua algorithm has been a particular area of focus for JP Morgan recently . It uses a technology referred to as reinforcement learning to create advanced signals on order routing and placement .
With reinforcement learning , which is a form of machine learning , the algorithm essentially learns from itself over time by looking back at previous signals that it has generated and evaluates performance . The signals will dictate whether the algo crosses the market or stays passive .
Reinforcement learning technology was first applied to a recently launched model of Aqua that is focused on navigating quarterly roll dates when futures contracts expire . It can be a high-
" The more challenges clients see in execution , the more opportunity there is for us to come in and help them , and the solution is increasingly the customised algorithm ."
Issue 67 // thetradenews . com // 61