Pragma Uses AI To Improve Algo Performance
Artificial intelligence has measurably improved the performance of execution algorithms in US equities and could do the same in other asset classes according to Pragma, the independent algorithmic trading technology provider.
David Mechner, co-founder and chief executive of Pragma, told Markets Media that the firm decided to apply modern artificial intelligence tools to its algos in the spring of 2018. He explained that the firm began to test the neural network in late 2019 with orders being sent to the AI algos on a randomized basis.
Following a number of controlled trials with clients, the AI algos were found to improve performance by between 33% and 50%, including the volatile conditions of last year. The AI algos react to signals to improve micro-trading decisions, such as whether to be aggressive or passive, routing, sizing, pricing and timing of orders.
â€śAI provided a material benefit which exceeded our expectations,â€ť Mechner added. â€śThe principles are applicable to any market-driven asset class and we are beginning to see positive results in foreign exchange.â€ť
The algos cover a universe of 3,000 US stocks, some of which are illiquid. Mechner said: â€śWe have seen a 50% increase in flows to the AI algos.â€ť
Mechner continued that Pragma has a pipeline of enhancements to further improve performance
Last year a long-awaited rule from the US Securities and Exchange Commission came into effect with the aim of helping the buy side achieve better execution by getting more data and information about their trades. Rule 606(b)(3) requires a broker-dealer to disclose routing and execution data for not held orders over the previous six months if a customer asks for the information. The rule is likely to lead to the buy side using more analytics to improve execution as electronic trading becomes more sophisticated and the choice in how orders are executed continues to increase.
â€śRule 606 shows there is a growing focus on broker obligations and that trend will continue,â€ť added Mechner.
He explained that, as a result, the bar is getting higher to compete in the top tier of algo trading.
â€śGetting a material, measurable benefit in average shortfall requires a lot of ingenuity,â€ť said Mechner. â€śIt took us a year and a half of intense research and development before we had our first version in production.â€ť
Algos also have to adapt to changes in US equity market structure. Last year three new exchanges launched in the US and firms have filed for regulatory approval of new order types.
Curtis Pfeiffer, chief business officer at Pragma, said: â€śThe use of AI allows us to adapt more quickly to market structure changes.â€ť
Breaking data silos is key to deploying automation beyond 'nuisance' orders.
They can be used on quantum hardware expected to be available in 5 to 10 years.
Streaming blocks change the basis of matching and price discovery so institutions can find new liquidity.
Clients can fine tune their pricing function via APIs and exposed user-defined settings.
Orders executed away from public markets can have measurable implications on execution costs.