Beyond identifying investment opportunities, AI is able to assist managers and banks in their trading strategies. Optimization and automatic execution are at the heart of the trading processes’ recent developments: Robotic Process Automation (RPA) with rebalancing automation, Blockchain for reliability and transparency of counterparts and reduce need for central counterparty, and Artificial Intelligence / Machine Learning (AI / ML). AI is useful to obtain more information from sparse historical models, or help identify non-linear relationships in order flow. ML can be used to create ‘trading robots’ that then teach themselves how to react to market changes.
Pre-trade analytics: for best price
Banks and asset managers are increasing their focus on pre-trade capture analytics. Key innovation trends identified in this area include:
- predictive analytics to model capital, balance sheet and tax impacts pre-trade
- real-time analysis of trades, position and margin across multiple asset classes
- streamlining request-for-quote and pre-trade negotiation processes historically performed through various channels
For example, machine learning could be used to identify groups of bonds that behave similarly to each other. By doing so, they can rely on many more data points, providing better estimates of price movements when the market is thin. The resulting tool groups bonds into broad, intuitively similar buckets and then collects the most comparable products together in each bucket, to score the liquidity of individual bonds.
TORA, provider of the industry’s most-advanced cloud-based order and execution management system (OEMS), launched an AI-driven pre-trade Transaction Cost Analysis (TCA) solution designed to help firms meet the best execution requirements imposed by MiFID II. TORA’s new TCA product moves beyond traditional TCA by using AI techniques to accurately estimate price slippage for trades before they enter the market.
Optimization of execution and minimization of liquidity risk: for the best timing
Optimization includes market impact analysis, especially the evaluation of the effect of a firm’s own trading on market prices. While this estimation is key to properly timing trades, market impact is hard to model with traditional tools, especially for less liquid securities, where data on comparable past trades are scarce.
AI can be used to help identify how the timing of trades can minimize market impact. Market impact models can describe as a starting point how previous trade affect an incremental one. The models attempt to avoid scheduling trades too closely together to avoid having a market impact greater than the sum of its parts. These models can be used to set out the best possible trading schedules for a range of scenarios and then tweak the schedule as the real trade progresses, using supervised learning techniques to make the short-term predictions determining those tweaks.
Traditional algorithms slice trades that are too large to be executed in a block into child orders. Execution performance will depend on the type of algorithm used, which will depend on market conditions, so algo switching picks the best suited algorithm at each point in time. The use of AI-driven adaptive algorithms allows the execution team to adapt to the impact of other market algorithms which themselves are acting based on the behavior of the original trading strategy. The approach is responsible for driving how orders interact with market liquidity.
While traditional algorithmic trading typically offers a level of predictability of outcome over a few minutes, AI-driven algorithms give predictions over a longer timescale and allow practitioners to apply algorithms to right timescale letting them understand when to execute rapidly and when to take longer, and when to speed up or slow down the overall execution program.
LOXM is the bank’s AI programme that executes client orders as fast as possible and at the best possible price by reading market conditions, on the basis of billions of past trades, both real and simulated.
The Hong-Kong based asset manager launched in 2016 a fund that independently identifies and executes all stock trades using multiple forms of AI. The automated system predicts price changes based on a host of data provided by multilingual social media, balance sheets and macroeconomics record.
Portware is a financial industry's leading developer of broker-neutral, automated trading software. It has been acquired by FactSet in 2015. Its flagship product, Portware Enterprise, is the first and only AI-driven thinking Execution Management System (EMS): a fully customizable, FIX-compliant trade management and execution system for portfolio, basket, single stock, automated, and algorithmic trading.
The added value of AI
For the most active systematic funds, as much as two-thirds of trades' profit are estimated to be lost due to market impact costs (Source: Day – 2017). Trading execution is thus a significant issue on the cost-reduction race.
CLSA, a Hong-Kong founded brokerage and investment group, estimates that the win-ratio of execution optimization is in excess of 80% compared with traditional algorithms. While identifying short-term price displacements is an important element when improving execution, its solution – ‘ADAPTIVE’ – also predicts future trends, volume, volatility and price.
The Head of electronic products at ITG estimates that his AI-tool “has been a game changer through improving execution by a very significant 2.3 basis points [in the U.S.]”.
Portware’s Alpha Pro AI-driven agent for trading, for its part, has delivered an estimated 133 bps in return on micro-cap stocks, 63 bps on small cap stocks and 17 bps on large cap stocks that it has handled from March 2009 to March 2017.
Digitalization of the trading execution process have also consequences on human resources. Thus, Goldman Sachs have replaced 600 traders at its US cash equities trading desk by 200 computer engineers over the last 16 years, leaving only two traders.
Some hurdles still have to be overcome
The need for transparency in the execution process is a major hurdle to AI adoption. To effectively understand and explain AI-based strategy behavior requires traders to reframe the problem and educate investors and portfolio managers.
Another challenge is to ‘educate’ AI algorithms to the consequences of their actions and not let them rediscover practices already tried and discarded by human traders. AI algorithms seek to fulfill their goals with maximum efficiency, and with no regard for consequences outside of their objectives. An AI algorithm set to a particular benchmark, say the opening price of a stock, might behave too aggressively in hitting that goal. Carefully considering the target and its second-order effects is crucial when employing discretionary AI in the execution process.
Author: Pascal Buisson - June 2018
