The Gateway to Algorithmic and Automated Trading

Every trader needs an algorithm

Published in Automated Trader Magazine Issue 03 October 2006

New breeds of aggressive algos are being designed for active DMA traders, whilst changes in market structure are making algos essential.

Richard Balarkus

Richard Balarkus, Managing Director and Head of AES sales at Credit Suisse

Credit Suisse's AES team is currently focused on the needs of the more active trader. Although the use of algorithms has exploded on both the buy side and sell side, Credit Suisse's AES team was still discovering many experienced traders who reject the use of algos. Chris Marsh, who heads the AES desk in London, said, "Many brokers were pushing algorithmic trading products that were little more than glorified trade scheduling and VWAP tactics. This may suit some but certainly not all styles of trading. In the real world, many traders need to be much more aggressive. They want to get done quickly and participate in available liquidity -- as long as they don't push or chase prices."

When the AES team analysed the execution performance of clients using their algos, it found many traders being too passive. For example, the majority of VWAP orders were set to run until the market close, and even strategies focused on reducing impact and
implementation shortfall were being used too passively. "The first algos were very successful at not paying the spread when compared to a human trader, but often it's not spread costs that matter, but trend and time risk instead. If a trader thinks a stock could move 70bp in the first hour of trading and the spread is 20bp, then if volume is present on the far touch it makes sense to lift a chunk if not all of the far touch and pay the spread," says Marsh.

At the same time Credit Suisse was trying to determine how to use algos in less liquid names where volume patterns are unpredictable and price correlations between names are weak. "We identified that simply joining the bid in a small amount would make the spread gap and offers disappear in illiquid names. We needed to avoid signaling risk. We also needed to make sure that when liquidity opportunities arose in these names, we got involved", says Richard Balarkas, Managing Director and Head of AES sales at Credit Suisse.

What emerged was a number of tactics that have taken the algo market by storm, and which started the trend for brokers to give their strategies combative names, perhaps the best known of which is AES's Guerrilla. "This strategy just waits in the bushes and when it sees a volume shock at your price levels it comes out and cleans up - then goes back and hides," says Marsh.

Originally devised as the answer to the small cap algo problem, Guerrilla has quickly become the algo of choice for traders in liquid names who want to complete trades quickly, at prevailing price levels, without fear of showing their size or chasing prices. "This is the algo for traders who say they don't like algos. It's the perfect algo for a DMA trader who doesn't have time to play the spread with small clips. "What really excites clients", according to Marsh, "is our ability to combine Guerrilla with other strategies - so we can gain real advantage from anticipating the trading patterns that will follow as a result of Guerrilla wiping out whole price points on the book."

In the US, however, Guerrilla and its "band of brother" strategies like AES Sniper have already been completely transformed in order to take advantage of the liquidity fragmentation that has arisen in recent months. "The NYSE and NASDAQ have been losing market share faster than anyone expected. There are upwards of 20 venues where you could get filled, and some of these have dark liquidity where the best prices are not even displayed in the book," says Dan Mathisson, Managing Director and Head of AES at Credit Suisse. Combined with new trade-through rules that require firms to fill top of book orders when they print a block, these changes have required the whole AES trading platform to undergo enormous changes in the last 12 months.

"That there is no longer a single dominant venue for trading presented many challenges to the existing operation," explains Mathisson, "like pulling data from multiple tapes simply in order to build a consolidated volume curve." Mathisson's team didn't confine themselves to the obvious approaches. "We had already built Pathfinder, which selects the best exchange to route an order based on factors such as price depth and latency. Pathfinder has now been trained to sniff out liquidity that has been hidden by others, and places our liquidity where we can use it to best advantage."

"Pathfinder has now been trained to sniff out liquidity that has been hidden by others, and places our liquidity where we can use it to best advantage." Richard Balarkas

A striking example of how radical AES order placement has become is the way in which Pathfinder seeks to place a limit order on a market where it least likely to be seen. This may seem counter intuitive, but Credit Suisse's AES team know that being top of book means block trades cannot trade through your limit, and being top of book on a less visible venue means you are less likely to be spotted and "pennied" (a scenario in which another trader's small order is placed just ahead of yours). This means traders cannot intelligently place a simple market order in the US without using an algorithm.

"We have always sold AES purely on the basis of its execution capabilities, not as a loss leader for other businesses, nor as a front end with lots of gimmicks to try and lock in clients," says Balarkas. "The additional complexity being introduced in the markets means we can differentiate by offering added value compared to other brokers - even pure and simple DMA is no longer commoditised. The good news is that all the hard work we have done in the US puts is in an excellent position to anticipate changes that might arise in European market structure post-MIFID."

  • Copyright © Automated Trader Ltd 2018 - Strategies | Compliance | Technology

click here to return to the top of the page
content