Algorithmic trading: do the numbers add up?

An algorithm can be defined as a series of calculated steps strung together to produce a desired numeric goal. First identified by a Persian astronomer in 825 AD, today algorithms are increasingly seeping into financial services through their use as trading programs, useful for shifting huge quantities of shares at a time. Trading algorithm can de defined as any type of mathematical model used for making transaction decisions in the financial market.

According to Matt Simon of the Tabb Group, algorithms can be thought of more as a tool to replicate the human or manual steps that bring together counterparties in a more sophisticated or more electronic manner: “The algorithms improve the trading process, as they provide a way to get through large blocks of shares without trying to alter the share price in any significant manner.”

There are five basic strategies that most algorithms currently being offered by the vendors can be grouped into, and as Simon points out, each has their own pros and cons:

• Time or volume-based algorithms. Traditional volume weight averaged price - VWAP - or time weight average price - TWAP.
• Liquidity seeking algorithms. Where the algorithm is specifically looking for shares in the market and looking to group them into larger blocks for more institutionalized trades.
• Price-based algorithms.
• Customized algorithms. These are where the buy-side or sell-side adds their own sophistications in terms of making the algorithm adapt to their own style.
• Portfolio algorithm. Used to find the exposure between the names in their portfolio rather than just focusing on the share price or market impact.

Reaction or pro-action

Most of the commoditized algorithmic trading programs being used today are more reactive than predictive. According to Simon, what this means is that they are based on share prices or volume, looking at historical data. “Being able to take more standard data like a news article and make repetitive decisions based on that would be a proactive approach that would imply the ‘intelligence’ of an algorithm,” he says.
Although this might seem like it has great potential, there are still proving to be many challenges that people are facing while trying to deal with so-called artificial intelligence: “They can be intricate to develop and the consequences of implementing something that can ‘think’ in this way can seem frightening,” Simon explains. “As the technology increases in sophistication, it can soon overtake the knowledge of the person implementing it. These things develop increasingly fast, and so interpreting what you’re doing on your own end can become increasingly difficult.”

Evolution in the market

One of the most exciting factors about the financial services industry is the innovative strategies that develop in the sector. These always promote the evolution of the market. In terms of algorithmic trading, Simon suggests that many of the brokers and independent providers that base their core business around algorithms “will be there each step of the way to help their customers. They won’t want their customers to feel that they’re losing control, or that they have lost the ability to handle what their technology is doing.”
Ultimately, when sorting through large quantities of shares, any trader will be thankful for not having to sit at his desk and do it all day. Anything that saves him some time in this case, will be highly valued. “But he needs to feel that he has control of the algorithm,” explains Simon. “Being able to be more aggressive, less aggressive, faster, slower: having that kind of knowledge remains important.”

And following on from some reports of trading algorithms spotting each other in practice, there are some brokers like ITG offering anti-gaming logic. They have reverse engineering products available, which, according to Simon, means traders can avoid getting caught when the market moves simply because there’s liquidity available in the stock, as opposed to moving the stock just to profit from it because the algorithms are moving it.

Drivers

The drivers for algorithms initially came from a desire to cut costs and take manual work out for the traders. “There has been a pressure for cheaper commissions, and on top of that we’re seeing fragmented markets where - in the US - the average execution size for orders is going below 300 to 400 shares per trade,” confirms Simon. There is also the well documented, continual struggle on the sell side in the top investment banks to maintain trading revenues on the equity side of their business, quarter over quarter.

Simon explains that we are seeing the following factors which have led to development of algorithmic trading: an evolution of less commission, less trading revenues for them, more fragmentation over the markets, and an ROI that’s becoming harder to produce on the equity side of the business. He gives the example of a sales trader, paying one to two cents per share versus upwards of six to seven cents. As he points out, “that doesn’t make much sense for an institutional trader, trading large liquid blue chip stocks”.

There has also been a general shift in the buy-side attitude, becoming less relationship-oriented, more confined, less reliant on the sell-side to prove their value. “Some of the agency-only models validate that,” says Simon. “They can get the executions done faster, cheaper, more efficiently. You wonder, what’s the value in having a trader trade Microsoft for me?”

Finding a place for algorithmic trading

Last year Tabb Group spoke to over 80 hedge fund managers, and published a hedge fund report that Simon co-wrote. More recently they published an institutional equity trading report, where they spoke to 60 head traders and institutions. The reports found that both groups had similar attitudes and feelings towards algorithmic trading.

According to Simon, on the traditional side, Tabb found 80 percent of traditional institutional asset managers were using some sort of algorithm because “it is cheap and efficient”. For hedge funds the number was lower. “Hedge funds are a little more sophisticated in what they do, and therefore more cautious,” he explains. “Professionals that are more quantitative-based are willing to take a shot at using an algorithm, because they can prove their value. With institutional traders, it is easier to implement a new algorithm, or new algorithmic style than to bring in a new trading platform or new way of speaking to somebody.”
Simon notes that he has seen brokers buying up some of the independents (he gives the example of Liquid Net buying Militus Trading). As well as brokers seemingly trying to implement independents into their execution management systems, faster automated trade is evolving at the same time. As Simon explains, this gives the anonymity to the buy-side that they are looking for. “People are intrigued by the concept because everyone wants less work, and it gives them the flexibility to work on the harder and more complicated trades but still fulfill their obligations of trading.”

The future

Predicting the future of algorithmic trading is difficult. “Every time we reach a point in the market that I believe to be saturation point – that it can’t possibly get anymore automated or developed – programmers take it up to the next notch, and develop something better,” he says. There has been talk about algorithmic development in how algorithms approach different markets. “It’s a tough call, because it’s already having such a significant impact: it’s around 25 percent of equity trades depending on how you classify it. That’s a significant portion and could grow, but until some of the more e-liquid asset classes - like fixed income and derivatives - become more automated it’s difficult to access that liquidity. It will be interesting to see how that plays out as well.”