A new generation of computer trading models is emerging as banks come under pressure to prove they are offering clients the best price on currency trades in an ever more competitive market.
Whereas investors might once have picked up the phone to ask brokers or traders for a quote, they are increasingly using complex models designed by the banks, known as algorithms, or algos, to transact large orders at the best possible price.
Pressure from regulators, who are keen to ensure a fair deal for investors in the $US4.5 trillion dollar-a-day market, is adding impetus to the trend.
"We're moving away from the old opaque world of FX trading," said Pete Eggleston, head of Quantitative Solutions at Morgan Stanley.
"Regulatory change and responsibility to investors are leading us into this world of best execution and transparency. Algos are part of the solution in helping clients adjust to these changes."
With a slew of new regulations coming into force on both sides of the Atlantic, the asset managers who deal with currency trading desks must ensure "best execution" for their clients and prove, rather than simply promise, they are keeping transaction costs down.
Spot currency trading will be exempt from reporting requirements outlined in the Dodd-Frank Act, a regulatory overhaul of financial markets passed by the US government.
But James Dalton, director of FX algorithmic execution at Citi, said some asset managers were already applying those rules to their currency transactions as the trend towards transparency was likely to spread.
"I am seeing for the first time when I talk to clients that questions are starting to come from the underlying investor to asset managers on how they measure transaction costs in FX," Dalton said.
"A lot of the main banks already have algo execution products or are about to launch. It's a must-have in the portfolio of execution tools."
Lawsuits involving US custodial banks BNY Mellon and State Street have prompted investors to look more carefully at whether they are getting good prices from their banks.
A report by consulting firm Greenwich Associates found a growing number of the world's largest institutional investors are analysing - and cutting down on - the costs of trading currencies.
Algorithms can provide electronic records of exactly how and when a trade is transacted, allowing investors to check the price at which a bank bought or sold a currency on their behalf against the price quoted in the market. Some banks offer a time-stamp to the millisecond for each part of an order.
As this activity, known as algo execution, becomes more popular, strategists are developing increasingly sophisticated models that can detect and adapt to sudden market swings, in much the same way as a human trader would.
The earliest models were inherited from equity markets, and simply broke up large orders into smaller amounts that were traded at fixed intervals, irrespective of market rallies or routs.
The next step was to equip models with different strategies for different market conditions. Passive strategies helped hide the order flow, while aggressive strategies were used at times of deep liquidity to move a large order quickly.
But an investor or trader still had to decide which strategy was suitable for particular market conditions, and the algo would plough on regardless of a turnaround in the market.
Now the more adaptive algos, like BNP Paribas' Cortex iX, can process feedback from trading activity and decide which execution rules are best suited to market conditions, helping capture the best price for clients.
"Think of second generation algos as blind and deaf ... With third generation algos we have given them the power of sight and hearing," said Asif Razaq, global head of FX algo execution at BNP Paribas.
Another feature of adaptive technology is that the execution of the order appears to be random, helping avoid detection from other algos in the marketplace which are on the lookout for trading patterns that they can exploit.
Protecting the herd
So far it has been the larger players in the currency market, such as Citi, Credit Suisse and BNP Paribas, which have forged ahead with the costly task of building these more sophisticated execution algos.
Some banks have even recruited people with doctorates in physics and mathematics from the world of academia to help construct models that will win clients and protect their currency trading profits.
Investment banks tend to lump currency trading in with bond and commodities trading. Overall, these units accounted for more than half of revenues at the top 10 investment banks in the first half of 2012, according to analytics group Coalition.
But the contribution that currency trading made to this overall trading revenue stream dipped to eight per cent from nine per cent in the same period a year earlier, Coalition said.
Unlike the rapid-fire models used by high-frequency trading firms (HFTs) to pounce on trading opportunities, execution algos are designed to ease large orders into the market without alerting other players to the flow, thus ensuring a good price.
By limiting price "slippage" - the difference between the price at which a market player wants to execute an order and the price at which they are able to do so - they are especially useful for what Aite Group analyst Javier Paz defines as "slowing-moving buy-side clients" at risk from predatory HFTs.
"The banks have to tweak their own execution algorithms that they provide to slow-moving clients so they can have that invisibility cloak. It's like protecting the herd from the velociraptors," Paz said.