How artificial intelligence can make FX brokers cleverer

COPYING AND DISTRIBUTING ARE PROHIBITED WITHOUT PERMISSION OF THE PUBLISHER: CHUNT@EUROMONEY.COM

By:
Paul Golden
Published on:

The ability of AI to help retail FX brokers is quickly moving from the theoretical to the practical; the result should be better operational efficiency and better trader services.

AI_cover_780


Most investors will know the disposition effect well, even if many are powerless to fight it. It describes the tendency for holders of falling assets to hang on to them, while also encouraging the lucky owners of rising assets to sell out – often earlier than they should.

But new research suggests that artificial intelligence (AI) can make FX traders better than natural intelligence allows. By collecting large amounts of data and analyzing and mapping it in real time, AI is providing a heuristic function, helping traders to avoid repeating past mistakes.

To demonstrate how this can work, trading platform Capital.com analyzed all clients who opened more than 10 trades between May and August 2018. The firm found that, on average, unsuccessful traders held losing positions for 4.7 times longer than did profitable traders.

Ivan Gowan Capital.com, 160x186

Ivan Gowan,
Capital.com

Capital.com then used machine learning to suggest relevant educational material for those traders who showed the strongest signs of this disposition effect in their trading activity, improving their entry and exit points.

The firm also used AI to calculate what would happen if all its clients put stop-losses in place (at the moment only about 30% use them). The modelling of possible results showed that if the unsuccessful traders were to use stop-loss orders on just 2% of the capital, they would more than halve the time they spent suffering losing positions.

The same modelling also predicted that the number of clients that would experience a margin close-out would fall by 75%.

This is useful stuff to know, particularly given the kind of performance that Capital.com saw in the group that it was analyzing.

“Our research has shown that while the majority of trades prove to be profitable, traders are often losing more money on their losing trades than the total amount they make on their winning positions,” says Ivan Gowan, CEO of Capital.com.

In the past, data analysts might have been able to segment their user database to an extent and offer those groups somewhat personalized products or services. But machine learning enables this process to take place at a much more granular level and a much greater scale.

Not just fast

There is much focus on real-time analytics, but analyzing data correctly is just as important as interpreting it quickly. 

Market participants will often say: “If I had such and such piece of data, I would be able to make more money”, without knowing how to apply the data properly. This kind of thinking leads to analytics becoming flooded with pointless data that can, at worst, cause more harm than good.

From a broker’s perspective, knowing which clients impacted its profit and loss is crucial. Being able to drill into live pricing to analyse what is going wrong and fix it is thus an important step, explains MahiFX director of trading and analytics, Alexander Ridgers.

Alexander Ridgers MahiFX. 160x186

Alexander Ridgers,
MahiFX

“External LPs’ prices can go wrong at any time, so knowing which instrument of which LP is causing the problem is a great example of how real-time analytics can spot a problem and provide the answer,” he says. 

“Accurate client assessment is the most important thing an FX broker can do, so having the best data in real time is something they should be willing to invest heavily in.”

The greater a broker’s capacity to understand what is happening in the marketplace and the quicker they can process this information effectively, the bigger advantage they will gain over their competitors.

AI systems can be programmed to use vast amounts of historic data as a reference point while analysing what is happening in real time and reacting accordingly. This can allow a reading of market depth in terms of trade execution on the bid and offer, assessing movement of orders up and down the order book or gaining a better understanding of hidden orders.

Continually reading these factors allows brokers to shape their pricing algorithms by predicting what will happen from historic performance – and, more importantly, from recent performance, where they are dealing in minutes, seconds and even milliseconds, explains Paul Webb, CEO of ADS Securities London.

“A machine’s ability to read a scenario where liquidity is starting to reduce has significant value,” he says. It allows a broker “to protect itself from being overly exposed in terms of how much liquidity it is putting into the market at a time when a potentially volatile market event is about to occur”.