ING To Expand AI In Bond Trading
Dutch bank ING is going to expand the use of Katana, an artificial intelligence tool, across its fixed income trading division after finding that it led to significant performance improvements.
Santiago Braje, global head of credit trading at ING Wholesale Bank, said at a briefing today: “Katana was named after a samurai sword as we wanted fixed income traders to respond to requests for quotes more quickly and with sharper and more consistent prices.“
There has been an increase in electronic trading in fixed income which is expected to continue under MiFID II, the regulations that come into force in the European Union on 3 January 2018. MiFID II extends best execution requirements into fixed income for the first time, and introduces pre-trade and post-trade transparency.
Braje continued that clients are sending electronic RFQs more often and for smaller sizes. It is also easy for them to invite more dealers to participate so there is more competition.
“This also presents new possibilities from the data that is generated,” he added. “One year ago we had no code, we took six months to develop a platform and we have been testing for two to three months.”
Katana has been developed by the bank’s financial markets global credit trading team in London and the wholesale banking advanced analytics team.
Frank Derks, head of advanced analytics at ING, said at the briefing that the unit was launched three and half years ago and now has 35 people, most of whom do not have a banking background.
“We have an end-to-end capability so we can put solutions into production,” Derks added. “We work on our own data science projects but also co-create with relevant fintechs.”
Derks continued that the unit is expected to grow to 300 people as it expands, especially outside Europe. “The strategy is part of the €80m being invested across the enterprise to create the next generation digital bank,” he added.
Katana has been tested by the emerging markets credit desk in London. Brahe continued that the tool allowed faster pricing decisions for 90% of trades, a 25% reduction in trading cost and that traders were able to offer the best price four times more frequently.
Braje said: “We have found that the combination of human and artificial intelligence performed better than either on their own.”
As a result Katana will be rolled out over the rest of ING’s fixed income trading desks next year. Braje said this fits with ING’s growth strategy in fixed income of covering more countries. “We can do more with the same number of people,” he added.
In addition to providing traders with a visualization of relevant historic and real-time data, Katana’s algorithms provide forward looking predictions of the price that will win an RFQ within a certain confidence range. Future developments include being able to price within Katana.
“The potential to improve decisions is huge,” added Braje.
In addition ING is working with a large institutional investor to co-create a solution using Katana which will help the client decide which bonds to buy and sell, rather than being able to respond quickly to RFQs.
Consultancy Greenwich Associates said in a recent report, Technology Transforming a Vast Corporate Bond Market, that there has been an increase in autoquoting, where liquidity providers are increasingly responding to RFQs below a certain size and/or risk threshold using algorithms to calculate a price appropriate for that particular request, without any input from a trader. The algorithms price quotes based on the individual client, current market liquidity, the dealer’s current risk position as well as other factors.
“As investors themselves start to utilize algorithms to automatically select the right counterparty for a given trade based on RFQ responses, the bond market will have its first foray into computers trading with computers,” said the report. “With dealers and investors alike now focused on electronic trading and empowered with data and analytics, execution quality and speed will allow new entrants to gain share where never before possible.”
Tradeweb, the electronic fixed income and derivatives execution platform, said last month that clients and volumes for its Automated Intelligent Execution offering, which automates request for quotes, have increased five-fold over the past year ahead of MiFID II.
Charlie Campbell-Johnston, director, head of integration and workflow solutions at Tradeweb, told Markets Media last month: “The ‘codifying’ of execution criteria through automation can be a significant help for investment firms needing to demonstrate the process of best execution.”
Today Liquidnet, the global institutional trading network, announced the launch of Virtual High Touch for fixed income. Liquidnet said in a statement: “It evaluates order characteristics, market data, liquidity conditions, and user preferences to suggest an optimal execution strategy for different groups of orders.”
Technology has enhanced capabilities of surveilling larger and more disparate data sets.
With Eugene Kanevsky, James Redbourn, and Joanna Wong, CLSA
AI/ML on the buy-side trading desk is a long-term program with short- and mid-term deliverables.
Aiden uses the computational power of deep reinforcement learning to improve trading results and insights.
BondDroid’s AI-generated prices are integrated directly into LTX’s pre-trade analytical tools.