INGs bond-trading bot does better than many humans
The software has lifted the number of times the bank has the second-best price on a bond in the market by 20%, while reducing mispricing — and thus trading costs — also by about 20%
London — An artificial intelligence (AI) platform developed by ING has helped the firm improve its bond trading performance, and the company now plans wider use of the technology, one of the Dutch bank’s senior trading executives says.
ING is also releasing a customer-facing version of the technology to help clients trade bonds, Santiago Braje, the bank’s global head of credit trading, said at an AI conference in Amsterdam on Wednesday.
In its first full year of deployment, the software — which ING calls Katana Edge — has allowed traders to make pricing decisions faster, improving the number of market-making trades the firm wins by 20%, Braje said.
The software has also increased the number of times the bank has the second-best price on a bond in the market by a similar percentage, while reducing mispricing — and thus trading costs — by again, about 20%, he said.
ING initially used the software with traders specializing in thinly-traded emerging-market bonds. The firm now plans to roll it out to other areas of its credit-trading practice, Braje said in an interview on the sidelines of World Summit AI.
The bank announced on Wednesday that it would begin offering a client-facing version of the software, called Katana Lens, that can help them decide which bonds to buy or, if they have a bond they want to sell, how best to price it. This system is being co-developed with Dutch pension fund PGGM, which is an ING customer.
Katana lens finds pairs of bonds whose prices have historically moved together and then looks for arbitrage opportunities during periods when the pairs’ movements get out of sync. The software can forecast how much the two bonds’ price movements will diverge and for how long, Androniki Menelaou, ING’s data science lead at its wholesale banking division, said.
Braje said that ING viewed Katana as a tool to help traders make better decisions — not as something that can replace them. For instance, while the system can spot bond pairs and possible arbitrage opportunities that a human might miss, it cannot understand if there are fundamental reasons for a price divergence between bond pairs — for instance, if one company is struggling financially or is based in an emerging market, such as Turkey, where political risk has deterred investors.
In these cases, the bond’s price might not come back into alignment with its pair, and it takes a human to understand those underlying factors and determine if the trade makes sense, he said.
’We view this as augmenting human intelligence,’ he said.
In the first six months of testing, PGGM traders using Katana Lens had made profits, even though the overall bond market was loss-making, he said.