There were strong words at the Equities Leaders Summit in Miami this week as a panel of buy and sell side debated how to best leverage data analysis to improve execution – arguing that ‘insights’ are all very well, but without a pivot to action, the term means very little.
According to Giuseppe Nuti, managing director at UBS, the million-dollar question is understanding how a parent order is doing. “Is it us moving the market, are we getting the liquidity we want?” The initial exploration is often centred around slippage, which of course is what the desk is solving for, but Nuti warned that this can be a very crude measure.
“When it comes to analytics, we’re looking at where the liquidity is coming from,” he explained. “It’s an interesting avenue of research – and it provides a base line for us to understand how the market is responding to our liquidity demands. The clearest sign of liquidity is actually around the relative value performance of the stock. When we look at it directly, it’s too noisy. But when we move to look at it in relative value terms, we get a very clear view of how the market is responding to our demands.”
That’s an insight that can help the trading desks to understand whether they should be speeding up, slowing down, how they should be behaving.
However, when it comes to looking at how you can systematically improve performance based on data, it can be a long process, warned Hitesh Mittal, founder and CEO at BestEx Research. “There is a lot of noise, but data is the only way to move forward.”
John Bright, head of global systematic trading at Fidelity Investments, highlighted the journey his team have been taking with regards to AB experimentation – pointing out that it has really evolved to “ABCD” experimentation on multiple levels. “First it’s data collection, under different market conditions and in different contexts. Then, what do you do with that data? The idea is to be able to constantly understand the interaction with the market. It’s constantly evolving. Automation is a huge key to be able to make that systematic improvement.”
Automation and AI are of course the big themes in the data space. However Marc Wyatt, head of global trading at T Rowe Price, notes that although more AI is being seen on the research side, it has not yet made it onto the trading floor – and there are good reasons for that.
“In terms of GenAI on the desk, I would strongly caution people. Make sure you know what’s going on, and what’s going in. The responsibility is on you, and when the regulator comes knocking, you’re going to want to know what’s in the box.”
What Wyatt wants to see is more market information. “When a block comes in, the trader wants to see all the info – what was the market environment the last time you traded it? We want to understand more specific market conditions for when the order decision hits that ladder, and make a routing decision based on that.”
The panel agreed that for 2024, normalisation is the name of the game.
“The noise to signal ratio is very high,” noted Nuti. “It shouldn’t be about data-driven insights, it should be data-driven actions. Understanding is great, but my job is to change things in order to drive optimal execution. A good example – recently we started an interesting project to understand where execution is happening in dark venues. We can actually work out quite precisely which venue a particular stock is trading on right now. Yes, it sounds like Christmas come early – find out where it is, send the flow, and boom. But it’s not magic. Maybe you’re on the same side, maybe you’re second in line, for example. As we observe the impact it has on our algos, statistically it’s zero. So it’s a great insight, but nothing actually happens. We need to be able to show actions. So we have moved wholesale towards managing those actions.”
Moving fast is a crucial element, as is independence and neutrality. Most people will only implement a new algo if they can prove there is a statistical benefit over the original one. But Mittal pointed out that sometimes it can take years of data to prove that – and you could be missing out on performance in the meantime. “Sometimes you have to look at the expected values,” he suggested. In addition:
“Over the last 20 years, when it comes to algo execution, they have been programmed to the bias of the trader. If a trader feels a certain way, that’s the way the algo gets coded. They need to become more data-driven. Every trader has biases. How do you use data to confirm or reject those biases? Be surgical about improving every aspect of trade execution.
Another trend is sell-side accountability.
“Buy-side traders are more informed, from a sell-side perspective, and they want to be more informed,” noted Mittal. “Gone are the days you could just go to a buy side firm and take some TCA analysis and some numbers and explain it away. There has to be more engagement with buy-side firms to really understand the rationale behind the trade.”
Wyatt agreed. “The future will be where the buy-side and the sell-side connect. We all talk about what we are doing in our own ecosystem, but I’m interested to see how that translates onto our broker platforms.”
Finally, recruitment is a key challenge within the data sphere – and hiring people from different industries is crucial to fill the gap and achieve the best performance, thinks Wyatt.
“Data scientists are agnostic, they work across industries – and they don’t think the way traders do. You need to leave your ego at the door. We have to let go of the idea of doing something because that’s the way we’ve always done it. It’s very helpful to let people who are not from the trading and investment space loose to figure out how to do things better. They are breaking legacies and workflows and I’m all for it – we’re all going to have to do less with more going forward, and this will help us do that.”
“The market isn’t stationary,” concluded Nuti. “We have to be cognizant of that, and adapt. What’s interesting for us is understanding which parts of the market are likely to change.”
©Markets Media Europe 2023
