A Closer Look At: The Launch of a New Algo

Mark Goodman, head of electronic execution, E&C, at UBS, talks about the latest addition to the bank’s algo suite.


Profit & Loss: UBS recently launched a new algo, called ORCA-Direct. Can you tell us a bit more about the design and logic behind it?

Mark Goodman: ORCA-Direct came about when we were trying to improve the technology available to our FX voice traders as they execute risk in the market.

As a general rule, the more liquidity they can access the better, so we made sure that we’re connected to the core ECNs and now have six on our platform. The other thing we started doing was looking at how we can use futures liquidity in correlated instruments to improve the ability of these traders to capture more liquidity when they need it.

So we’ve built technology that brings in the spot ECNs, our own FX principal liquidity and the liquidity from the CME FX futures. By using the futures as a correlated instrument to the spot, the traders are able to capture in the region of 15-20% more liquidity, which essentially means that they can get more done at a faster rate.

P&L: But now this is being rolled out to clients?

MG: Yes. So ORCA-Direct has been live internally since July 2017, we’ve been building up the data set around it, tweaking and improving it, optimising the performance, we began offering it to clients this year and have been fully rolling it out this week.

We don’t think that it makes sense to build one set of technology for your traders and another set for your clients. Yes, there are some different requirements, but really you’re trying to solve the same problems.

P&L: What are the other features of ORCA-Direct?

MG: Well as soon as you start going through fragmented liquidity pools, you get the problem of “ghost liquidity”. If I decide that the best price is on EBS and the next best is on Hotspot, when I hit the price on the former the price on the latter might then disappear, meaning that my performance is decaying because I have to go back to the market again.

We’ve solved for that by how we’ve built out our infrastructure and how we co- locate our servers. We’re trying to make sure that our hit rates are as high as possible and they’ve actually gone from about 50-55% to around 80%, which is a significant increase in capture. And this is massively important because when you go to the market to take liquidity you’re providing a signal. Now if you get the liquidity that you need then you’re happy to have some signaling cost, but if you don’t get all that you need then you have to go back to the market again and you’ll find that the performance of that trade has decayed against you.

P&L: That is a significant increase in the hit ratio, is that just from infrastructure and co-location optimisation?

MG: It’s partly from that and partly from the third problem that we’re trying to solve, which is the classic issue of last look versus firm liquidity. In FX, if you decide that you don’t want to give out any signals then you will probably want to avoid last look liquidity and only trade on firm prices. This increases your certainty of execution but there’s probably some useful liquidity that you’re excluding by using this approach. Conversely, if you don’t distinguish between firm liquidity and last look liquidity then you’re probably going to leave signals in the market and not get all of your liquidity in return, causing performance decay.

So what we’ve done with ORCA-Direct is looked at how to mesure the value of last look liquidity. This is where a lot of the innovation has been, we’ve done a lot of quant-based research based on machine learning techniques to understand how to normalise the last look prices versus the firm prices and work out which one to go to first.

So we’re essentially creating a synthetic order book where we’re adjusting last look prices to be able to measure them on the same basis as firm prices specifically for the order at hand, and that helps with the hit rate as well because we’re going for last look prices where our data is telling us we have a higher than average chance of getting filled.

P&L: In what type of scenario do you think clients should consider deploying the Orca-Direct?

MG: It’s essentially an algo for clients wanting to get a large amount of liquidity done with a fairly high urgency. If they’re trying to spread an order out over time, maybe they don’t have an alpha view and want to follow the market, they might need a more classic-style algo like TWAP, or one of our more intelligent third generation-type algos such as UBS TAP.

What we’re doing is using ORCA-Direct to underline how we access the market. So you can think of ORCA-Direct as handling the micro child order level and these other algos as handling the macro decision-making.

P&L: In general, do you see clients generally preferring to use simple TWAP or VWAP algos?

MG: I don’t come across many clients that only want to use one type of algo and, having watched the evolution of algo trading in equities, I think we’re seeing the same evolution in FX but on steroids. So many of the FX clients that we speak to are going straight to the more sophisticated level of algo trading.

Galen Stops

Share This

Share on facebook
Facebook
Share on google
Google+
Share on twitter
Twitter
Share on linkedin
LinkedIn
Share on reddit
Reddit

Related Posts in