AC: How do you and your team use FX algos?
JR: We’ve always seen ourselves as a pioneer or a continual leader in the use of electronic trading and algorithmic execution. So, just to set the stage, we were very early adopters of electronic trading for the futures markets.We actually created our own in-house algos for the futures markets before the banks had their own algorithms. We were also very early adopters on the London Metal Exchange (LME) and were one of the first that we are aware of that started to use bank-supplied algorithms in that space. The LME is sometimes thought of as a traditional trading pit market only, but it does have an electronic market as well, LMEselect. There are only a handful of brokers that have leveraged their ability to tap into that electronic market and hook up their algorithms to it.
AC: When was that?
JR: We started exploring electronic trading in the futures exchanges in the early 2000s. We were writing our own futures algorithms and trading electronic DMA around 2003-2004. Then, LME followed futures. And as you are aware, the foreign exchange market has recently gone more electronic and the algo adoption on that marketplace has more or less exploded over the last 18 months or so.
AC: How have you been using FX algos?
JR: We are exclusively focused on optimising trade execution. Our research team does extensive work on the alpha generation models that produce the orders. The trading desk is tasked with getting the best execution.
Effectively when those research models produce orders, each order comes to the desk with a designated benchmark that we’re trying to target. So, for example, on a slower trading trend-following system we might target the time weighted average price. For a faster, more intraday-type model, we might target arrival price. Those are pretty standard benchmarks.
We have a team of experienced traders with deep market knowledge and they get the order, know the benchmark and select a bank-supplied algorithm that’s designed to meet it. And then, depending on current market conditions, the trader has the ability to tweak those underlying parameters to help customize the algorithm, whether it’s the trading window, speed, the urgency – any of those types of metrics and others – to target the best price for our investors.
Another thing that’s unique is we have our own proprietary, real-time transaction cost analysis (TCA) application.As we load up and are trading a variety of algorithms, the traders have the ability to monitor each and every order in real time against that stated benchmark and compare that benchmark to the performance of that order’s average price. The idea of post-trade TCA – end of day, end of week or end of month – has value to a degree. But what we really felt we needed was the ability for the trader to see how the algorithm was performing and then step in and intervene if the deviation to that benchmark, in terms of slippage, crossed a threshold that was too great.
AC: There seem to be two schools of thought as to whether to let algos do their job or to go for a more hybrid, human-algo approach.
JR: We fully embrace the spirit of algorithmic trading and it’s not that the traders are intervening on a frequent basis. We see it more in certain times of maybe lowered liquidity, where the algo isn’t programmed or isn’t expecting it, around holidays or those types of events, where it becomes apparent that the algorithm is not optimized for the current environment and that human interaction is warranted.
AC: What about with big breaking news? Is that the sort of situation when the human element comes into play?
JR: It would really be specific, on a case-by-case basis. Campbell is a long-standing systematic manager. To our core, everything we do is as systematic as possible and that includes trading. So it’s not our preference for our traders to jump in, because that goes against the grain of what we do. We do have a full Investment Committee so if, God forbid, there was a major natural disaster or a terrorism event and we needed to reduce the entire risk of the book or the portfolios, that would obviously have to go through a process of approval and then those orders would be generated and given to the trading desk to execute. That could be picking up the phone, that could be loading them into an algorithm, or that could be staging them for point and click trading. So at that point we would have the ability to protect our investors the best way possible. But in a regular, normal market environment, typically the orders are loaded into the algorithms and then monitored, with very little intervention.
AC: You said you have the means of observing in real time how you’re performing against your benchmarks. Can you talk about what that performance is like?
JR: We’ve seen as we have evolved a meaningful improvement in our overall slippage numbers. Part of that is our research group getting better as to when their models deliver orders to the trading desk for execution. That means looking at the more liquid times, the more optimal times to be in the marketplace for any given instrument. The other piece is the traders becoming more refined with how they’re using the electronic tools.
AC: You also said you’ve developed your own algos in the past. For execution, are those primarily from providers or is it a mix which includes your own?
JR: The execution algos are predominantly supplied by banks. We previously have done some of our own algo development but what we’ve found over time is that we would much rather have our internal resources focused on the alpha generation processes. So as banks became much more mature and the availability of algorithms became plentiful, we felt we could get best of breed execution algos by basically outsourcing to the banks. If you think about the big banks, they’re all competing against each other so they have large IT staff, they have pretty deep pockets in terms of resources to fund those initiatives. Given that competition amongst each other, they’re always trying to improve and deliver best of breed.
AC: Do you use a variety of different suppliers?
JR: It’s a mix – more than one or two but less than a dozen. We clearly want to partner up with what we call our strategic execution partners and be sure that we get adequate support, in terms of being responsive, modifying algorithms if needed, and providing day-to-day, on-the-desk support if there are line issues or technology issues that need to be troubleshot. There’s kind of a fine line with having too few partners and too many partners, so we’re always trying to work around that line.
AC: What do you think about the current offerings and what sort of developments are you looking for from providers?
JR: That probably takes us to the most interesting aspect of Campbell and electronic trading within the FX base currently, in that we have partnered with Pragma, which is one of the better known third-party independent equity algorithmic shops. We do business with them on the equities side, but knowing that we have deep experience in electronic trading and deep experience in foreign exchange, they approached us about a year and a half ago about developing custom FX algorithms. We’ve been live with them for going on six months and we’re doing some very interesting things, not only on the front-end algo development side but also on their ability to see into the microstructure of the FX market and provide us with some very interesting data. They are also able to provide some very powerful TCA.
AC: You mentioned that they are strong in equities. It seems that a lot of the providers have in the past taken their cues for FX algos from what’s already out there for equities.
JR: The interesting part is that a lot of the banks initially just took their equity offerings and effectively tried to port them over to the FX world.
What you find is the FX marketplace microstructure is unique and different from the equity world. So there are real opportunities if you’re building FX algos from the ground up, as opposed to just trying to take the proverbial equities square peg and stick it into the FX round hole.
AC: How do you see the partnership developing?
JR: It’s a very strong partnership for us, as a firm. Our research group has done several meetings with the quants at Pragma to get a better understanding of several topics, such as issues around trading costs. So not only is it a relationship that is specific to equities and FX, but it’s also a broader business relationship and we’re leveraging their knowledge in new and different ways.
AC: Is Campbell & Company open to new investors?
JR: Yes, and we have seen some fairly significant growth over the last two years.
AC: Which FX markets are you in?
JR: We’re in all the majors and we do some of the emerging markets, but it’s more what I call the “more developed” emerging market countries; we are not currently engaged in the more peripheral emerging markets.
AC: Finally, what sort of trends do you see in the marketplace these days?
JR: Clearly the divergent global central banks paths between where the ECB is right now and, say, the Bank of Japan, versus where the US Fed is, has created interest in several different pairs and it’s created volume and heightened volatility.
Anytime we get more players in a space, like the euro or the yen, it helps to decrease spreads and ultimately helps us with our trading costs. That’s probably the biggest thing we’ve seen – interest has really been high lately given that divergence of global central bank paths.
It’s been largely helpful to our strategies, but can also present challenges as we saw with the recent Swiss National Bank moves. Overall, we view divergent central bank policies as driving the types of sustained trends that provide potential opportunities for firms like ours.