Corvil » News Coverage

Traders Magazine: What do you make of the boom in high-frequency trading?

Donal Byrne: It does seem to have a new lease on life in the last couple of weeks. I’m not quite sure what’s going on. But it’s been all over the place.

TM: By some estimates, high-frequency trading represents upward of 70 percent of total equities volume. Does that sound right to you?

Byrne: Yes. I think that’s very credible. Certainly, on any given day and probably on average, orders relating to automated/high-frequency/arbitrage--it’s all getting pooled together--probably constitutes at least 70 percent of the order flow in this space.

TM: What do you make of this? Is this good for the markets?

Byrne: I think time will tell. That’s what people are trying to figure out. Certainly, it’s good for some set of folks who are successfully implementing high-frequency trading strategies. So, someone is getting some good out of it.

I think the way you have to look at this is that the trading industry and the trading environment have always leveraged technology. And technology has always been a differentiator in the world of trading, even with the first introduction of basic computers to do better front-office executions or better back-office clearing. People have always leveraged out technology. And people who have done that successfully have tended to gain market share and tended to increase profits and tended to grow their business. I don’t think we should be looking at high-frequency trading as something that is very different, or Machiavellian, or maybe hard to understand, where people suspect there is an ulterior motive at play here.

TM: Many people seem to be doing that, of late.

Byrne: I think, as a technologist, you look at it and say that: OK, people are leveraging fast computers and fast algorithms and observations to improve their trading strategies, and technology is playing a huge role in that. And there are numerous other examples within other industries where these types of technologies have been introduced and have created changes in the marketplace.

If you look at the whole role of Fedex, the nature of the business was on the basis that you could ship a package overnight, when everyone else took much longer to do so. And they figured out that there was an addressable market. That market didn’t grow for a time; it didn’t blow away every other carrier. But it created a business for itself, and that was on the basis of leveraging logistics and technology.
So, high-frequency trading is leveraging a technology; it’s leveraging an observation that there are price deltas in time. And it’s taking advantage of that.

TM: But what do you think of the quality of the liquidity high-frequency trading shops supply, and what they mean for liquidity, in general?

Byrne: The new breed of market makers--the high-frequency, the automated market makers--are now playing a critical role in most liquidity destinations, in that they do provide greater sources of liquidity. And they are playing a very, very important role in that regard. I think that most traders on all sides recognize that.

Now, on the other hand, they are managing to profit successfully from that. And it is a case that this is a new breed of liquidity player in the high-frequency space. But they are now an integral part of making the markets work on a day-to-day basis. And without them, there simply wouldn’t be as much liquidity out there.

TM: What are your opinions on co-location? Should everyone be co-locating at different market centers?

Byrne: We work with lots of clients who choose to co-locate. And we work with lots of clients who choose not to. We work with lots of clients who are determined to co-locate, who, after a while, decide not to, and vice-versa. Let me give you the one-minute tour on co-location.

The predominant benefit and the predominant reason people would co-locate is to reduce their latency on the assumption that latency is dominated by geographical distance. Now, we have quite often shown that that’s not always the dominant factor in latency. When it is the dominant factor, then co-location can reduce your latency, and therefore, is likely to improve your trading strategy--if it’s a high-frequency or latency-dependent trading strategy.

TM: Do you see a downside to it?

Byrne: In general, co-location is a good thing. However, it comes at a price. And the trade-off is that many people in the high-frequency game are now trading at two or multiple liquidity destinations, or execution venues. And so, therefore, they have to make a trade-off between: do I have to co-locate and pay the price that co-locating my end venues involves, or should I centralize. And there are cost/performance trade-offs there. And when you have regulations that require you to check best price on everything before you execute, it’s something where you have to really check the price coming from every single venue, which can become quite expensive to implement once you get into a large number of venues that you’re trading across.

So what we’re finding is, with recent regulation, a lot of folks now are starting to say: I’ll re-centralize my trading engines. I’ll put measurement appliances at co-location points, so I can measure latency precisely. And then I’ll optimize my strategy across N venues, as opposed to just one venue. When it was predominantly single-venue trading strategies, then co-location was making a huge amount of sense.

TM: Are there other issues to contend with?

Byrne: Getting back to my original point, many times when people co-locate they remove the geographical limitations. But they’re still left with latencies introduced by the fact that when you have higher levels of traffic in peak times, then things slow down.

So, that’s no different from on our roads, it takes you longer to get home at 5:00 p.m. than it does at 3:00 a.m., even though the distances are the same. And we find that those effects can quite often dominate on a dynamic basis. So, if everyone else wants to trade when you want to trade, just because you’re closer (through co-location), you’re still competing for a set amount of resources. So, trying to overcome that is an additional part of the challenge that co-location does not always overcome. In general, co-location is a part of the answer: sometimes it can pay off, and sometimes it won’t provide you with as much benefit as people think.

TM: Let’s talk about latency management, your bread-and-butter. Who are the main customers for your latency management products?

Byrne: I can describe them without mentioning the names. There are three dominant groups of hosts that we sell to. On the trading side today we’re finding that the sellside and the automated market makers and the high-end, high-frequency shops, the prop trading shops and the guys who have the single-page Web sites, they are the predominant consumers.

Downstream of that, what is interesting is that there is a different set of both competitive and collaborative dynamics that always follows this space--whether it’s latency or high-frequency trading, etc. Of course, the venues, themselves, end up competing for this huge volume of order flow. Latency performance, quality of execution, price, liquidity ... all of those things are huge factors in how those venues compete. So, what we’re finding now is there is a slower up-tick, relative to the high-frequency side on the execution venue, as it relates to being transparent.

But there’s a more rapid take-up of the basic technology internally, because there’s a huge demand for them to get their own house in order, firstly, so they can represent to the outside world that, not just that they’re fast, but that they’re fast and in control of the speeds--because a lot of what the traders are looking for is predictability, the consistency in the performance and execution that they’re going to receive.

TM: Can you explain that further?

Byrne: Once they have predictability, they can feed that into their algorithms and achieve a predictable outcome. What traders hate is any sort of non-predictive behavior. If they understand that one venue is slow and the other venue is fast, and that’s predictable, then they have an arbitrage opportunity. If they have no idea from trade to trade as to what the particular performance is going to be, then it’s pretty hard to assess and get some sort of quantitative or predictive result out.

Now, sandwiched in-between that is a bunch of providers. What we find is that when people deploy our technology--both at the trading end and the vending end--what often happens is, in-between, for both the market data providers and also the people offering the connectivity or hosting or whatnot, they also have to have a comparable set of tools and capabilities, such that, all three parties end up talking in the same language. So, if I call you up and you’re a market data provider to me, and I say: I think you’re slow today. You can look and say: Well, actually I am--or I am not. But we’re at least speaking the same language.

TM: What other trends are you seeing in the latency management space?

Byrne: One trend that’s out there today within trading is risk and transparency. Now, when you project that onto the high-frequency world, it takes on slightly different meanings. But they have, in essence, the same concerns. Trading parties wanted to have predictability and visibility of what the latency performance was going to be as they look to execute at a given venue.

Now, what’s interesting about this space right now is transparency can have multiple meanings or interpretations to folks, which include: Does that mean that every trader trading to a given venue gets to see exactly the same thing? That’s one interpretation of transparency.

But, of course, that would remove the competitive dynamic. And so folks would say: That’s bad. It removes arbitrage opportunities. But there’s a transparency along trading partners, which is: If I’m trading on your venue and you’re using a set of providers, I should have transparency across all of the people that you’re trading with.

TM: Does this transparency concern the traders involved?

Byrne: What we’re finding is that there’s a big demand for transparency against people within the individual trading loops. And then, of course, there’s a desire to make sure that the competitive advantage is not made apparent to people that you’re trading against. That’s part of the challenge here in leveraging technologies: You have to come up with technical answers, but also answers that fit the competitive and collaborative landscape that’s out there.

So, you have to provide a means by which I, you and our other partners get to see the exact same thing. But at the same time, you have to be able to protect ... if you’re a customer of mine as a venue, and you feel that you have a particularly competitive trading strategy, then it’s incumbent upon me not to dilute or divulge that to some other party that coming in to execute on my venue. And so, with transparency comes a reduction in risk, meaning: If I’m a trader, and I have absolute transparency and latency, then I’ve got a greater predictability that I will achieve my fill rate, and therefore I have a lower risk of having a poor execution.

So, that’s an angle relevant to the high-frequency world about what we would call both risk to being able to achieve a fill and then transparency, in terms of us being able to see exactly what’s happening outside our own four walls. This is something that is evolving, that isn’t there today, and which people are trying to resolve.

TM: As you probably have front row seats to this, you know that market venues have different levels of latency. What do you make of the unequal latencies at the exchanges?
Byrne: Our firsthand experience shows that some exchanges are “fast” and some exchanges are “slow.” Some “fast” exchanges are sometimes slow and some “slow” exchanges are not as slow as people think. So what we make of this is that there is a lot of uncertainty from traders on who is really fast, when are they fast, and who is slow.
But we must remember the high-frequency trading game is one of relative speed, not absolute speed. So, what traders really need to understand is the likelihood that they can trade faster at a given exchange relative to their competitors, i.e. other traders. The advantage that a fast exchange brings to a trader with the latest low latency technology is that the total speed differential is relatively larger. Therefore, the relative advantage is larger. Contrast this with a slow exchange where, irrespective of how fast a trader can react, the relative speed advantage is marginal for the trading strategy to succeed.