Interview with Peter Van Kleef, Lakeview Arbitrage International

In the first of our interviews featuring keynote speakers at the upcoming High Frequency Trading World events in London and New York, Mike O’Hara from the High Frequency Trading Review speaks to Peter Van Kleef, CEO of Lakeview Arbitrage International.

Prior to his role at Lakeview, Mr Van Kleef managed significant hedge fund type investment portfolios and quantitative trading departments for among others Cooper Neff, Salomon Brothers, HypoVereinsbank and Credit Lyonnais. He has over ten years of experience in the development and running of sophisticated automated trading operations. He holds a MBA degree from the Owen Graduate School at Vanderbilt University, Nashville, USA. He is a frequent speaker on complex arbitrage strategies with a focus on volatility arbitrage and high frequency algorithmic trading. He is also a well known consultant to the investment community with regards to trading, risk management, operational and strategic issues. Lakeview provides a large range of trading and risk management related services.

High Frequency Trading Review: I’d like to start with the same question I ask everybody, because every time I ask it I get a different answer! What’s your definition of High Frequency Trading?

Peter Van Kleef: Basically, any trading that’s faster than humanly possible. Any kind of trading that a human can’t execute anymore, that’s what I would call high frequency trading.

Then I would make a distinction between high frequency trading and ultra-high frequency trading. High frequency trading would be something you can do in software, whereas ultra-high frequency trading is something you can try to do in software, but you are probably better off doing it in hardware.

HFTR: Is that because you’re going down to microseconds of latency with ultra high frequency trading?

Peter:   Exactly. But if you look at the reaction span of a human being, anything that’s beyond that I think would be my definition of high frequency.

HFTR: Would you include where the machine is actually making the trading decisions, as well as submitting the orders into the market?

Peter:   I think in general the term applies to where you need a machine to either calculate something beforehand and make a decision or a machine to execute.  I would include as well where a machine is just calculating stuff and making a decision while execution potentially  still gets executed via some other means or some other machine.

HFTR: What do you think is the impact of high frequency trading on the different market segments, in terms of buy side (asset managers, hedge funds, prop traders and so on) and sell side?

Peter:   I would say the leaders in our market clearly are the small prop trading firms because they drive innovation.  Usually you’ll see whatever is the latest and greatest technology in the small trading shops first. Then it goes to bigger sell side firms, then it goes to advanced hedge funds, to normal hedge funds, and maybe a couple of years later it gets to the mainstream in terms of normal asset managers.  So I think there’s a natural progression in terms of people driving the technology, usually due to small firms needing to innovate, finding an edge in the market and their being nimble and small, so they can implement new technologies much quicker.  If you have legacy infrastructure that cost $50 million, like a big investment bank might have, throwing all that out and buying all the latest and greatest kit is not going to be an option.  But if you’re a small firm with just 50 guys and some new kit that costs just a hundredth of the price, then you can easily buy that and implement it immediately.

What then happens, the bigger firms see that they are losing out in terms of business and profitability to the smaller firms. They don’t get the trades done anymore because someone else is faster and better in the market.  So they then do some research in terms of finding out who that is and what they do better, and try to catch up.  Then that usually filters through to the customers of the investment banks, which are the major cutting-edge hedge funds that get the insight on what the banks are doing now. Those hedge funds try to replicate it and it goes to bigger hedge funds, then finally it becomes common knowledge, and people start talking in conferences and it gets dispersed to the general public and the big asset managers.  So that’s the national progression I would see in technology.

HFTR: On a side note, you mentioned the general public, one of the things I’d like to get your thoughts on is the misinformation and misconceptions around high frequency trading on the part of the public and the media. Anybody who’s involved in this space can see that there seems to be a big backlash against high frequency and algorithmic trading.  What do you think is the reason for that?

Peter:   I think it’s part envy, part misunderstanding, part fear of new things. That’s dangerous in that most people don’t really spend enough effort and enough interest to actually find out what high frequency trading is and what it does.

You have to be realistic that computer-based trading increases liquidity and efficiency in the market by a huge magnitude, compared with for example the “old boy’s network”, you know, calling the broker over the phone and dealing at whatever price he would be able to get for you based on your relationship.  High frequency trading is what keeps prices in line across different venues, across different markets, different currencies, different securities and so on. That price adjustment happens reasonably quickly once news comes out, so there are tons of benefits to it.

Saying this is all bad just because some people don’t know how to implement it and then have some screw-ups in the process, that certainly is not fair and usually comes from the people that didn’t spend the time to get things right first time.  Of course anything that is of huge public impact, that can influence the lives of millions in terms of if the stock market drops 50% in 10 minutes, that destroys a huge amount of value and impacts everybody.  That is where the regulator has to step in and make sure that the market continuity and market performance continues under stress and that the necessary precautionary measures are taken. Preventive measures have to be taken at the exchange level, not only at the broker (gatekeeper) level. The markets have to have mechanisms and safety features in place that protect them from incompetent and neglignet members and participants. That’s something that has to be a clear process because usually the regulator only acts after the fact not before, because if he has no suspicions of something being about to happen, it’s pretty difficult to do pre-emptive regulation.

It’s a question of the regulators a) realizing the value of high frequency trading for keeping prices in line, providing liquidity day in and day out, b) focusing on weeding out the bad guys, who actually are few and far between because usually the bad guys get carried out of the market sooner or later anyway and c) forcing marketplaces to have effective protective measures in place to protect themselves and the functioning of the market. 

HFTR: In what way?

Peter:   Well, there’s probably people that made millions in the “flash crash” and those usually are the good guys (in a technical sense) and there are some guys that lost millions too and usually those are the guys that shouldn’t be supported.  I mean exchanges, cancelling trades because there were mis-trades? That’s basically supporting the bad guys in my opinion. Anybody that sells Proctor and Gamble at one cent deserves to be carried out of the market because he basically doesn’t know how to program a system properly or how to run an infrastructure properly. These sorts of things shouldn’t be protected. Finally, there should be no mis-trades any more in the 21^st century. With proper rules and safety features, I don’t see why there should be such a thing as a mis-trade. At least for “professional” market participants. So I think it’s an iterative process, seeing what the regulations will do and then going at it again.

The big danger of course is regulators trying to regulate something that they don’t understand. That’s an interesting challenge for the regulator, to work with people involved in the market, because the regulators need to get educated before they regulate.

HFTR: So you’d support of the correct use of circuit breakers rather than exchanges breaking trades after the fact?

Peter:   Having circuit breakers is a very simple but very effective measure, which is not a new feature.  Circuit breakers have been around in the market for decades, pretty much since the early days of electronic trading.  So it surprises me that this wasn’t, or isn’t standard in most of the markets already because such a trivial feature would prevent things dropping to zero and back again in a flash. The importance here is not to hold the market up for too long a period – which could create other risks – but to have it just kick in long enough to stop bad systems damaging the market and short enough not to disrupt the market.

The worst thing that an exchange can do is breaking a legitimate trade.  If they allow a trade in the first place they should never break it. Of course if this is a retail person, you can’t expect them to have the knowledge and the systems and the know-how, that’s a different story, you have to still protect the innocent. But a JP Morgan or a Brevan Howard or some other professional firm? That’s not an innocent third party.  If they screw up, tough luck but so be it. If they don’t know what they’re doing they shouldn’t be in the position, they should be carried out of the market immediately and rightfully so.  If you do a trade as a private person, as a non-professional in the market you should be protected, but on the other hand I don’t see why the other guy should be protected.  It’s not as if when Michael Schumacher runs out of gas that they say “okay let’s stop the race, let’s refill everybody and let’s give him a second chance”. If he didn’t plan his stops correctly then nobody cares.

HFTR: Good analogy. Moving on, I’d like to talk to you about staffing and recruitment. There are obviously various different roles evolving now within this space. How do you see the skill sets changing and how does a firm go about recruiting and keeping the top talent that they need to run a successful high frequency trading operation?

Peter:   Well, things have changed. It’s not like the old days any more, when traders had to be aggressive and adrenaline-driven. I think they are a dying breed because those are the danger points in high frequency and electronic trading and I see that some firms are still trying to recruit that kind of traditional style trader.

There are a couple of important things, the first is a mindset for risk but controlled risk. Someone who understands that they are basically playing with a loaded weapon.  These are people that usually would be happy to program missiles for the military, or who would program medical operation systems, stuff like that. People who realize that they’re doing something that can have a huge impact within a very short period of time and can potentially destroy massive amounts of value and impact a massive amount of people. So we look for people who show responsibility, that’s very important.  What you definitely don’t want is a traditional trader type that basically says okay if I go down the firm will go down with me and then I just go to the next shop and brag how big I was and get the next job. I think that breed is dying out.

In addition to that, you need traders that have some kind of IT skills. Not everybody needs to be a hard core programmer but they certainly need to be able to talk to hard core programmers, they need to be able to communicate with IT staff.  And ideally you want an IT guy that has ideas and some experiences in trading, OK, the IT guy doesn’t need to be an expert trader and the trader doesn’t need to be an expert in IT but they need to know from each other what is difficult and what is easy, and what the challenges are.  Also it’s worth looking at people who have an interest in the market but are working in other, unrelated industries.  So you might have people from biology, chemistry, electrical engineering, different fields that face similar problems.  For example in chemistry and biological engineering you might have to issue medicine to distinguish cancer cells from normal blood cells and that’s similar to distinguishing a market that goes up from a market that goes down or goes nowhere.  People who come from that kind of background may have good ideas and creative new input and that’s something that’s very valuable, on the entry level, that you have people from different subject fields coming in with fresh ideas and new views on things. In high frequency trading, it’s very important to be able to step back and look at the problem and see if there’s an alternative solution that’s much more efficient, much quicker than the old one rather than just pushing the old one.

HFTR: So how do you go about finding these people?

Peter:   Usually what is a great idea is having some kind of partnering with universities, where you might have a contest with a prize of say ten thousand pounds for the person who comes up with the best algo.  Then if someone comes up with an algo that really makes money and your firm earns millions out of it and it’s going to cost you ten thousand pounds, you’re going to have potentially fifty or a hundred guys developing strategies for you for free and you only have to pay for the thing that works.  I mean it’s a no-brainer. But there are very few firms doing things like that. Very few firms spend the time and effort to actually get in touch with educational institutions and try to establish some relationships there, and they’re a great source of people.  Another great source is having people come in for an internship and doing various trial things, so you can see how people think.  With entry-level people, for my part it’s much more important how they think and how quickly they can learn than what they know in the beginning.

Intuition is also important. For example, if I sell ten thousand of a leveraged contract then I potentially have a huge position and for a lot of people in this business tend to lose focus of the numbers.  So for them it’s like a hundred is the same as ten thousand, as a hundred thousand, or a million. Sometimes when they get results where a person from the outside would say “okay this thing is stupid I mean that can’t be the right number”, they say “well but my computer told me it’s the right number”. So they don’t develop intuitive understanding of what makes sense and what doesn’t make sense.

That natural intuition is often the most difficult skill to find.  You find a lot of people that are great in programming, people that are great in mathematics, you find a reasonable amount of people that have some risk appreciation. But people that have all that and then have an intuitive understanding for what the impact is of what they’re doing, what the numbers mean, are not so easy to find and that’s usually what the small firms are really good at, recruiting that talent.  So what may seem like ridiculous interview questions for entry-level people, like how long do you think the Nile is or how many people are there in North America, they’re there to see if candidates have a rough idea what numbers mean, what size means, what the impact is of certain things. That’s usually how you get good people at entry-level.

On the more experienced side, it’s often good to have people who have failed in something because ideally they’ll have learned a lesson from it.  I think that’s something you find out in an interview rather quickly, if somebody is going to make the same mistake all over again or if they actually learned something from their mistakes. Again that’s very tricky of course because it’s very hard if you hired someone that already screwed up, people will say, “why did you hire that guy, you knew he’d already screwed up”.  Well I would argue that somebody else paid for him to learn a lesson!  What also helps on the senior level is get people with family. That naturally forces people to accept responsibility and once oyu have kids you understand quickly that what you do is not only impacting you alone. 

HFTR: OK, but having the right staff is only a part of the equation. To stay ahead of competitors, any high frequency trading firm has to continuously improve its technology.  I’d like to get your thoughts on what kind of HFT technology can be bought “off the shelf” versus what has to be custom-built or adapted.  Can you take us through the “build versus buy” arguments?

Peter:   Sure.  Basically most of the components can be bought fairly readily these days, even very advanced stuff, which is a good thing because if you want to get up and running with something like FPGA programming then you need to send people on courses and get clued up on that and it’s probably going to be at least a 2 year project. Whereas you can hire firms that have been working with that technology for years, maybe for machine control or in telecoms for example, and so they have all the experience.  Basically you just need to introduce them to the field of finance and algo trading and they can do it more or less immediately.  There’s a lot of knowledge that you can buy out there but people are still hesitant to go out and buy stuff.  A lot of times also it’s due to the fact that either they don’t know where to get it or feel that it’s overpriced, or maybe they are only aware of the overpriced stuff and they don’t know the alternative.

In the old days all the trading firms developed their own in-house software and there was no other way because you really couldn’t buy anything.  These days you can probably buy what would fill 60, 70, 80% of the needs of even the most sophisticated trading operation. There may be certain elements like special analytics, special screens, how you want to look at data, of course the algorithms that you write, you can’t easily buy algorithms that make you millions for a hundred thousand bucks yet and I’m sure it will be a while before you can.

Having said that, everything is getting commoditized and that’s really a development that has happened over the last 3 or 5 years.  There’s a huge commoditization and industrialization of the business that nearly everything that you need you can buy, or if you can’t buy you can have a service provider who can do it for you, in a lot of cases much more efficiently than you can do it yourself. This is the really big change.  The real edge for most people these days doesn’t come in infrastructure but really comes in the smartness of the algo and that’s what I think most of the conferences and the talks and articles are neglecting in terms of the logic, the intelligence, the behavior of the strategy, the know-how of the strategy for example to switch from a normal market to a fast market, how does the behavior change, how does it react to different market environments?  That’s the real juice in terms of productivity and making money.  This is why people who have that understanding but don’t have the technology or the funds for the technology and don’t have the capital will be selling that stuff, because it’s valuable and it’s something that going forward will be an interesting market. You’re going to be able to get the technology and funding basically off the shelf.

HFTR: So the future will be a hybrid of buy and build?

Peter:   I think the best solution will be to go for some sort of open source or commercial package, where you get the basic needs filled and then you customize on the back of that. Building everything from scratch is crazy these days.  It doesn’t make sense for anybody, not even the most sophisticated cash rich organization because it’s just burning money. In any industry that has grown, people outsource the key components.  In car manufacturing they don’t do the seats, they don’t do the tyres, they don’t do the wheels, they even have somebody to do the engine and they outsource the design.  Even Ferrari doesn’t design their own cars, they go to Pininfarina!

HFTR: OK, so continuing on the technology theme but drilling down a little, you guys are doing some interesting work around FPGA and NVIDIA GPUaccelerators.  Can you tell us a little bit about what you’re doing in that space?

Peter:   Basically you have two technologies there that are used for slightly different things.  GPUs usually provide scalability in terms of calculations that you can do in parallel.  For example option pricing calculations, some implied calculations for correlation dividends, things like that.  Any calculations that you can parallelize can give you a huge advantage.  You gain massive savings in terms of electric power consumtion; computing power and heat emission.  Calculating risk scenarios that might take hours in some cases can be brought down to minutes or even fractional seconds, because in the complex risk analysis you can paralyze calculations.  So that’s one feature where GPUs are used.   But then GPUs, at the moment, you still don’t have perfect integration with the rest of the trading infrastructure so you have some overhead in terms of sending the data from the normal CPU, from the normal bus in your computer, to the GPU, and getting it back.  So there’s some latency involved with that. For the moment you can’t buy the chips alone and build your own system, but I am sure that will come. 

HFTR: And FPGA’s?

Peter:   FPGA basically can give you ultra fast response time, way below millisecond response time on the market. You would use those with pre-calculated values from the GPUs uploaded to the FPGA, pre-trade, and then when the market update comes, the FPGA card basically already has a memory of what the response will be and can send out that response immediately without having to go through the bus of the computer to the CPU to the storage to the memory and back, basically doing some calculation on the fly.  So you use one technology for one part and the other technology  for the other parts. You use FPGAs for feed handlers or for reacting to the market with electronic eyes and trading engines and stuff like that. And you would use GPUs to pre-calculate the values that maybe the FPGAs are loaded with.

HFTR: So do you see exchanges moving towards using FPGAs for matching engines, for example?

Peter:   Well for me it seems ridiculous that the exchanges are still rolling out new exchange systems in software.  All of them are competing and saying they want to be the fastest, so they should be serious about speed.  If they want to be serious about speed they would have to do it in hardware of course.  For the moment none of the exchanges (as far as I’m aware) is even planning that far.

HFTR: OK, final question. What do you see as the future of HFT?

Peter:   I think the key challenge going forward will be the integration of regulation, marketplaces and participants, in terms of getting everybody on one table to actually try to come out with a sensible structure for the whole thing. We have this unorganized, heavy hardcore development with “loaded weapons” everywhere, it’s a little bit like dropping parachutes with tons of loaded weapons somewhere in a developing country and letting people figure out what to do with them. You might say okay there’s natural progression that someone will be the leader and basically run the country but it’s not necessarily sure it’s the best way of doing things and I think that’s something that’s missing from a lot of discussion.

We need to look at the economic impact, the potential social impact of those things and find a balance there. The attention is really on technology but technology doesn’t give you the edge in the end.  Yes, technology is a necessary ingredient to be competitive but the edge really comes from the intelligence and smartness of people.  Being more aggressive, being smarter, being more creative in the algorithms and stuff like that.  I think that’s missing as well.

Another thing that few people realize is that on the algo trading side, this is actually the first time that intellectual property in trading gets captured somewhere. Basically if you make sure you know what is in the code and what actually gets executed you can now, for the first time, objectively judge if a strategy is working because you get access to the intellectual property.  It’s not like your trader leaves and nobody has any idea how he was making money, whether he was making money because he was skilled or basically he was just lucky.   With a human being you’re never able to tell if the guy is lucky or the guy actually has a strategy because human beings will not execute strategies consistently every time the same way.

With algo trading and high frequency trading, this is the first time really that you can figure out a if something is actually making money on a consistent basis and how to capture the intelligence and the know-how, because you have access to the program code and can follow what it’s doing.

HFTR: An interesting point to end on. Thank you Peter.