The Gateway to Algorithmic and Automated Trading

Local knowledge

Published in Automated Trader Magazine Issue 19 Q4 2010

Bernard Orenstein leads the ten-strong team behind the automated high-frequency eStats Revolution Fund, based in Sydney. They’ve been managing their own money for 31⁄2 years, and in 2009, they managed a return of 182.73%. The fund’s leveraged, the drawdowns would also get your attention, and from November, if you happen to have US$1 million to invest, they’d like to hear from you. Larry Levy went to speak Bernard Orenstein about risk management, volatility and the evolution of an interestingly idiosyncratic automated strategy.

Bernard Orenstein

Larry Levy: Tell us a little bit about your background.

Bernard Orenstein: My first real job after studying computer science at The University of Technology, Sydney was with BHP (now BHP Billiton Ltd). I started work in their research laboratory in Newcastle, north of Sydney, in 1988. My focus was on high-risk, high-reward applied IT research, mostly in minerals and petroleum exploration. One example was virtual reality.

This would be used to create visualisations using multi-sensory techniques that interpreted minerals and petroleum geo-scientific data. We used a multi-sensory approach because as humans we have limited bandwidth as to what we can take in. The research we did was going from 2D to 3D visual but also employing auditory and haptic feedback. So we would build fully immersive environments that would enable many human senses to facilitate the interpretation of these multi-dimensional data sets.

Larry Levy: Do you feel that your modelling experience was useful for trading later on?

Bernard Orenstein: Yes. Another thing that helped was a project to develop multi-agent systems for automatically interpreting 3D seismic data to help determine exploration drill locations.

I thought that it would be great to take some of the skills and techniques I had learnt at BHP and apply them to the financial markets. I had the vision of fully automated trading, and in my head was the idea of having a system that could make the trading decisions. Also, a few of us had started an investment club, and I had found that my discretionary trading was pretty poor. I didn't handle the emotional ups and downs particularly well.

I started at a boutique hedge fund in Sydney in '98, and within a few weeks became their main software guy. The firm was into automated futures, which was relatively high frequency for that time. It was automated for simulation, reconciliation and order generation. But the orders were emailed across to the broker. That was a bit fuzzy. We would receive fills and sometimes there were errors and quite a lot of slippage. The lesson, apart from occasionally getting bludgeoned in the futures market, was that manual execution was error-prone and expensive. Slippage was also really expensive.

When I left the hedge fund, three colleagues and I, (all co-incidentally ex-BHP), set up Object Trading, which is an independent software vendor in the exchange DMA infrastructure space. This was in late 2000. We wanted to take full programmatic control of orders from models through to the exchange and back again. After a few years we had developed connectivity to a sufficiently large number of exchanges, that by about 2003 my passion for trading was reignited and I thought I could get back into trading at a whole new level.

Larry Levy: Where did eStats come from?

Bernard Orenstein: Initially, it was more of a part-time research project. Then I got more and more involved and gradually started building up a team. As to the model, I first wanted to answer the slippage cost question. It quickly became clear that locals in the trading pits were the main beneficiaries of slippage. I set out to really understand how a local worked in the pits and represent some of their business rules inside a computer programme. That was the seed of eStats and it took a long time to manifest.

Larry Levy: Tell us about your performance over the last 3 and a half years.

Bernard Orenstein

Bernard Orenstein: In 2007, we decided we would take the move from research and testing to production trading where we actually intended to make money. The performance has been pretty strong over the last 3 years but also quite volatile. We have had 3 significant drawdowns over that time but we have managed to pull in about 85% per annum after all fees, including the performance fee of 49%. The biggest drawdown was roughly 38%. In 2009 we had a strong period and we did 185%, also after the performance fee of 49%. We only charge a performance fee on trading profits, and there is no management fee.

We have spent a lot of this year doing structuring work in terms of setting up typical hedge-fund company structures and getting financial services licensing and getting our track record audited. We've been getting our administrator onboard and that type of thing.

Larry Levy: Was money easy to raise?

Bernard Orenstein: The fund itself has not been launched yet so the capital is purely from us and what we have built up over the years.

Larry Levy: How was the economic crash for you?

Bernard Orenstein: It was tough for us when Lehman went down. At the time, we were just trading a handful of markets - the eMini S&P and Eurostoxx futures. It was our first experience of ultra high volatility. Our model is fundamentally liquidity-provision, which relies on a fair amount of mean reversion and we went into quite a deep drawdown. It was pretty tough and there was a question, because we were systematic, of what to do. Do we turn it off, scale back or what? We had a bit of a weakness at that time because we didn't have any kind of systematic scale-back regime. We had a risk committee and an investment committee which convened and on a discretionary basis we pulled out of the market for about a two week period. We then developed a rudimentary scale-back scheme and then re-entered the market.

Larry Levy: Is it a volatility-based scale-back model?

Bernard Orenstein: No, purely drawdown-based.

Larry Levy: Were you affected by the flash crash?

Bernard Orenstein: No, we weren't. Over the last 18 months our risk systems have evolved quite significantly so we actually pulled out about 2 hours before the flash crash occurred. We basically consumed our risk budget for the day and then our systems turned themselves off.

Larry Levy: Would you do things differently if you could go back and, if so, how?

Bernard Orenstein

Bernard Orenstein

Bernard Orenstein: Every time we have had a significant drawdown we have done some soul searching, research and learnt from that experience. Where we are now is a long way ahead of where we were a few years ago in terms of our risk management. We are a lot more comfortable. Even in terms of managing deep drawdowns we do this systematically. We do have a philosophy of continuous improvement. I don't think we will ever rest on our laurels.

Larry Levy: Tell us about the eStats Revolution Fund

Bernard Orenstein: This is a Bermudan-based master-feeder structure that we are launching in November. We have been trading on a prop basis for some years but over the last 15 months or so we have uncovered what we believe is more capacity than we can potentially consume in any reasonable time frame on a prop basis. We believe the capacity for our current model in the futures market is around about 100 million US. This is at our high leverage, a normal fund would trade this at around $1 Billion US So rather than leave the money on the table our view is to involve external investors and take a performance fee. Take the capital we have on a prop basis and put that alongside investor capital and trade together.

We happen to be structured as a hedge fund but really it is a prop-trading market-making style of firm in hedge fund clothes. In the prop space the traders or strategy providers will see the 50/50% type profit split all the way up to an 80/20% profit split in favour of the trader so we are taking that type of model and turning it into a fund.

Rather than aligning with a single capital provider what we are doing is introducing a diversified investor base. The last thing you want when running a business with a single capital provider is having the rug pulled from under you for whatever reason and by having a diverse investor base we alleviate that risk.

Larry Levy: Tell us more about the strategy. What are market micro-structure dynamics?

Bernard Orenstein: We focus on historical data that we have collected from exchanges over the years with full market depth. We look for patterns in these data and it's in these patterns where we bring our research edge to the market place

Larry Levy: You profit from counterparties needing immediacy. Can you explain that?

Bernard Orenstein: The core basis of the model is an electronic market maker. They provide liquidity in the market and they basically sit there in the bid or the offer reasonably agnostic and provide that liquidity to those that demand immediacy. Let's say there is a retail trader, an institution or a trend follower who has a signal and they want to act now. They will cross the bid:ask spread and the locals or firms such as ourselves will sit there on the other side of those trades. Our return is largely a function of the bid:ask spread, sometimes wider and sometime narrower.

Larry Levy: Do you benefit from the statistical algo in the times the market is doing very little and you are picking up ticks from genuine two way trades?

Bernard Orenstein: We basically trade the noise rather than the trend. We believe the market is dominated by noise and occasionally there is signal. We attempt where possible to recognise when that signal comes and stand aside or abide. We are only able to do that to some degree of success and we are certainly not able to profit everyday.

Larry Levy: So statically the market spends most of its time in the noise area and so unlike other funds that wait for direction you basically developed a way to profit from the noise?

Bernard Orenstein: Yes that's correct. One of the positives is exactly that along with diversification benefits. If you look at the correlations with our returns against the S&P500, CTA or the hedge fund indices, the correlation against each is really low, around about the 0.15 mark.

Larry Levy: So you are not benefiting from the trends rather just the two-way markets.

Bernard Orenstein: Yes that's right

Larry Levy: How does market depth affect what you do?

Bernard Orenstein: It's really important, particularly in our simulations and our signal generation. Market depth is the food that we eat with regards to the way that we develop our models. It's a bit of a barrier to entry but there are a growing number of firms that do consume full historical market depth. It does take technical and intellectual horsepower to work through that.

Larry Levy: You study and develop algorithms to some extent based on market depth which determines your reaction?


Bernard Orenstein: Yes. We have a simulation platform that we have developed over the years which allows us to replay full exchange information, historical market depth and tick data alongside our own trading systems. We see our results on a historical basis.

Larry Levy: Do you find there are major differences between different markets? For example different currencies like Euro vs Yen vs S&P or do you see the same patterns across the board?

Bernard Orenstein: We see the signatures are quite different across markets from a market microstructure perspective. The underlying principles of a market dominated by noise rather than signal, we believe, is sound.

Larry Levy: Tell us about how you develop and test the algorithms. The technology, people involved, typical tests times.

Bernard Orenstein: We have a research process where everybody in the organisation is involved. There's an ideas database. My role is to scan the ideas database and then help decide where we allocate resources. We run a portfolio approach with our research programme, and manage spend in multiple categories between strategic research to process improvement.. But fundamentally, trading impact is really important to us. Everything we do needs to revolve around trading impact and this is what drives our portfolio approach.

Trading impact is the impact on the bottom line. What impact would an idea have on our net P&L, P&L per trade, capacity, et cetera? Any research idea in the initial stages would require a bit of guesswork in terms of its impact. The approach is to wash risk out of each research idea by allocating resources and then adjusting the trading impact. And we estimate the probability of success, and come up with a better quantifiable cost of the research endeavour or its duration. Passion is a strong driver. If there are two ideas that pass the trading impact criteria but someone feels particularly passionate about one of them, that one gets it.

Larry Levy: You say that everybody's involved?

Bernard Orenstein: Everybody in the organisation can develop software. That is a key skill they must have before joining the team. We don't limit the research just to the guys with PhDs or the maths people. Everyone has a perspective coming from different backgrounds and it's when those perspectives come together that we see good results. One guy on the team has a PhD in automated high frequency trading and another was a motor mechanic. He doesn't have any tertiary qualifications but has a very practical bent.

Larry Levy: So you are using a fair degree of empirical data and observation to build into your modelling?

Bernard Orenstein: Our trading ideas come about through a mix of approaches.

Larry Levy: Talk about conventional risk management, circuit breakers, stops, et cetera.

Left to right, back row: Adam Grey, Robert Moxham, John Marchetta, Dr Zac Harland, Milan Peters. Front: Bernard Orenstein

Left to right, back row: Adam Grey, Robert Moxham, John Marchetta, Dr Zac Harland, Milan Peters. Front: Bernard Orenstein

Bernard Orenstein: We have a range of risk management techniques. These span financial techniques through to operational techniques. At the highest level we run the portfolio with a maximum 50% drawdown so if we breach that 50% we terminate the trading program. That is from an equity high so it will be from any prior daily equity high through to maximum drawdown. If that happens we close the fund, return capital to investors and go fishing. We then come back later to decide what to do. If that happened it would mean we had some sort of excursion well beyond theoretical statistical expectations that would indicate something is broken. We would take some time before turning back on.

Larry Levy: What else?

Bernard Orenstein: We have a risk overlay which uses a fairly traditional linear scale back technique. We also have a combined time and drawdown based overlay where we reduce aggression if we encounter clustered losses. We also have maximum daily risk budgets and maximum budgets for the portfolio and daily budgets for each individual market system. Also we have session budgets for Asia, Europe and US and their individual systems. We have intra day risk limits in terms of exposure, open positions, order quantities and they are all enforced by the pre trade risk management software that we use. That's the commercially available software called FrontRunner from Object Trading.

Larry Levy: How far can the fund go down before you panic?

Bernard Orenstein: We have had three significant drawdowns of the order of about 30%. I don't think you would see us panicking until it reached about 40%. Between 40-50% there would be concern but we are quietly confident in our risk management systems and it would have scaled back. We have a management team which incorporates our risk committee and investment committee. If we saw something particularly untoward or something seemed "broken" then we would do something about it. We wouldn't run the system into the ground just for the sake of it. We would have to have very good reason to do anything discretionary.

Larry Levy: Who are the key people, what do they do and how long have you known them?

Bernard Orenstein: Steve Woodyatt is my business partner of about 20 years and we met each other in the late 80s at BHP. We have worked together since then and launched 3 businesses together and we sit on each other's boards as non-executive directors.

Steve and I founded Object Trading where he is the MD. I am the MD of eStats. It's basically an opportunity to provide external perspectives on what is going on in the complete trading space. We are both fully aware of the strategic plans for our businesses and we help develop those plans. So one of us will always be boots-and-all involved with their business and the other one can provide a bit of perspective.

Rob Moxham is our Director of operations and one of the Bermudan fund directors. He runs the trading operations and execution technology. He largely runs the business on a day-to-day basis. He was the first person we brought on board that came from the sell side. He's ex-ABN AMRO Futures in Sydney and Singapore, and HSBC in Hong Kong. We shared a vision together back in 2003-2004 when I launched eStats and I needed a clearer that appreciated what I was trying to do.

We have a senior researcher called Zac Harland from the UK. We brought him out here around three years ago. He has a PHD in automated high frequency trading. He brings an academic perspective with an applied flavour and a certain sense of rigor into the organisation.

In total, we have a team of ten. Six full-timers and four part-timers. We have people that span a broad range of things. We have a chief technology architect, trading operations people and masters of finance graduates. Virtually everyone deals with software.

Larry Levy: Does your location, near Sydney, come with issues in terms of speed of connectivity and low latency, or do you host at the other end?

Bernard Orenstein: We have data centres in Chicago, Frankfurt and London and smaller data centres in Hong Kong, Singapore and Sydney, so our location is largely irrelevant. We simply have a trading desk which is an operational monitoring desk rather than a source of order generation.

Larry Levy: Are you sub-3 milliseconds to a lot of your exchanges?

Bernard Orenstein: Yes. We are sub-1 with whatever we control.

Larry Levy: Do you co-locate like these guys that are almost in the same room as the exchanges?

Bernard Orenstein: Yes we do co locate. Latency is important to us but it's not ultra-important because we aren't a latency arbitrage-type firm. So we are in the space where hundreds of microseconds matter but 10s of microseconds don't.

Larry Levy: That brings me on to talking about the software and hardware that you use along with the communications. What do you use for operational risk management?

Bernard Orenstein: On the operational risk side we have a few things. We have a Command and Control Centre (CCC) dashboard application - that runs at each of our data centres. It's hooked up to each of the trading systems and each trading system has a dead man break. If it's not being monitored then it turns itself off. It cancels the orders and flattens any position it might have.

Larry Levy: If there is suddenly a global event of some kind and a centre is cut off from everything else it will automatically cut out?


Bernard Orenstein: Yes. That's the basic principle that runs through the entire organisation. Be conservative unless our research and risk management systems determine we can act more aggressively. We have a CCC user interface back in Sydney where we can see what is going on. The CCCs do the monitoring of the trading systems and get a snapshot of the entire picture of each system and if there is anything that is out of spec according to the rules then it will automatically alert the operator.

Larry Levy: Is it a 24 hour monitoring system? Do you have something to alert you when you are out of the office or do you always have someone sitting in the office?

Bernard Orenstein: We have someone sitting in the office 24 hours a day. That is an emergency type of role only. There isn't a lot to do as the systems turn themselves on and off. At the start of the trading week they come on and are turned on and off automatically through the trading day. Then they turn themselves off at the end of the last trading session globally. From time to time there can be technology issues however and that is the job of the Operator to monitor.

Larry Levy: If I say to you that you have to go away for 4 weeks and leave all these things running and everyone else has to go away with you would the systems run themselves satisfactorily?

Bernard Orenstein: Yes that would work but I wouldn't be comfortable doing it though.

Larry Levy: You wouldn't be surprised to come back and see your system has made you money then?

Bernard Orenstein: Over a four week period I would have expected some glitch along the way but I would have thought it could run for a week or two without any sort of glitch. In a month there would be a 60-70% chance it could run without drama.

Larry Levy: Are they independently redundant? If one thing closes down would the rest carry on running?

Bernard Orenstein: Yes. One of the nice things is we aren't a dealing desk. The worst-case scenario for us is that we have to stop trading. It's not like we have to get up and running again in a few minutes. We don't have clients screaming down the phone saying where is my order I want to do this or that. We just stop trading.

Larry Levy: You do expend a huge amount of effort on the risk-management side from many different angles, don't you?

Bernard Orenstein: Yes, it's pretty relentless. There is always new risk that we uncover that we do look to address.

Larry Levy: Is that essentially because you are a highly geared operation and that's why you have to expend a lot of effort in this area? So because you are a high risk high reward firm you have to make sure everything is just right?

Bernard Orenstein: Exactly. In some respects we drive the high Octane vehicles in the trading space the risks are real and we are highly leveraged. We take our risk responsibilities very seriously - not just for our prop funds, or our investors, but for marketplace health too.

Larry Levy: What is your leverage?

Bernard Orenstein: It's hard to say when you are talking futures. As a rule of thumb a million dollars invested with us would turn around on an average day 250 million dollars of notional value of futures traded. That's a really rough number that spans commodities, foreign exchange, equities and bond portfolios.

Larry Levy: Would that be at any one moment in that day or is that across the entire span of that day?

Bernard Orenstein: That would be across the day and at any one moment an exposure of one to two million would be pretty normal

Larry Levy: That's based on current AUM?

Bernard Orenstein: That's just on a per million dollars invested.

Michelle Polome

Michelle Polome

Larry Levy: So It's 2:1? At any one moment in the day if you have 1 million invested you have about 2 million at risk?

Bernard Orenstein: Well that's not at risk. If there was 1 to 2 million exposure at any one time then that would be pretty normal. On a more extreme level it could head out towards ten times that amount as well on an intra day basis.

Larry Levy: Talk about software, hardware and communications, please.

Bernard Orenstein: On the software side I have mentioned Object Trading in terms of providing the exchange infrastructure layer and the unified API. Our trading software really sits on top of that API so we have a high performance code.

Larry Levy: Do you write in C++?

Bernard Orenstein: It's a mix of languages. C++, Delphi, C# and PHP . We also use Business Rules Automation Technology to represent our trading rules. We use a product called Common Knowledge from a company called Object Connections which myself and Steve, along with some other BHP buddies, run and are actively involved with.

Larry Levy: How about the comms? Who do you use for them and how is that structured? How redundant is that and how do you work that?

Bernard Orenstein: We use a Chicago-based network service provider. They take care of our data centre and our communications. We don't have to use any specific carriers because we are co-located.

Larry Levy: In general why is it configured that way?

Bernard Orenstein: Simply for low latency. We just need to be competitive. It's just an irritation to be honest. Whenever an exchange introduces co-location in order to be competitive we need to set up a data centre there. If Intel comes out with a new chipset or HP have a new model then we have to buy it. Being part of the arms race is annoying but necessary. I'm sure many of our peers would agree on that one.

Larry Levy: What do you do when a fuse blows?

Bernard Orenstein: If there is some sort of network failure between our data centre and the exchange then our CCC would detect a break in communications and then alert our operator. The operator would run a standard operating procedure which is fairly straight forward. Number one is cancelling the working orders of the trading systems exposed to the failure. Step two is determining any net open positions that may exist and step three is flattening that position through our broker.

Larry Levy: Do you like technology?

Bernard Orenstein: Not really. I think I am very much an applied technologist. I don't really like technology for technology's sake. I do like the fact that we are using cutting-edge stuff but I'm not a tinkerer and when I come home the last thing I want to do is sort out the family network or try and get the printer working or anything like that.

Having said that, I am passionate about applying technology to the cause. I have a passion for automated trading and if you were to ask me what was more important, the research, the trading systems you bring to the market or the technology, I wouldn't be able to answer. I would say it's 50/50. I think it's important for firms to have feet in both areas.

Larry Levy: How do you recruit staff?

Bernard Orenstein: Where possible we try and reach into our own network for staff. That isn't always possible and so we may occasionally use recruiters. The important thing is not to rush recruitment campaigns. For us, a key driver is passion. We are obviously assuming all candidates can write code and do everything else well, but we are looking at what they will bring to the organisation and how that complements what we already have. If someone has passion and it's aligned with our own then they are keen to get the job done and it works a lot better when people want to be there.

Larry Levy: Who is the fund for? Why did you make the decision to target it at them and what particular features of the fund are designed to attract that investor base?

Bernard Orenstein: De-correlation is particularly valuable to investors. We target investors that seem to be quite systematic about their portfolio construction and would aim to introduce de-correlated components. That is something we are certainly able to offer. We have looked at correlation in the range of about 0.15, plus or minus about 0.05 with all of the major traditional investment or alternative fund indices. We need a million dollars' minimum investment.

This is a leveraged product and as a result we want our investors to understand the costs of leverage. Our fund is tuned to maximum 50% drawdown. An investor might only be comfortable with a 10% maximum drawdown and they may want to have say a 10 million dollar exposure. They can keep 8 million for themselves and just give us 2 million to manage and that way they are managing their own cash. We aren't in the business of collecting huge amounts of cash and managing it. We just want to create return for our investors and fees for us.

Larry Levy: Please describe your typical working day. Does your routine vary according to market conditions?

Bernard Orenstein: I don't think there is much of a change to my routine. My working day is pretty much get up at 6:30am and do a few family chores while checking emails and seeing how the performance has been through the night, the US session. I get to the office around 7:30am. We are located at the beach so every other day I try and get out for a run and a swim before work. Other days I might be on the early Asian shift so I'll be on the operations desk. While on operations I will have a chat with the previous operator for about half an hour just to talk over any issues or any research that operator might have done overnight.

At the moment I do some of the software development, so I try and sneak a few hours of that in during the day but invariably there are a heap of interruptions that do occur from all over the place. We normally have a handful of meetings during the day with staff related to research activities, HR issues, technology type issues along with the regular Monday morning meetings and Wednesday afternoon research meeting and Friday board meetings. Around about 6pm I head home.

Larry Levy: You are in a drawdown at the moment at about 20% so I could rightly ask if your edge has gone and what makes you think you are going to bounce back?

Bernard Orenstein: As a trader you never know with 100% certainty that you will bounce back. However, the systems that we have built really operate at the core of the way markets operate from a market micro-structure perspective. So while markets operate in this particular way we expect the edge to be there statistically. We have experienced deep drawdowns in the past and we have always recovered quite strongly from those. I wouldn't be comfortable at all being a short seller of our own equity curve

Larry Levy: Have you calculated your chances of annihilation?

Bernard Orenstein: We used our maximum drawdown at risk technique and we calculate it as a 1-in-100-year scenario - excluding the possibility of a black swan event That would exclude our risk management overlay so if we ran without our risk overlay and scale back then that would be the sort of probability we are talking about. With the risk overlay our level of aggression does reduce as we approach 50% so under normal market conditions that probability in infinitesimally low.

Larry Levy: Do you live to trade or trade to live?

Bernard Orenstein: I would say I live to trade. I love what I do and I have no desire to retire. This is a passion and I hope to do it as long as I'm able.