Daniel, what is your background and how long have you been in the currency trading and investment business? What attracted you to the industry?
I hold a MBA from UCLA and a PhD in economics from Sorbonne University (Paris). My 35 year general management career spans across Hong Kong, Los Angeles and now Geneva where I have lived for the last 20 years. I have traded either personally or professionally options, futures, indexes, stocks and forex. In 2003, I created the Autopilot methodology (a multi-timeframe momentum play) which I licensed in several countries. Exponential Ltd was created in September 2005 in Hong Kong and Geneva, and started marketing the intellectual property developed by Mern Capital - as it became FINMA regulated in October 2010. University training attunes oneself to study abstract concepts and processes and I suppose this is the original influence which brought me to the financial arena.
When was Mern Capital founded? Who are the key people involved in the firm and what are their main day to day responsibilities?
Mern Capital is a market research partnership focusing on algorithmic trading and portfolio design. It began informally as Exponential Ltd joined forces with two University of Geneva graduates in the field of statistics, maths and market finance in order to design and back-test expert trading systems, and combined them into an original trading matrix covering numerous currency pairs, entry time frames and strategies.
After a year or so - in October 2010 - a partnership agreement was drawn up whereby my two partners would be in charge of overseeing trading operations, monitoring performance, ensuring the system’s redundancy and responding to a wide array of real time alerts if need be. Additionally, they would develop new algorithms and beta test them within expert systems to be validated and activated into the matrix. By delegation of Mern Capital, Exponential Ltd is in charge of legal and regulatory aspects, administration, client relations, banking and marketing.
Why did you decide to delegate the management of client accounts to Exponential Limited, which is also based in Switzerland?
By law, a partnership is just a contract between several parties. It is not a corporation and therefore cannot be regulated. In our context, delegating the management of client accounts to Exponential Ltd was a logical move.
What sort of clients and investors are likely to be attracted to your services? What core services do you provide?
Our client base is diverse. Exponential Ltd manages HNWI (High Net Worth Individual) client accounts under a power of attorney and a discretionary management investment agreement. We also have relationships with money managers in several European countries and we provide trading signals to firms or platforms offering a managed account program. Exponential Ltd is a regulated manager and does not hold client funds. We favour managed account set-ups as they address investor’s issues such as profitability, transparency and liquidity in a simple and effective manner.
What investment programs do you offer and how are they tailored to fit differing levels of risk appetite?
Our research efforts resulted in three successive generations of expert systems portfolios. Our latest program is offered in two risk versions (balanced and aggressive) with attractive risk return profiles. Both versions use the same trading engine but with different sizes.
How would you describe your approach to trading and in what ways has your proprietary and systematic methodology been designed to yield a higher level of risk-adjusted performance?
Our systems are CPU intensive as they contain on average more than 1,500 lines of code. It includes the typical components (setup, triggers, filters, risk management, money management, etc.) as well as other proprietary routines. The general concept is the combination of a hyper-diversification of market exposure via multiple smaller open positions and a probabilistic trading logic. As we added new currency pairs, new entry time frames and strategies, we obtained a sort of matrix allowing calculation of internal parameters such as correlation, volatility and others. We then worked on the risk reward profile of the composite output of the matrix instead of any individual system performance.
We found that the numerous sets of pairs, time frames and strategies increased our chances of identifying specific configurations exploiting unusual random small patterns or unusual price moves. Using automation, our coverage of scattered market events and anomalies was augmented, as not all sessions, pairs, and trading hours are equally good for trading.
What benefits do you obtain by generating a high number of trades and how do you make use of these?
Typically, numerous open positions (200 to 250 per month) will be allocated a smaller trading size, and therefore a smaller stoploss. A fat tail occurrence is much less likely to affect all of our exposure, and will not become a critical condition. This benefit became obvious during 2011 with events such as Fukushima in February, the market meltdown in summer and the SNB intervention in the fall. Some stops were hit but impact to performance was minimal. Also, one looks at leverage differently when exposure is spread across many open positions rather than just a few. Broadly speaking, the actual leverage is less than the theoretical leverage due to imperfect correlation between pairs and strategies, to long/short offsets and other factors. Consequently, overall risk is lowered.
How does Risk Management influence and shape your overall investment philosophy and in what ways does your use of multiple systems, time frames and trading sizes add to the effectiveness of overall risk control?
We fully automated the back office in the sense that we run peripheral systems to monitor the trading systems. For example, we run a tick by tick comparative analysis of spreads, which alerts us to execution problems, sensitivity to spread issues, random variations affecting performance. We compare the feed quality of providers and negotiate accordingly.
When looking at drawdowns, we use the most severe measure which is the intra trade drawdown, not the daily, weekly or monthly setting. We do not stop at communicating a dry number such as X %, but study the worst 5 or 10 drawdowns in duration and amplitude and look at their shape and the power of the following run-ups. If the account balance surges up powerfully after a drawdown, then this is a sign of good health. Reporting drawdown in this format reveals rather more than just a static figure.
In what fundamental ways do you think your systematic investment approach to trading differs from other money managers, some of whom use discretionary trading techniques?
We don’t need to spend much time on emotions and how they affect trading performance. I shall refer you to the seminal work of Roland Barach (Mindtraps) and the abundant subsequent literature. I do realize that there are top performing non systematic CTA’s, but they are few and far between. Discretionary trading may have deep historical roots dating back to the early days before modern technology, but when it comes to monitoring so many pairs, time frames and strategies around the clock, acting on a signal in a timely and sustainable fashion, I fail to see the human advantage in terms or risk and cost, and I prefer to replace words like intuition or insight by probability and confidence intervals.
The fact is that a powerful machine can monitor many systems with many currencies with a high degree of reliability (if programmed correctly of course) and there is simply no way a human being can monitor that much data on a 24x6 basis.
To what extent have you developed research agendas and analytical programmes to help improve the design of new trading strategies and to maintain the performance of your trading engine?
Active expert systems in our portfolios are monitored continuously. We use many measures of risk or performance (average pips per trade, Sortino and MAR ratios, % win, win/loss, profit factor, net pips, VaR, etc.).
Before we dismiss a system due to its individual performance, we look at the way it benefits the overall matrix. It might be less profitable but cover a unique and specific occurrence (combination of currency pair, timeframe and strategy) and thus bring in a profit not obtainable otherwise.
Candidate systems are in the sidelines in beta test waiting to be validated and included into the mix. So we deal with these trade-offs and the systems come and go as a result of this quality control.
Additionally, we carefully monitor trade execution (spreads, swaps and slippage) and fragment our orders into smaller bits in order to get a better price. As trading sizes grow, you need to look more closely at order books, liquidity pools and providers, etc.
How do you go about back-testing your trading systems and signals to confirm that your strategies will perform as required?
Initially, we performed a 10 year back-test using a classic walk forward methodology. We also use live data to conduct Monte Carlo type simulations on a regular basis.
How frequently do you fine tune your proprietary algorithms and what steps do you take to ensure that your trade execution pathways are continually optimized to reduce latency and better meet your investment objectives?
Latency is not a critical variable in our process. Our systems have nothing to do with high frequency trading, scalping, grid trading, etc. The average time in trade is about 24 hours. We consider mostly daily volatility and focus on entry and exit prices rather than order book timing.
In terms of review process and fine tuning, selecting monthly or quarterly time intervals seems rather arbitrary, and we prefer to rely on statistical and real time alerts generated by the back-office. Various issues are flagged by the monitoring systems, signalled and logged, and we then investigate the matter. Several times a month, we have a performance audit and make sure everything is proceeding as planned.
How have you managed to configure your trend and range strategies to gain better entry and exit prices?
That is a key point which I unfortunately cannot describe as it is proprietary. I can confirm we researched this issue thoroughly and are happy with the results. For the benefit of your readers, I shall mention one of our routines which is the coding of dynamic stoploss and takeprofit levels. Just imagine opposite trailing stops moving these levels, as opposed to fixed values entered upon entry. There were significant improvements in profit factors due to this adjustment.
What about money management?
We are keenly aware of the critical importance of money management and carefully studied the works of scholars such as Van Tharp, Ryan Jones and Ralph Vince. Our routines modify the trading size as a function of equity, buffering the decline and accelerating the run-up.
We had to customize the code since we are working with the composite of a matrix output, not a single position.
Can you share with us one feature of your methodology which significantly improved your process?
Sure. We all want top of book pricing and best fills for a given size. Unfortunately, spreads deteriorate as the trading size increases. So, the best practice is to stay in the lower pricing bands. To address this issue, we have integrated random “fragmenters” into our systems. These routines randomly split the trading size into smaller fragments while also selecting random time intervals between split orders. The full size eventually gets executed, in a lower band, at a better price, and is virtually impossible to “read” by a third party.
Please give us a few statistics about your latest portfolio?
Here is a brief report by the numbers:
MQL expert systems: 35
Currency pairs: 12
Time frames: 4
Average trades per month: between
200 and 250
Average time in trade (hrs): 24
Sharpe ratio > 4.0
Sortino ratio > 8.5
MAR ratio (trade to trade) > 6.1
% Win: 78.6%
Win/loss ratio: 0.33
Profit factor > 1.19
Balanced program: trading size: 0.34 standard lot per US$ 100,000 AUM, average leverage: 1.5, maximum drawdown (intratrade): 6%.
Aggressive program: trading size: 0.68 standard lot per US$ 100,000 AUM, average leverage: 3, maximum drawdown (intratrade): 12%.
How did Mern Capital perform during the financial crisis and in what ways did your investment approach shield you from the worst effects?
Actually, 2011 was our first full live year and a banner year at that. The three sigma events which occurred (the Fukushima catastrophe, the market melt-down in July-August and the SNB intervention) did not impact our operations significantly nor prevented us from achieving a very attractive end of year return. To be fair, I must say that the model did not perform so well in the low volatility environment during the first part of 2012, with static ranges and a general lack of trends. The bright side is that the third generation issue which kicked in around June shows great potential and is making new highs (more than a thousand closed trades in six months resulting in a MAR ratio higher than 5).
You present your methodology as decorrelated. Can you elaborate?
Well, we are currently watching an uncertainty game played between the forces of behavioural finance, the school of fundamentals and market manipulation. The background is globalisation, systemic risk/counter party risk, fat tail events, massive redemptions, the invisible hand and, clearly, many other disruptive factors.
We feel that in this context, the traditional benefits of a classic mix diversification have been dramatically reduced and so a forex component can effectively spread the risk and help mitigate the lack of visibility. Without discussing coefficients and the nature of correlation, let’s just say that the foreign exchange market is a valid asset class, that it is very liquid with fewer gaps and that many small spot positions managed together within a matrix stand a higher statistical chance to produce a stable outcome in various intermediate strength scenarii.
What electronic trading platforms do you find most appropriate to use and what factors influenced your choice? How did you go about building your trading desk IT infrastructure and was the trading software and connectivity technology provided by third party vendors?
We tried a few solutions without success, namely Easy language, Java, C++, before finally selecting the Metaquotes MT4 software. We are well aware this is a retail solution that comes with a few drawbacks. However, we re-coded a few features, and can tell you today that we are quite satisfied with our choice in terms of stability and functionality after having processed more than 10,000 trades with it over the last 22 months. We either connect directly or via a bridge to partners who use the API FIX standard.
On the logistics front, our servers are hosted in a state of the art Tier 3+ Swiss data center boasting all the modern bells and whistles. There, we enjoy high CPU capacity, system redundancy and institutional level connectivity. Our sessions with the servers are based on a VPN SSL protocol.
You have mentioned using the Metaquotes/MT4 software. Can you describe its advantages?
Many operators focus on speed, latency and the capacity to push through zillions of orders. This is not our game as our approach is based on hyper diversification and probabilistic outcomes within daily volatility. Pre MT4, we considered various platforms such as TradeStation/EasyLanguage and some Java based systems.
In the development stage, we found that various flaws, bugs or structural constraints impaired the performance beyond an acceptable level. There was better compatibility between the MT4 platform functionalities and our process, even though the platform was retail oriented. We found that once several internal commands were recoded, the result was quite adequate for our specific need.
We can confirm with our experience MT4’s capacity of dealing with large expert systems (around 1,500 lines of code) and portfolios of around 40 systems.
Its stability (having closed more than 12,000 trades over the last 23 months, in all sorts of market conditions across trading sessions and 21 currency pairs) has proved excellent. We do use electronic bridges to access the widespread API FIX protocol and we do not feel that migrating to a C++/Csharp coding environment at this time would significantly improve the liquidity, slippage or general performance of our operation, as the cost/benefit of a conversion of systems and back office is questionable.
What do the logistics of a fully automated trading firm look like?
Well, we have uploaded our entire process, not just the trading systems and their platform. So, we have also hosted in the same place our back office, our databases as well as a comprehensive set of real time alerts such as ping checks, broker disconnects, excessive leverage, missing stoploss orders, etc. Three people are on call, equipped with secure links to the servers. In practice, we had only one real emergency in two years. This arrangement not only works very well, but is also quite cost effective. You would have to be a medium size bank in order to obtain the same level of security and system redundancy. Research wise, we use a distributed topology, which optimizes staff schedules and productivity. As most clients reside overseas or conduct business remotely, we just keep a small office for administrative duties.
In what ways has technology made your business more sustainable and increased trading and investment opportunities for the firm within the FX market?
We are benefiting from convergence. Modern trading platforms, better coding routines, streamlined executions, increased broker competition, lower commissions, improved access to liquidity pools and increased stability - all these factors contribute in one way or another to our success. This type of fully automated operation was simply not possible just few years ago unless you were a major bank ready to invest millions per year in the technology alone.
“Let’s just say that the foreign exchange market is a valid asset class, that it is very liquid with fewer gaps and that many small spot positions managed together within a matrix stand a higher statistical chance to produce a stable outcome in various intermediate strength scenarii.”
What new strategies and products have you been exploring as part of your continuing efforts to widen the range of currency investment solutions and customized services you make available to clients?
For 2013, our trading programs will be complemented by a bond component, investing the portion of capital not used by margin. This will enhance the overall performance and add another dimension. Of course, there is an interest risk to take into account, but we will hedge the currency risk through manual long term spot positions. We have looked at gold and some futures, but did not find enough potential so far to bring a new product to market.
More generally, we anticipate that China’s upcoming power play will jumpstart new FX flow and liquidity pools in Asia and create new, attractive opportunities. We look forward to applying our models in these markets and have pre-positioned ourselves in Asia thanks to our Hong Kong presence.
Looking ahead, where do you see the main challenges facing Mern Capital as you seek new ways for capturing and exploiting systematic investment opportunities with currencies?
This is probably the most interesting and difficult question. It seems we have reached a technical vantage point after all these trades in spite of our young age. We now have stable technology and logistics and we can focus our energies on business development and new AUM.
We anticipate however that at some point in time, we shall face a new challenge with the liquidity limits of our model. When we approach it across trading sessions around the world, the constraints will start impacting our performance, incrementally at the beginning, and thereafter significantly. We are preparing for this by developing the tools and techniques we need in order to circumvent the limit or at least reduce its effect. Let me close by saying that this problem will be a nice problem to have. For further details please contact email@example.com