Please tell us a little about your background and what you each particularly like about the currency trading and investment business?
Stephane Coquillaud (SC): I have a quantitative background in Finance and Insurance, being both a qualified Actuary while also holding a Master in Computational Finance from Carnegie Mellon, Pittsburgh.
I also graduated from ESSEC (École Supérieure des Sciences Économiques et Commerciales), a leading French Business School. I came to working on Foreign Exchange markets first from a risk management angle as I ran the market risk team for a few years at Dresdner Bank on FX, commodities and short term rates prior to switching to a trading role in emerging markets exotic options. After that, I ran the FX exotic derivatives trading business at Credit Agricole.
David Hitchins and myself then moved together to TD Securities, in the aftermath of the credit crisis, when the risk appetite for most complex derivatives reached an all time low. This is where I got involved in algorithmic trading. Three years later and now here we are having launched Alacrity FX.
David Hitchins (DH): I started working on the FX markets immediately after I got my BSc in Combined Science back in 1997. First at West LB, where I learnt the basics as a junior FX Options trader, based in Dusseldorf for a few years.
I then landed a job at Bank of America on the FX Options desk and was quickly given more and more responsibility, first with some overseas assignments in Tokyo, New York and then at the London desk a few months later I was promoted to the role of Global Head. I actually ended up after just a few years being one the youngest appointed managing directors in the Bank, running quite a large team of just over 50 people. Later on I ran similar FX options businesses at Credit Agricole, and then TD Securities and UniCredit prior to launching Alacrity FX with Stephane Coquillaud in late 2011.
What core activities does Ganymede Partners undertake and what kinds of clients and investors are attracted to the services of the firm?
SC: Ganymede Partners’ core business is hosting an alliance of early stage hedge funds. The aim is to provide emerging investment managers the full range of corporate services needed for them to reach a business critical mass and perhaps become independent at a later stage.
Amongst other things they facilitate all the regulatory arrangements, legal and compliance, fund structuring and most importantly distribution. They have extensive experience in raising capital for alternative asset funds from a wide range of institutional, family office and individual investors.
Thus managers experience a similar infrastructure benefit as if they were running a stand-alone strategy as part of a large hedge fund group. However they will retain ownership and control of their legal entity and brand. The key advantage is economy of scale.
In the current environment, the most significant challenge of an early stage manager is to reach the $100 - $200m of AUM that enables investment in the infrastructure needed to meet investor due diligence standards. GMA (Ganymede Multi-Manager Alliance) is structured so the required institutional standards are met from day one.
In addition, Ganymede Partners also provides hedge funds with finance solutions such as leveraged share classes, liquidity facilities, or other tailored payoffs.
Last year you launched the Alacrity FX trading strategy. What does this aim to exploit and how does it seek to achieve returns?
SC: Having a long track record in trading and risk managing complex portfolios of derivatives, we had been exposed for years to the effects of the higher order dynamics of the markets, such as skewness, kurtosis, jumps, stochastic volatility, correlations, mean reversions, break outs, etc.
Our idea was to aim at detecting recurring patterns, fundamental dynamics in the micro-structure of the markets which are persistent in very large samples of high frequency data sets and over long time ranges. Then use these findings in the short term to trade systematically and profitably.
For this purpose, we quickly dismissed classic top down approaches when trying to establish an explicit “theory” of the markets and instead deployed sophisticated machine learning techniques combined with large computational power to evolve trading strategies according to predetermined and suitable target behaviours.
What are your main day to day responsibilities within the company?
SC: I look after the day to day trading, which includes running the trading algorithms and overseeing the downstream middle and back office processing. Since all our processing is fully automated from front to back, with absolutely no manual intervention involved, this still leaves me freedom to keep working on research and development matters, which takes up most of my time and is actually my main interest. At the risk of being a bit blasé it can become rather boring to sit there and watch an automated algo trading. The real excitement is in the conception and implementation of it!
DH: I am also involved in overseeing the trading with Stephane. As the more eyes - the better. Even though our processes are fully automated, there is still always the odd glitch that will necessitate interaction with the brokers, manual reconciliation, or a deep dive in the execution work flow. This being said, my main role within the company is to run the distribution of the strategy to institutional investors.
Alacrity FX is a systematic and fully automated program. What currency pairs is it trading?
DH: We focus on the main currencies: USD, EUR, JPY, AUD, GBP. And trade all the crosses between these pairs. This is probably one of the most liquid markets existing and this allows us to be able to trade on a 24 hours basis, 5 days a week. The execution engine is designed in such a way that we do relatively small trades, several hundred times a day.
The vast liquidity available in the underlying forex market coupled with the fact that we never take large liquidity off it allows us to have a highly scalable framework with no foreseeable capacity issues.
What processes are involved with the eventual trading strategy that is executed for any currency pair?
SC: A number of processes have to run in parallel in order to facilitate our trading and their importance to the whole result cannot be underestimated:
On the front office side, we record the price action aggregated over multiple ECNs, continuously tick by tick. This tick by tick data is sent to the hundreds of strategies for crunching and generating trade signals. When triggered, all these signals are then aggregated in an internal liquidity pool in order to allow for offsetting of the buy and sell orders where relevant, so that we only trade out in the market the net value of all the signals, which obviously saves some transaction costs.
Simultaneously, we have an overlay tactical capital allocation set of quantitative strategies, which reallocates dynamically the risk across the various strategies, hence re-weighting constantly the trade sizes in the internal order pool. Ultimately, we have a smart order routing process, which decides which ECN we will channel the orders to, based on optimal price, liquidity and latency among other things.
On the back office side, we send the trades to our prime broker every five minutes, for them to match with the same information received. These trades are then split by the prime broker for each managed account we run, and the reverse given up to each of the custodian brokers holding managed accounts. We then send to the custodian brokers the list of trades they should have received from our prime broker and they can then also perform an independent matching. We do the same with the investors so that they can match with the information they have received from their custodian broker.
Furthermore we receive a twice a day recap of all the trades from our prime broker and we can therefore perform a final reconciliation against our back office. This procedure, even though it sounds a bit complicated, ultimately enables each party in the process, be it Alacrity, its prime broker, the final investor or its custodian broker to perform an independent reconciliation provided by two contradictory sources of information.
What is your current minimum account size?
DH: For the moment $2 million minimum due the fact that we do not operate on a managed account platform yet. We therefore have quite a lot of piping work to do for each new account as per the processes previously described, which is why it makes economic sense to us only beyond this level.
This floor will however be shifted as we are soon to launch on a managed account platform.
What is your current and projected AUM?
DH: We launched with $15 million and are currently working at on boarding two additional customers for an additional $5 million.
What kind of leverage do you use and does it vary for any reason? Can the client choose leverage factor?
SC: Absolutely. Each customer will have their own tolerance and aversion to risk given an anticipated level of return. We make a point of understanding these and thus derive different traded leverages on each individual account as a result of this exercise, while always maintaining a relatively conservative approach.
So far we are being asked to trade between 0.75 and 1.5 of leverage generally which corresponds to very classic targets of double digit expected returns and single digits volatility of daily returns. However, absolutely nothing would prevent us from tailoring a highly leveraged product for a high net worth risk taker for instance, or conversely from designing a very conservative version of our product for a pension fund, for example.
“Ganymede Partners’ core business is hosting an alliance of early stage hedge funds. The aim is to provide emerging investment managers the full range of corporate services needed for them to reach a business critical mass and perhaps become independent at a later stage.”
Our aim is to derive systematic profitability by classical trading means, i.e. going long or short on the market. The tools we have designed to analyse the micro-structure of the markets and the occurrence of recurring patterns allow us to do this profitably over a long period of time by constantly adapting to market evolution, whilst staying immune to possible changes due to the increasing scrutiny of regulators in the electronic trading space.
How do you go about back-testing your trading models to confirm that your strategy will perform as required?
SC: We are building our strategies on a number of years of so called in sample data. Once we are happy with these, a second selective process takes place where we look at the performance of these strategies over several years of out of sample data. Such a validation step is a key element to ensure a candidate strategy exhibits enough statistical confidence to be promoted to live trading.
In other terms, the name of the game in our approach is to avoid over fitting. Over fitting is a typical issue in machine learning where your strategies end up capturing too much noise and not enough of the underlying process which you are trying to model. This translates into great performance in sample and bad or at best random performance out of sample.
To avoid such a negative behaviour, well known techniques exist, such as cross validation, regularisation and other Bayesian approaches, which all help in the model selection process. Time after time, we have elaborated upon our own protocol and are constantly improving it.
Does Risk management influence and shape your overall trading and investment philosophy and in what ways does your operational methodology add to the effectiveness of overall risk control?
SC: Yes most definitely.
There is a large gap in the perception of risk in general as a trader working in a large investment bank alongside other traders on one hand; benefiting from an army of trade booking support, middle office, back office, risk management, risk control, and compliance staff and being a small emerging manager team on the other hand, with limited resources and daily meeting your investors eyes, investors who have trusted you and are taking risks because they believe in what you do. Emotionally it is just not the same thing.
Partly but not exclusively as a result of this, risk management and control take very naturally a prominent place in everything one does as an independent investment manager. It just happened that David and myself have similar ideas about many things including risk management and we decided to implement a parallel process to the trading. Call it a sort of algorithm monitoring the algorithm so to speak. We take a cross monitoring and joint view about implementing market risk limits in the format of various cross sectional trading limitations, trading, strategy, and currency pair allocation at portfolio level.
But also, this way, we perform operational risk monitoring which involves limiting or measuring such things as trading frequency, execution latency, connection status to various liquidity providers, clock synchronisation across the various servers, trade rejection notifications and and numerous other operational events that could negatively affect trading if not under control. All this goes beyond simple market risk exposure limitations. Any exception to this framework generates an immediate electronic notification to relevant distribution lists.
How frequently do you fine tune your strategy and what steps do you take to ensure that your trade execution pathways are continually optimized to meet differing market conditions?
SC: We do re-tune but the frequency of this is something we consider to be a proprietary matter at this point.
Is latency an important variable in your trading process and if so what steps have you taken to reduce its impact?
DH: Yes and no. It is important in one sense because there is clearly a cash premium attached to low latency and just like anybody else, we hate to leave money on the table due to sub par infrastructure performance. Reducing our pre trade and execution latency is what allows us to increase our trading frequency, and thus to improve our risk return ratios ultimately. So it is always an area of improvement, under constant monitoring and something we look at enhancing the best we can.
For this reason, we monitor the processing time of our counter-parties and systematically detect and remove those which are behaving as ‘last look’ liquidity providers. Why bother being co-located and putting a lot of effort to target a few milliseconds of acknowledgment time to reach the LP if subsequently one allows him to use a free option of 200 or 300ms to decide or not to honour the price he published?
On the flip side though, we are just trying to earn money by simply smartly generating alpha, taking long and short positions in the markets, and benefiting (or not) from intra day market moves, which really is good old fashioned trading!
The difference is that we are using cutting edge technology and artificial intelligence to achieve this. This is very different from actual High Frequency Traders for whom a tenth of a millisecond can be a critical amount of time and which could even affect their overall business model. For this reason we believe that there is value in building a business which is somewhat resilient to the latency shock waves that could be induced in the near future by the regulators.
In what fundamental ways do you think your firm’s investment approach, and use of technology and market intelligence differs from many other managers?
SC: Many of the most common and historically successful approaches are well known, top down, low to medium frequency trading techniques: macro/fundamental, event driven, long/short, trend following, relative value, arbitrage, etc. To name a few, and irrespective of the asset class and the location traded.
We do not fall clearly into any one of these categories to the extent that we are up front completely agnostic as to which technique should be used. We allow for many of these approaches to be possible in our framework and let the data and the performance drive the optimal trading style in any given market environment. This translates somehow in a weighted portfolio of trading styles, constantly being dynamically reallocated.
As a result, our performance, whenever good or less good, is vastly uncorrelated to most of these standard strategies and our distribution model consequently appeals to those looking for genuine alternative alpha sources, as a diversification investment tool.
What electronic trading platforms do you find most appropriate to use and what factors influenced your choice?
DH: We looked at many systems when we started the project across the spectrum of vendors. These ranged from the few well known retail products, which you can get for free, up to a few hundred bucks a month to the largest institutional complex product offerings, necessitating in certain cases multi million dollar budgets.
We realised after some research that we needed an institutional grade solution when it came to flexibility, customisation, technology and robustness, which supported the full suite of functionality in generating and deploying alpha seeking strategies, while staying at intermediate budget levels, always being careful not to let costs run away with themselves, before having generated the PnL.
Ultimately, Deltix proved to be this package. It is at once powerful and fully functional at a reasonable price relative to other institutional alternatives I could mention! Beyond technical and financial considerations though, what we really found with Deltix as a bonus, are genuine long term business partners. Firstly, this shop is a real powerhouse, populated with the smartest quants one could think of, whilst the senior management, fully aware of the challenges facing start up investment managers like us, is always careful to help in making the most efficient decisions on the technology side.
How did you go about building your trading desk IT infrastructure and how was the trading software and connectivity technology provided?
SC: Deltix runs under the MS.NET framework on Windows 64 bit. The data is processed by Deltix’s own proprietary event-oriented polymorphic database engine. For the trading environment, we are running three servers, each with two quad core Xeon processors and a 3 GHz speed, with 24 GB of RAM on each one. One is dedicated to live trading and the other to UAT validation pre production as well as live data recording. The third server is a backup for each of the first two and can therefore replace either the trading or the UAT on the fly.
All this infrastructure is located at an of Equinix Data Centre in New Jersey. For the R&D environment, we are running in another data centre a cluster of multiple computing nodes each equipped with six dual core AMD Opteron processors and 32GB of RAM in order to distribute the calculations and therefore speed up our back testing.
Looking ahead, where do you see the main challenges facing the Alacrity FX team as you seek new ways for capturing and exploiting investment opportunities with currencies?
SC: We are trying to diversify our trading along two axis. First and foremost, our existing strategy is asset class agnostic, it being an artificial intelligence computational framework. Whereas we were working within forex departments in investment banks and were therefore only allowed to trade FX, it is natural that the first implementation instance of the trading program naturally focused on this asset class.
However, this somehow fired back at us, as we have learnt the hard way, moving from the sell side to the buy side; since FX is a vastly under invested area by the largest institutional hedge fund investors, compared to equities, rates, credit or commodities. We are obviously very biased in saying that, but we feel there is a missed opportunity here for the largest multi strategy fund of funds community not to add a really genuine uncorrelated alpha source to the main classic risk factors they usually trade.
So while we believe the invested amounts in FX specific trading strategies can only grow going forward and that we are ideally positioned to capture market share as this happens, we have to recognise that it was not the easiest way to start up a business, even if we had no choice.
We are therefore very actively looking at extending the current framework to other asset classes. Additionally, we constantly explore alternative approaches, be it in the field of machine learning techniques in the high frequency spectrum or more conventional top down modelling in the low to medium frequency range, though that stays in the test lab for the moment!
Are you pleased with the performance of Alacrity FX to date and has it met all of your investment objectives so far?
DH: As of April 2013, our trading trading program is live now for 28 months and so far we have accumulated a total trading performance just over 20%, on an annualised volatility of the returns of 7.5% and without drawing down more than 4.5% from peak to valley during that time.
We feel we are where we should reasonably be in terms of Sharpe Ratio given the medium frequency, short term horizon, single asset class features of our trading program.
However we are pleased but not satisfied. As the president of one of the largest systematic funds in the world recently told me: “One ought to try to push oneself outside one’s comfort zone, right? Otherwise we will not evolve”. Agreed!
What options do investors have for accessing Alacrity FX?
DH: At the moment we are offering access to the program via managed accounts only. The client investor can chose any custodian broker of his choice to open a managed account and we are using standardised inter broker dealer, give up and reverse, give up procedures to channel the trades from the various trading venues through to our prime broker and ultimately distributing these to the various custodian brokers. Full reconciliation takes place daily to ensure that each position in each managed account is as it should be at the close of the business day.
Furthermore, we are currently pursuing several diversification strategies to offer access to alternative investment routes to our trading program. Firstly we are aiming to be available on a selection of the major managed account platforms in the next few months.
Secondly, we are working at seeding the launch of a cross asset class fund in the medium term. Thirdly, we are also exploring with our partner Ganymede alternative structured routes such as issuing performance linked notes for those players who would like to participate in our trading performance without being able to explicitly allocate capital via a fund or a managed account.
What are your plans in the future?
SC: This is perhaps your most difficult question! To be perfectly honest with you, we are now in action mode and we are simply are not reflecting that much. We know the next few months are going to be critical for Alacrity FX, for all the reasons previously explained. Therefore the best we can do is live in the present by staying completely focused and doing our best day in and day out.
Personally, as long as I am working in a challenging environment and excited with my day to day work, I am very happy, this helps me not to be too anxious about the near future.