The boys are back in town

Beta-Gamma Research is an algorithmically focused software house based in London with an office in Australia. The principal role of the company is to provide software solutions which enable banks and financial institutions to enhance their profitability, efficiency and risk management. Dr Paul C. Tolman, the founder, has a PhD in Mathematical Physics from Cambridge University and is former Head of Quantitative Trading at Royal Bank of Scotland Financial Markets. He is an expert in concurrent programming and applied Bayesian statistics. Paul’s uncle, Howard M. Tolman, is CEO at Beta-Gamma Research and has extensive experience in banking and in particular FX IT.

First Published: e-Forex Magazine 42 / e-Forex Interview / January, 2011

The boys are back in town

Howard, algorithmic trading continues to gain in popularity within FX. What sort of firms are  approaching you for help with deploying Algo trading techniques and what types of tools and strategies are they looking for?  

HT: Before answering this I would like to mention why Beta-Gamma was formed. Our founder Paul Tolman took the view that there were a large number of institutions that could benefit from using algorithmic based trading technologies but could not justify the expense associated with it. We identified our principal market as being banks in the second tier and below, however we have attracted interest from Hedge funds, wholesale and retail distribution platforms as well as banks. It’s fair to say that the principal interest in the banking market at this time is for infrastructure which is why we developed the Aggregation tool. Until a bank has reasonable rate management in place they cannot really reap the benefits which our flow and risk management tools can offer. It’s important to stress that we are firstly a product based company. We are not generally in the business of writing custom made algorithms. 

Trigger Trader is the core algorithmic trading system provided  by Beta-Gamma. In what ways do the execution algorithms and trading tools provided by this solution provide competitive advantages?

HT: Trigger Trader is the product name we use for a collection of algorithms which manage risk and flow and as such it is a reactive tool. Our first bank deployment is a black box solution but there is absolutely no reason why the trading strategy algorithms could not be rolled out on any aggregation platform including our own. In this way  we are providing dealing room automation which reduces direct costs. But the principal advantage is that the flow management algorithms can make money for the user. Efficient flow management has the capability of raising profitability while at the same time reducing risk. Of course the user has to arrange the workflow in such a way that the flow can be managed efficiently in the first place. That is not always the case.   

Do you believe we are moving away from a pure “black-box” approach to algorithmic FX trading towards a more transparent environment where algorithmic toolkits developed by providers like Beta Gamma are more easily accessible?

HT: Frankly yes. Of course we are always going to see black box stuff developed and deployed. After all people are still trying to break the Bank in Monte Carlo. Dr. Paul is quite capable of writing proprietary trading algorithms which may well make a lot of money but if we wanted to do that we should start our own hedge fund and put our money where our mouth is. Having said that, this is something which we may well do in the future. I can’t speak for the whole market but Beta-Gamma’s design philosophy is to build a huge amount of client configurability into the core applications and to then make it easy for business users to react to market or internal circumstances instantaneously. 

What we cannot do is predict the unpredictable. Events like 9/11 do happen and if you are the wrong side of a risk position at that point, well you can’t beat bad luck.  We provide a myriad of configurable parameters within the trading toolkit which are accessible in real-time and can provide a whole series of different outcomes depending on appetite in different areas. 

What factors are important in helping trading firms decide whether it is better to deploy off-the-shelf or customised FX algorithms?

HT: At the end of the day it comes down to what you want to achieve. Cost is also an important factor. If you want your own team of quants then be prepared to pay a lot of money for it. They don’t come cheaply and are not always good. Likewise if you want a customised Algo developed by a vendor for prop trading then be prepared to shell out as you’ll have to pay up front even if it doesn’t work in the end.  As I said earlier it comes down to what is actually required. There is a huge difference between wanting to manage an existing situation more efficiently and profitably and starting from scratch. Give some rigorous intellectual analysis of what you have and what you want to achieve. If you conclude that your current set up is leaving money on the table then work out what the solution might be. My preference would be to buy an off the shelf solution with high levels of flexibility, but I would say that wouldn’t I.  

Do you see any particular risks associated with customising algorithms and does customisation make attempts at benchmarking Algos much more difficult?

PT: When the algorithms are our own, configurable parameters are always checked to be in safe ranges. Further a number of in-built risk controls limit the positions taken and the number of trades placed by

any Algo. We also provide the option for our customers to automatically measure the performance of execution Algos by benchmarking them, for example against the expected fill rate of classic smart order routing execution. We also provide a full simulation of the aggregated market, ensuring that the behaviour of Algos in volatile or wide markets can be tested before they are sent live.

What steps do you recommend that clients should take to measure the success of a specific algorithm or combination of Algos?

HT: It might sound like I am being flippant but the key to being able to measure the effectiveness or otherwise of any algorithm or combination of algorithms is proper management of the underlying data and frankly not every trading house has this capability. I do not want to go into too much detail as this is an obvious requirement but suffice to say that proper analysis of performance can only take place if all of the components of why trades were executed are capable of being collected properly and this includes, in the flow management sense, the data coming from the underlying client trades or other flows. The design of Beta-Gamma’s applications allows a comprehensive calculation of performance against specific benchmarks which are subject to configuration by the client. This is to my mind a very important part of the proper management of FX positions. It is obviously important to measure the profitability of electronic distribution channels by client but not everyone has this capability. Once again I think it is down to the client being rigorous in determining what they need in the way of reporting and ensuring that they have the methodologies to deal with it. Fortunately for us Beta-Gamma’s applications have this capability built in.

In what ways can we expect to see algorithms becoming more “intelligent” and flexible with respect to their order management and strategy execution capabilities?

PT: Modern manual traders, whether clearing or prop, are used to working with the machines. We feel a key element of flexibility is to allow the traders to control the Algos and not vice versa.  We try to get away from the closed “black box” mentality of execution strategies and instead provide a framework in which traders, provided with at least a qualitative understanding of what an Algo does, can configure and customize it to perform different roles depending for example on the currency pair or the context in which the algorithm is being used. Essentially this allows human intelligence to be overlaid on the application technology.

The Beta-Gamma Research offering currently consists of four products, all designed in different ways to work together in creating a more efficient and profitable execution and Algorithmic trading environment:

FX Aggregator: which can  sit alongside other aggregators has a client or browser based front end. FX Aggregator gives access to depth of market, which can be viewed via a bar chart instantly showing who and how deep the market is in real-time. FX Aggregator features various execution principles, including the ability to  trade passively as well as actively, and can take into account differences between trading venues. The system shows rates very clearly, by displaying the main last digits of a currency in a far larger font than the first two, making it easier to really see the rate quickly by just glancing at the screen. Colour coded “Power Dealing Buttons” make one click execution easy. The aggregator can connect directly to banks and or liquidity providers or to to an existing aggregation system on the organization’s intranet.

Trigger Trader: is a modular algorithmic trading system and can be “optionally integrated” into FX Aggregator according to Dr Paul Tolman. Features include market access, order placement and management, position and risk management, execution algorithms and performance reporting.

FX Blotter: provides real-time open position information, mark-to-market (by currency pair and by product) as well as session profit & loss and a fills table, which can be sorted and filtered. It is available as a stand-alone module or as part of FX Aggregator. In addition to the above several add-ons are available to FX Blotter: These deal with Profitability Analysis, Order Management and Risk Management. The Risk Management module uses Beta-Gamma’s proprietary Bayesian short term model of  variance-covariance to provide the ability to calculate real-time value-at-risk (VaR) over any time frame, and can provides automatic calculation of optimal hedging trades.

FX Toolkit: is an earlier product from the Beta-Gamma stable aimed mainly at quantitative market analysis at daily frequency. Some of the models and methods are however used in Beta-Gamma’s high frequency products. Quantitative models include Model View which compares different portfolio models, Bayesian covariance – an advanced model of volatilities and correlation model and  Bayesian gambler - which provides an objective view of the optimal portion of the available capital to “bet”, given current market conditions. Other models include Market Coherence Index - designed to produce strong signals ahead of major market events, an advanced Stop-loss calculator, FX Backtester and Signal-Stabilisation – which uses an application of Bayesian control theory designed to stop the common problem of model “whiplash” with minimal impact on profitability.

The boys are back in town

Do you believe there will be more demand for specific FX algorithms as high frequency and systematic model driven trading spreads into emerging markets and is applied to less commoditised currency products and more complex order types?

HT: I would have thought that this was more or less inevitable. The FX market has always been extremely innovative.While some things have not moved as fast as they might have done  the improvement in the quality of the technical components which drive electronic FX execution in recent years has been impressive. In fact is already happening. I was speaking to a day trader the other day who trades in only one exotic currency pair and makes a nice living on the proceeds. FX will surely follow along the same path as equities did. If anything, technology is moving faster than ever and this is going to create opportunities for those people who are smart enough to handle it and espouse the necessary advances quickly enough. One of the drawbacks of fast moving technology is that shelf lives of particular strategies or technologies are getting shorter and shorter. Take ultra high frequency latency arbitrage as an example. It is not that the advantages go away completely but as time goes on the bottom line suffers as technology expense rises inexorably and profitability is eroded through competition    

Beta-Gamma recently launched FX Aggregator. What is this aggregation engine designed to do and what benefits does it provide?

HT: Although we have just launched this application we have been using the aggregation capacity within a black box trade out system for the best part of two years. I probably made a mistake in positioning the trading strategy algorithms ahead of the aggregation capability. We launched the Trigger Trader flow management product in February 2010 and while it created a great deal of interest it became obvious within a few months that the market was very heavily focusing on trading infrastructure. As such we put the cart a little a little way ahead of the horse. FX Aggregation is the science of pulling together all of the component FX data feeds and execution channels in order to maximise opportunity and ensure the best possible rates of execution. The strength of individual applications from different vendors of course varies however some of the technology is already outdated.  We are very proud of our Aggregator. It’s design, thanks to Paul who is the ex Head of Quant trading at RBS Markets, is state of the art, written end-to-end in C++, and it is extremely fast and has the significant advantage of being designed for use by professional traders using easy to use tools that allow them to manage trading algorithms in real-time.   It assists traders in their job, making them more productive and more profitable. 

The boys are back in town

In what ways are traders able to configure the built-in execution algorithms within FX Aggregator?

HT: We provide a simple algorithmic input GUI which the trader can use to install de install and amend trading algorithm parameters in real time. These algorithms can then be attached to Power Trading buttons under the control of the Trader. Of course a huge amount of the day to day work of a dealing room can be handled electronically but the market is also heavily influenced by events happening in the real world. Our applications allow traders to take account of these occurrences. Traders can also enter a series of what ifs to customise their trading activity to deal with any eventuality. 

In what ways are traders able to configure the built-in execution algorithms within FX Aggregator?

HT: We provide a simple algorithmic input GUI which the trader can use to install de install and amend trading algorithm parameters in real time. These algorithms can then be attached to Power Trading buttons under the control of the Trader. Of course a huge amount of the day to day work of a dealing room can be handled electronically but the market is also heavily influenced by events happening in the real world. Our applications allow traders to take account of these occurrences. Traders can also enter a series of what ifs to customise their trading activity to deal with any eventuality. 

The boys are back in town

Do you think there are any significant knowledge transfer lessons that FX can learn from the continuing development of algorithms in other markets?

HT: Undoubtedly. However where we are presently is in the information gathering phase. I think it is true to say that the principal characteristic of those people who develop trading strategies within a specific asset class should be a deep knowledge of that asset class. While there are no doubt common characteristics within asset classes specifying them is a real intellectual exercise. 

Do you see the evolution of more advanced FX trading strategies impacting on the future development and implementation of FX algorithms?

HT: Yes the two are interlinked. But FX algorithms are not new. I had some discussions around automated execution over the Cognotec platform based on historic analysis of market data in the mid 1990’s. The guys concerned were very bright and had access to an array of Crays at the ZIT. It didn’t come to anything but there was activity in the space even back then. Where it eventually leads to I don’t know but FX is not a totally predictable market whatever anyone says. It needs real people with real brains and emotions. Right now we are trying to help those people improve their performance.  

Looking ahead where will leading technology providers, like Beta-Gamma be looking to extend the functionality of algorithmic FX trading toolkits to take strategy execution to new levels?

PT: Beta-Gamma is active in quantitative research in areas such as risk and directional forecasting.  We employ this research both to developing new and enhancing our existing execution strategies, and in providing quantitative tools as plug-ins to our applications.  

Our work on simple multi-language APIs is intended to allow clients to easily integrate their own algorithmic strategies into our applications.  In addition, we are researching user-interface enhancements to allow manual traders to more easily exploit the opportunities in the aggregated market.