John Howard: Perhaps we could start with some background. You ran a very successful discretionary desk with, I think, a strong global macro element for Mizuho for many years. What prompted your move to become an exclusively systematic manager?
Philip Stoltzfus: It was an evolutionary process. Prior to 1998 I had really not been involved much with systematic trading; the group I ran at IBJ and later Mizuho was primarily a fixed income relative value business with, as you say, a global macro component.
We had around 12 traders supported by programmers and analysts. During the Asian crisis and rouble devaluation I finally became convinced of a characteristic of our RV trading that I had long suspected - namely, that while most of the time the returns of our traders were non-correlated, and did not have a particularly high beta to the fixed income markets, we seemed to have a high vulnerability to regime shifts in volatility. As a result, the risk of a simultaneous draw-down for all of our strategies when volatility increased was very high.
I therefore decided that I needed to find a strategy that would be a genuine diversifier, and that would do well in raised volatility conditions. I had recently hired Scott Ganis, who would become co-founder of Thayer Brook Partners, to help me develop our risk management and performance measurement framework for IBJ's proprietary trading team. When I discussed the issue with Scott, he suggested we look at introducing a systematic momentum capture strategy, because the P&L profile would be positively convex, and therefore would benefit from higher volatility environments.
Scott had extensive experience in model design. He showed me a trend following model that he had developed, and we concluded that while it satisfied some of our criteria, the risk/reward needed improvement - the draw-downs following trend reversals were too severe. We spent around a year and a half developing a model that could compensate for this problem in a robust and parsimonious way, and once we had done that, we introduced the program in April 2000.
John Howard: And that led you to set up Thayer Brook - how?
Philip Stoltzfus: Over the next few years I concluded that the strategy performed its diversifying role exactly as advertised, and that the risk/reward over the long run was superior to most discretionary strategies for a number of reasons - the discipline of the strategy, the favourable relationship to high volatility, the indifference to consensus, and the transparency of pricing and valuation.
I felt that our systematic fixed income strategy could perform the same useful diversification role for other investors in the market, and decided therefore to create Thayer Brook Partners and launch our first product, The Thayer Brook Fund. We also knew that the models that we used would be equally effective on other asset classes, so we developed and eventually launched a second product, The Thayer Brook Global Diversified Fund, which also trades stock index, energy and commodity futures in addition to FX, Short Term Interest Rate and Fixed Income futures.
In all cases our goal is to create an absolute return, a convex return stream relative to the underlying instruments traded, and a favourable relationship to volatility.
Thayer Brook Partners
Thayer Brook Partners was founded by Philip Stoltzfus (CEO) and Scott Ganis (COO) in October 2005. The firm now manages $136 million in assets including the investments of two separate funds - The Thayer Brook Fund Limited and the Thayer Brook Global Diversified Fund Limited.
The former was the firm's first fund and applies a portfolio of trend capture and risk management models to fixed income, currency and short term interest rate futures. It is an absolute return strategy seeking to profit from significant moves in the global government bond and foreign exchange markets and has returned 17.6% in the past three years. Similarly, The Thayer Brook Global Diversified Fund seeks to benefit from significant moves in the markets it trades, and adds commodities, energy and stock index futures to those traded by The Thayer Brook Fund. The global diversified strategy was launched in April of 2009 with a three year track record showing a 44.0% return. The strategy currently has $26 million in assets invested in it.
At first glance Thayer Brook Partners might seem unlikely candidates for Me and My Machine. They don't generate thousands of orders per second, and their trading strategy isn't dependent on them constantly finding ever lower latency routes to the exchange. The compelling factors that interested us about Thayer Brook start with the decision by one of the firm's founders to spurn the comfort zone of his successful discretionary trading background for an entirely systematic approach.
Having started out with a fixed income focus, Thayer Brook's use of systematic risk management models within their core trading strategy has enabled them to trade successfully across multiple asset classes - something that many systematic managers set out to achieve, but somehow get lost along the way.
John Howard: Would it be fair to say that you're now biased against discretionary strategies?
Philip Stoltzfus: Not at all. I continue to follow the RV world relatively closely, and would not rule out the addition of a discretionary RV strategy to the Thayer Brook platform of products in the future. However, our plate is pretty full at the moment.
John Howard: Which instruments do you most actively trade?
Philip Stoltzfus: The Thayer Brook Fund is a fixed income, short term interest rate and FX strategy. It trades US, European and Japanese government bond futures, US and European short term interest rate contracts, and the Euro, Yen, Sterling, Swiss, and Australian $ FX contracts. The Global Diversified Fund trades all these plus US, European, and Japanese stock indices, oil and natural gas, grains, meats, metals and softs.
John Howard: And how many models do you currently run?
Philip Stoltzfus: We currently have three model sets, each of which is designed to perform a fundamentally different role within the portfolio, ranging from momentum capture to leverage management to diversifying position entry over time.
John Howard: Describe the philosophies that underpin these models.
Philip Stoltzfus:: The principal objectives are that we want to deliver an absolute return and a convex return stream relative to the underlying instruments traded, and we want to capture significant market moves for our investors with good risk reward. Those are fairly straightforward goals - doing it is another matter.
Scott Ganis: We, of course, want to rigorously guard against data mining and curve fitting; we are parsimonious in our model construction and the combination of models within the portfolio. We are generally believers that each model has to bring something fundamentally different to the table in terms of performance, and should not simply be included because of an apparent low correlation to other models.
We believe that models have to be dynamic enough to capture changing market behaviour, but robust enough to do so without being burdened with preconceptions about how individual markets work.
John Howard: Are the models purely price driven or do you also incorporate fundamental inputs?
Philip Stoltzfus: At the moment we focus on price and volatility, but we are agnostic on such matters - we are always on the look out for methods that meet our design and test criteria, and don't like to pre-judge what is likely to work.
John Howard: What is the typical lifecycle of one of your trades?
Scott Ganis: Investors tell us that according to their definitions the median duration of positions in our portfolio is either on the short end of medium term or the long end of short term. Generally speaking we are trading 1 to 2 times per day in each of the instruments we trade, so our volume is relatively low.
John Howard: How do these orders find their way to the market? Which OMS/EMS do you use?
Scott Ganis: All of our systems, including order management systems etc., are designed internally. We prefer to have total knowledge and control over these processes.
John Howard: Aside from the decision of whether or not to deploy a model into production, can you think of any way that discretion might enter the trading process - for example, in the deployment or withdrawal of individual models based on a fundamental view, the selection of parameters, et cetera.
"If you can't follow the discipline of systematic trading and let the models do what they are designed to do, you should try doing something else for a living."
Philip Stoltzfus: No, we don't use fundamental overlays or anything like that. Obviously we have a fiduciary responsibility to our investors, and are vigilant in terms of monitoring liquidity and market access. If an instrument were to become obsolete we might make a fundamental decision to remove it from the portfolio. Otherwise, no.
John Howard: How often do you find yourself wanting to flatten a core position or perhaps finesse an order because it conflicts with a personal view?
Philip Stoltzfus: Well, you don't really want to go there - that is the slippery slope. My background is as a discretionary trader, and of course there are times when I am not happy with a position that the models are taking. Sometimes I am right, and sometimes I am wrong. If you can't follow the discipline of systematic trading and let the models do what they are designed to do, you should try doing something else for a living.
John Howard: Describe the structural and macro changes you've seen in your markets in recent years and the changes you expect to see moving forward.
Philip Stoltzfus: This is really too big a topic! I have been trading since 1986, and there have been innumerable changes during that time, whether in fixed income with the dramatic increase in credit investment and in the use of derivatives, or in the tremendous widening of the range of investors wishing to include commodities within their portfolios, or in the huge advances in technology resulting in, among other things, the nearly exclusive use of electronic execution platforms, or in the abandonment by governments of even the pretence of laissez faire monetary and fiscal policy in favour of a highly interventionist style.
To be honest, these changes create certain technical challenges, but do not impact the fundamental validity of the method, as I think can be seen by the consistency of the returns profile of the last several years with what has occurred in the past.
John Howard: You spoke earlier about volatility. Let's talk about that a bit more.
Philip Stoltzfus: Volatility is a key performance driver, so regime changes in volatility are important to us. In 2005 and 2006, volatility sunk to generationally low levels, for reasons that I think were partly understood at the time and partly understood only in retrospect - the latter stages of a secular fall in inflation, driven in large part by the introduction of the Chinese workforce to the global economy, the use of derivatives to deliver yields on risky assets to groups of investors that had traditionally focused on more conventional products, and endless liquidity made available by central bankers who were willing to operate relatively accommodative policies given the seemingly friendly inflationary back-drop.
Investors became comfortable with risks that would have seemed unacceptable before, not because they were irrational, but because the volatility of the investments had dropped and therefore seemed less risky. That volatility compression was a self-reinforcing process, and ended by the comprehensive blow-up in risk assets in 2007 and 2008. Needless to say, the ultra-low volatility environment of 2005 and 2006 was ideal for the systematic approach, and the higher volatility environment of 2007 and 2008 was a lot better.
John Howard: And since then?
Philip Stoltzfus: In 2009, we had a situation where risk assets were readjusting from extremes set in Q4 2008 and Q1 2009, and a US fixed income market that could neither rally materially because of the evident recovery of the global economy, nor sell off substantially because of the active government support - both through promises of continued low rates and direct action such as quantitative easing.
We therefore ended up in an environment of high daily volatility but low month to month volatility, exactly the opposite of what we want to see in the managed futures space. However, I don't see this as a return to the 2005 and 2006 low volatility consensus, but rather a consolidative phase. Many of the risk elements that were in play in 2007 and 2008 are still there, but simply shifted to other participants. In particular I am thinking of the transfer of leverage from the banking sector to the public sector.
John Howard: So how do you cater for these paradigm shifts in volatility when it comes to creating models and risk managing them? Describe the creative process.....
Philip Stoltzfus: The design process is generally driven by a trading objective or requirement.
The original objective, as I have mentioned, was to develop a strategy that would cope with regime shifts in volatility in the fixed income space - once you have a goal, and have an idea as to how that goal should be accomplished, the other steps fall into place. Once you are operating a methodology, the design process is oriented around enhancing what you already have, either through system refinements or trying to identify new methods that are fundamentally different.
New ideas are generated by being absorbed in the process, just as in any line of work, and then augmented by what we observe in the market, what we read, the insights of the young PhD's we employ who bring ideas from their own research backgrounds, and of course, the periodic bolt from the sky.
John Howard: What percentage of ideas make it from conception to production?
Philip Stoltzfus: The hurdle rate to be included in the strategy is high, because the strategy works already. There are plenty of ideas that don't make the cut, not necessarily because they don't work, but because they are not clearly different from what we are already doing. We generally are pretty successful in terms of achieving our development goals, whether in terms of creating a model to fill a perceived gap, or in major projects such as the Global Diversified Fund.
John Howard: We talk to other systematic managers, and they nearly all outline the ease with which a manager can fall into the curve fitting trap - whether by fitting model to timeframe, the selection of parameters, weightings for capital allocations to different models, etc etc. What steps do you take to avoid curve fitting?
Philip Stoltzfus: This is obviously the biggest risk for the strategy - the virtue of systematic trading is that it lends itself to testing, but the testing process is always vulnerable to curve fitting. The first requirement is to be vigilant, of course. It helps if you have extensive market experience, because people who have successfully traded the market know that there are no answers, only methods.
A good understanding of statistics is essential - you have to understand the limits of the data you are using. There are standard techniques using blind periods and so on, and our own proprietary methods. It is also important to understand that if you are applying technology in a way that would not have been possible in the past, you need to handicap that historical data.
John Howard: Once in production, how far does a model have to fall outside the window of expectation to be withdrawn?
Philip Stoltzfus: The issue here is that we have expectations of how a model should work based on the performance drivers on which it depends. If those performance drivers are absent, that does not mean we should abandon the model. On the other hand, if the performance drivers are present and the model does not respond to them, it would be a problem.
John Howard: So what do you feel you might do differently to other managers where risk management is concerned?
Philip Stoltzfus: We regard our risk management methodology as a key competitive advantage and strategy differentiator. Our initial use of managed futures was in the context of a portfolio of fixed income, short term interest rate, and foreign exchange futures, all of which were highly correlated. As a result, we did not have the crutch of instrument diversification to lean on, and we had to develop risk management methods to handle reversals in a portfolio where all instruments were headed in the same direction much of the time.
This stood us in good stead when we eventually developed the Global Diversified Fund with its broader portfolio of assets. In general, we think even quite primitive methods can have some success in capturing momentum; the trick is in hanging onto gains. Therefore, the primary job of a managed futures manager is to get his risk management right, and we think we add a lot of value here.
John Howard: Because you assess risk differently or use different risk metrics from other managers?
Scott Ganis: We don't want to be tied down to or limit ourselves to only a select set of risk metrics. For example, we have found analysis of the type shown here [see chart] to be very useful in assessing performance drivers and risk control. Almost any strategy can be examined in this way and the results are often quite interesting.
Here you can see an analysis we did that shows the impact of our risk control systems. The chart shows various portfolio rates of return for one of our strategies, in blue. The chart clearly demonstrates that volatility has a positive effect on both the size and number of winning periods, but that this effect is not symmetric - the number of losses and size of losses does not appear to rise in the same manner.
In fact, you can see that this is due to our risk control, because when we remove the risk control systems from the portfolio - these results are shown in small red points - the positive observations are not changed in any meaningful way, but the distribution of returns with respect to volatility is now much more symmetric - we see more losses of increasing magnitude without the risk controls.
The graph below plots returns against an index of the volatility for instruments traded in a portfolio. At low levels of monthly volatility, positive and negative months cancel each other out, but as volatility rises, the number of big positive monthly returns increases dramatically, both in absolute number and relative to negative months.
Y = Periodic Rates of Return X = Volatility of Underlying Markets (Indexed)
Philip Stoltzfus: We want each model to be fundamentally different, by which we mean, operating on the basis of different performance drivers. That really is the most important thing. We have a variety of statistical measures we look at, but we are not dogmatic - what metric you look at is somewhat determined by the individual problem you are trying to resolve.
Philip Stoltzfus, CEO, a graduate of Princeton University, is the former head of Mizuho London Branch's proprietary trading business, and has over 20 years' trading experience.
Philip developed the portfolio strategy used at Mizuho to combine the trading activity of multiple traders engaged in fixed income, derivatives and foreign exchange markets. He developed a managed futures strategy as a diversification technique in 1999, which evolved into the model portfolio method now used by Thayer Brook Partners.
Scott Ganis, COO, is a graduate of the University of Chicago, where he studied economics, statistics, and applied mathematics before going on to obtain an MBA from their Graduate School of Business.
Scott has over 25 years' experience in the field of trading model design. He joined Philip's proprietary trading department at Industrial Bank of Japan (later Mizuho) in 1997. There he designed and implemented the group's risk management framework, and spearheaded the research and development of the group's managed futures strategy.
Anthony Murray, Marketing Director, joined Thayer Brook Partners in March 2006. He has over 20 years' experience in the financial markets, most recently as Managing Director at JP Morgan London, where he was responsible for Fixed Income Futures and Options Sales in Europe.
At JP Morgan Anthony was also responsible for the global coordination of Futures and Options Sales to Hedge Funds. Anthony previously held similar positions at Morgan Stanley and Bear Stearns.
Thayer Brook Partners has a total of 7 staff.
John Howard: Do you have any plans to extend your activity to cash markets?
Philip Stoltzfus: We have already expanded into other asset classes with the launch of the Thayer Brook Global Diversified Fund. There is scope within that product to broaden the instruments traded, and of course we can add to the FX and fixed income instruments we trade within the Thayer Brook Fund if we see a reason to. Our main research focus however is on further model development, both in terms of refining existing models and introducing new ones that can deliver something genuinely different.
We don't need to substitute cash markets for the futures markets we are trading, as that unnecessarily complicates what is the cleanest and most transparent strategy in the market, managed futures. However, should we wish to trade a market that provided great diversification, low transaction costs, and transparent, verifiable valuations I would obviously consider it. It is not really a priority at the moment.
John Howard: Putting your discretionary hat on for a moment, what opportunities do you anticipate the market to offer going forward into 2010?
Philip Stoltzfus: Put it this way: we are currently seeing a lot of interest in our fixed income product, The Thayer Brook Fund, because investors are worried about what can happen to the government bond markets. The emergency measures that policy makers have put in place will be withdrawn at some point, with unpredictable consequences precisely because the measures themselves were unprecedented.
Government debt in Western countries has accumulated at a rapid rate, and is likely to continue to do so, and neither central banks nor politicians are likely to want to hike rates or rein in stimulus in the face of uncertainty and particularly, high unemployment. So the odds of a policy error have increased on the margin, and also the imbalance between growth rates in emerging economies and those in Western economies may result in conflicting policy objectives.
John Howard: Some would say that the world has become a smaller place. Has harmonisation of central bank policy across key markets made your job as a systematic trader that much harder?
Philip Stoltzfus: I don't see the world as a smaller place, but rather as a much bigger place, with huge new players such as China, Brazil, India and Russia - aside from opening up new markets and being massive investors themselves, these countries influence global economic outcomes in very important ways. We are many years past the day when US government bond investors could exclusively focus on US economic events or the Fed to decide when and what to buy.
As for harmonisation of policies, the synchronous decline in global economies is one reason why the recession was as sharp as it was, and the common policy response a reason behind the dramatic moves in assets over the last couple of years. So synchronous action has been a benefit to systematic traders in the fixed income space, as it has reinforced trends.
If by "harmonious" you are suggesting "successful" and therefore likely to result in low volatility conditions going forward, I would have to say that on the basis of recent experience we should have a healthy scepticism about the accuracy of those policy makers and whatever crystal ball they are using.
John Howard: Finally, what are your plans for 2010 and beyond?
Philip Stoltzfus: Our main focus at the moment - having launched the Global Diversified Fund in April 2009 - is on a short term, day trading model that we have been working on. We have a long research pipeline, but the hurdles to be included in a portfolio that already works are pretty high. We will keep you posted on those developments.
John Howard: Philip, Scott, thank you and best of luck.