Back to the Beginning

Sabre returns to its roots with the launch of a new CTA fund

Originally published in the September 2009 issue

The returns and investment strategies of hedge fund firms often run through cycles reflecting economic conditions and the comings and goings of portfolio managers and partners. It is unlikely, however, that many firms can match the evolution and change that has occurred at Sabre Fund Management over its 27-year life span.

The company was originally founded in 1982 as a commodity trading adviser (CTA) but, under a new generation of principals and portfolio managers, since 1997 it has focused on the design, management and trading of quantitative equity strategies. As one of the early pioneers of London based managed futures funds, Sabre has returned to its roots with the July launch of a quantitative CTA strategy. “We long wanted to diversify the business with a non-correlated asset class,” says Melissa Hill, Sabre’s managing principal in an interview at the firm’s offices located just around the corner from Buckingham Palace.

“We had been looking for a first class team to come and join us,” she says by way of introducing senior portfolio manager Alex Guillaume and portfolio manager Dobromir Tzotchev, both of whom joined Sabre in September 2008. Though the fund only launched this summer, the strategy ran successfully from 1st January 2006 within Peloton, the now closed multi-strategy firm founded by Geoffrey Grant and Ron Beller. From then until 31st August (including a period with no assets under management) the Adaptive Trading Fund made annualised returns of 10.3%.


Notwithstanding the difficult climate for new fund launches, the newly christened Sabre Adaptive Trading Fund has attracted firm investor interest. Hill attributes this to the fund not being a true start-up. “Alex and Dobromir have managed sizeable amounts of money before,” she says, noting the former’s strong pedigree from his experience at Aspect Capital, an established London CTA, and Société Générale. “It does seem that investor appetite is now returning,” Hill says, adding that a well known fund of hedge funds has invested with a 12-month lock up. She adds: “Because we are a quant firm and the existing team’s experience dovetails very well with the new team we are finding that our existing clients are interested in the futures strategy.”

The firm runs two other funds. Both of these are based on the Style Arbitrage, systematic equity market neutral methodology, developed by chief investment officer Dan Jelicic. The Anaxis Sabre Style Arbitrage Fund is a white-label fund distributed by Anaxis Capital. It features in the Credit Suisse Tremont data as a top-10 performer with year-to-date returns of 34.36%. The flagship Sabre Style Arbitrage Fund, with the lower risk target of 6%-7% per annum volatility, is also up a healthy 24.92% for the eight months to 31st August.

The current business mix reflects changes of focus made during the long track record of the firm. In its first CTA phase the firm attracted some of London’s top talent including David Beach and David Harding, both of whom went on to set up their own successful firms. When original co-founding partner Peter Swete left, Colin Barrow of ED&F Man, the precursor of industry giant Man Group plc, was brought in by co-founder Robin Edwards to share the business risk and to assist with a new operating plan. In 1996, Barrow, by then chairman, recruited Hill, who had been looking to join a new business venture, to head up investor relations and marketing.

“The business plan changed to sponsoring and developing new managers, as the remaining original founding partner wished to step back from managing client funds” she says. “The plan wasn’t intended to specifically focus on quant, itjust happened that the first strategy we brought in-house, the Sabre Market Neutral Fund, was based on a statistical arbitrage methodology – a quantitative equity market neutral fund. It was such a success that we concentrated on building the team in this area from that point onward.”

The success of the business with quant funds led to another restructuring and redistribution of equity in 2003 with Edwards exiting his day-to-day relationship with the firm in order to develop a new business and Hill taking over the management. Then in 2005 Hill led a management buyout and with CIO Jelicic, bought the majority equity stake from Barrow to add to their existing holdings. “It is a very different business from the one that was started in 1982,” Hill says. “We have a new generation of highly talented staff, most with PhD or Masters level education in complementary science and maths disciplines. We have become a research and development led firm, not a marketing focused business.”

Research focus

The core principle of the portfolio managers for Sabre Adaptive Trading is to focus on research and development in order to improve the quality of returns. This involves fine-tuning existing strategies as well as developing new ones in order to diversify the sources of alpha. Sabre Adaptive currently trades seven systematic strategies with truly diversifying profit drivers such as trend following, mean-reversion, pattern recognition and market comparisons. The target returns are cash plus 8%-14% per annum with annualised volatility of 12%.

The fund trades seven strategies. There are four directional strategies including adaptive, centroid, edge trigger and level trigger. The three others are spreads strategy, interest rate driven FX and dynamic FX carry. Several other strategies are also in development. They include a sector-adjusted equity valuation model, a dynamic carry model on emerging market currencies, a commodity/FX strategy and a crude oil trading model using the intra-day correlation with equity markets and exchange rates. “These classic trend following strategies – level trigger and edge trigger which use breakout and momentum filters respectively – work well in the long run,” says Guillaume. “We believe they are quite similar to what has been traded at large CTAs like Winton, AHL or Aspect, at least in the early years. We think we have made some enhancements on the standard trend-following models in the way we scale positions.” The typical methodology is to derive a combined signal from the underlying ones produced at each different frequency. Trading frequencies typically vary between a few days and a year. The overall signal is then scaled into a position by using a short-term volatility estimate. This volatility estimate is based on a proprietary model that Guillaume and Tzotchev developed at Peloton. They also use long term VaR capping to avoid taking too large a position when the signal is strong and the volatility estimate is low – something that occurred with equities during 2006-07. In such an environment with strong positive signals and low volatility estimates, the fund used a long term VaR estimate to cap its positions in these markets.

The adaptive strategy shows how the managers have innovated in an ever changing market. “This strategy is not assuming that the behaviour of the market will be trend following,” says Guillaume. “What you see in managed futures is that most funds assume a trend-following behaviour but they find that the short-term frequencies do not add a lot of value. This is true between, say, one and 10 days. So some funds decide to shun these strategies and not allocate to these frequencies. Our approach is agnostic. We do not impose a trend-following behaviour. We let this be dictated by the behaviour of the market over the recent past. We developed an adaptive model, which adjusts dynamically, the weighting of the different trading engines (or frequencies).”

If a market has been consistently mean reverting at a particular frequency, the model will have negative weight on that trading engine. This means that the model may capture mean reversion in the short term if it is consistent enough. “Over the medium and long term frequencies, we are typically trend following but again with an adjustment being made daily to the weighting that we have on these different frequencies,” says Guillaume. “So we have only one additional coefficient in this model, which is how fast we learn the behaviour of a market,” he says. “These are typically medium term coefficients (ranging from one to three years depending on the asset class). This is the parameter we optimise per asset class. So with equities we see a lot of mean reversion in the short-term frequencies. Another important feature is that it has a continuous position function and adjusts to the behaviour of the market. This is important as it eliminates unnecessary jumps in the target position.”

Centroid strategy is most complex

Among the directional strategies, the centroid one is probably the most complex and also the least transparent. It employs a combination of statistical methods such as principal component analysis and stepwise regressions to select the price filter which has the most relevance on predicting the next day return. A multiple regression produces an estimate of the next-day return for a given market. These return forecasts are then used to build an efficient portfolio with a specific VaR constraint. This is based on a rank based portfolio optimisation method which uses the properties of centroids (hence the name of the strategy). “We control the overall risk of the strategy,” Guillaume says. “In that sense, this strategy is less directional that the trend-following strategies. This is because we control the overall VaR of the portfolio. It allows the fund to take a long position in one equity market and take a short offsetting one in another.” Altogether, the four directional strategies comprise approximately 65% of the portfolio.

The fund’s other strategies compare one market to another. The spreads strategy focuses on equities, fixed income and commodities. A spread market is made of two correlated markets. The spread is adjusted for market risk by using a dynamic beta adjustment. The adaptive strategy is then utilised on the spread market. Typically, spreads tend to be highly mean-reverting in the short-term frequencies because these markets are quite highly correlated – for example the CAC 40 and Euro Stoxx 50 are highly correlated even though they may not move exactly in line. From this, the fund seeks to capture the short-term discrepancies that tend to revert. In the longer frequencies, there might be some trends which can be captured by the adaptive model. This allowed the fund, for example, to capture the outperformance of the DAX in 2008 as it drew resilience from its exposure to China. “This strategy is very consistent,” says Guillaume. “The only downside is that its capacity is not as high as for other strategies. The residual risk is small compared to the notional exposure because the spread has a much lower volatility than the two legs. You can’t scale that strategy as much as you would be able to on directional strategy. Because we are relatively small to start with, it is a good strategy for us to have. It hasn’t been a very good period for CTAs since we launched in July, but we outperformed our peers thanks to the spread strategy which made money in both months.” The fund also trades some interest rate spreads including eurodollars versus euribors and the two, five and 10 year bond futures in Europe versus their US counterparts. A total of 16 spreads are traded across equities, fixed income and commodities.

The last two strategies focus on FX markets. The interest rate driven FX is another market comparison model which uses information from the bond markets to trade FX. “The idea is that bond markets are a little more efficient in predicting the moves of central banks,” says Guillaume. “FX markets are probably a bit less efficient because you have a lot of non-trading participants”. The strategy takes a signal derived from changes in swap rates differentials (using maturities between two and 10 years) and measures it over relatively short periods of time. This signal is then scaled into a position using the same methodology of volatility scaling and long term VaR capping. The fund applies this model over 12 major currency pairs.

The last strategy is dynamic carry. “FX carry tends to work in the long run but it faces large drawdowns during period of high risk-aversion,” says Guillaume. “In a sense, carry can be viewed as the risk premium earned for taking this risk,” he says. “The returns earned over two or three years, can be suddenly vaporised by such drawdowns. Our idea is to reduce our exposure to carry during these periods. We try to do this by building a risk model that is very broadly based and looks at a number of factors across different risk dimensions.” The portfolio managers look at FX implied volatility, equity implied volatility, swap spreads, corporate spreads and emerging market spreads. They also consider risk factors for, say, the CHF versus a basket of currencies or the performance of equities versus bonds. “It tells you about which market regime we are in, in terms of risk aversion,” Guillaume explains. “It is not the level of risk that matters for us but the change in risk aversion. If we have a sharp increase in risk aversion over a short period, this is where the model will get cautious and reduce its exposure to carry.” The portfolio managers calculate that dynamic scaling gives a 20% improvement in the risk/return profile over a static carry strategy.

Strategies started at Peloton
The adaptive, level trigger and edge trigger strategies were started at Peloton in January 2006 with the others added subsequently. At Peloton, Guillaume and Tzotchev were one of many trading teams in a diversified multi-strategy fund. The fund traded fixed income, FX, equities, convertibles, asset backed securities and emerging markets with each team being allocated a VaR number defined by the risk manager. “When we joined Sabre we came with all the IP that we had built at Peloton,” says Guillaume. There they had six front office quantitative traders and six developers. Altogether, 15 researchers worked on some aspect of the system that is in place at Sabre. “As Peloton disappeared completely, we were able to bring with us the code which was built over a period of two and a half years’ of development work,” he says.

As a pocket of Peloton’s multi-strategy fund, the systematic team managed up to $200 million based on the VaR allocation. “Peloton was really managed like a proprietary trading firm where each desk or trader was allocated a VaR budget,” says Guillaume. “This VaR budget was generally decided on the basis of past success” he says. “If a trader was successful the VaR was increased, if a trader started to lose money the VaR would be decreased. It was probably not the best set-up for systematic trading. With systematic trading you don’t want to have an allocation overlay which basically redeems from your strategy after a drawdown. Typically these are periods where a strategy will outperform. It worked the other way around as well. When we were doing well, our nominal capital was increased substantially.”

The inevitable question is what did the former Peloton portfolio managers learn from their experience in a failed business? And, moreover, how is that applicable now and what has it done to make the Sabre offering a better fund than it might be otherwise? Guillaume cites several positives from the Peloton experience. “The infrastructure at Peloton was very good,” he says, adding that his first role there was to build the quant and risk systems for the investment manager. “We were able to trade seamlessly in all types of instruments in all asset classes. Also, for a systematic trader, it was very beneficial to trade next to prop traders in fixed income and FX and emerging markets. They gave us a number of ideas to develop new strategies. For example, our interest rate/FX strategy is a systematic version of one that an FX trader was employing.”

In retrospect, Guillaume says one of the reasons Peloton failed was because it expanded too fast into other strategies beyond core strategy run by the founding group of former Goldman Sachs macro traders. “They hired some good traders but they probably did not understand their strategies well enough,” he says. “That is one lesson: don’t expand too fast in strategies you don’t completely understand, especially in terms of risk management.”

For good reason, Guillaume and Tzotchev don’t feel they are in a start-up environment. Now that the fund is trading, the next challenge is to continue raising capital. Here the pedigree of Sabre and its long experience in the quantitative sector is bound to provide a useful calling card. Sabre’s resources gave the new fund the staying power to launch a year later than expected at a time when investors are beginning to emerge with a modicum of appetite for risk.

“The most difficult thing is the fact that the industry just basically shut down from September until about the end of the first quarter of this year, while everybody was trying to assess what was going on in the hedge fund space and who would stand a chance of making money going forward,” says Hill. “Obviously there is a significant degree of risk aversion that is present. Obtaining money for strategies is much harder than it has ever been. But I think we stand a better chance than a lot of people because of Alex and Dobromir’s reputation, because they are established managers and because we can show their live track record from Peloton. Along with Sabre being a well known name in the quant space, I think the whole package has come together.”