Bowmoor Capital CIO, Gareth Abbot, has developed a distinguished trend following CTA strategy that has been managed since 2013 (and an earlier version of it dates back to 2006). The Global Alpha strategy has been recognized by The Hedge Fund Journal’s CTA and Discretionary Trader Awards 2023, receiving the award for the Best Performing Fund over 3, 4 and 5 years in the Trend Following (AUM < $50 million) category. It has annualized at 26.84% (gross) over the past 10 years.
Abbot, a mathematician and Fellow of The Institute of Mathematics and its Applications (FIMA), was fascinated by maths from an early age, and still agonises over the riddle of the odd perfect number, and views, “Beating the market with a system entirely based on mathematics as one of the greatest challenges for mathematics”.
Abbot has worked in the markets for more than 20 years, helping Duncan Duckett set up a derivatives desk in 2000, and was at Harmonic Capital Partners during its early years of growth.
Beating the market with a system entirely based on mathematics as one of the greatest challenges for mathematics.
Gareth Abbott, Chief Investment Officer, Bowmoor Capital, London
Bowmoor Capital is led by CEO, Richard Liddle, a former RAF pilot, who has served in Iraq, Libya, Syria and Afghanistan, and says, “I am bringing elements of military discipline and risk management to the business”. Liddle previously was part of a team who raised over $500 million for a real estate fund.
COO Brendan Mulvany, whose home near Bowmoor Lake in the Cotswolds provided the inspiration for the firm’s name, brings trading savvy and much more to the table. He has spent all of his career trading derivatives, futures and options, including at Dresdner Bank and Bank of America. He traded as a “local” in the German Bund pit on the LIFFE exchange for 8 years. After that he setup and ran highly successful trading desks at Refco, Kyte Futures, GHF Futures and Marex. This also involved high frequency trading, with programmers in Singapore
The performance objective is to double capital (gross of fees) every three to five years, which implies annualized returns of between 14% and 23% (based on the rule of 70). Both the real money track record, and simulations reaching back to the 1970s and 1980s, show that this objective would have been achieved, on average. Profits frequently outweigh losses, thus performance has positive “mathematical expectancy”. Additionally, the Sharpe ratio has exceeded one and the Sortino ratio has been higher still at over 3, due to the downside volatility being dwarfed by the upside volatility.
The return objective was also met during the QE era of central bank market manipulation that wrong-footed many trend followers. Though the fund was down 6.71% in 2018, a year bereft of trends, it had several other very strong years. “We did have some short-term underperformance over 1-3 year periods, but overall, we recovered, and had some spectacular years. The best was up 72% in 2014, and up 49% in 2017,” says Abbot, who judges that, “The post GFC market climate did diminish opportunities, but now we are back to more normal markets, which mean non-normal fat tail price distribution in statistical terms”.
Bowmoor pays more attention to maximum drawdowns than Value at Risk (VaR), though VaR is published in its reporting. Abbot fully expects one drawdown around 20% in every 3-to-5-year macro cycle as part of the strategy.
In the CTA world, Abbot admires Dunn for their consistent commitment to trend following over 50 years, and has also drawn inspiration from others such as Ed Seykota, Jerry Parker, Michael Covel, Perry Kaufman and Richard Dennis but also reaches back further for deeper academic foundations to the likes of David Ricardo and Charles Dow.
The performance cycle aligns with the over 100-year-old Dow theory, which identifies bigger picture macroeconomic trends in economies, macro and government policy of 3-5 years as being key drivers of markets. “There are always fundamental forces, behavioural biases, human nature and government policy driving markets. Charles MacKay’s Madness of Crowds and Popular Delusions never ages,” says Abbot. Though Global Alpha does not use fundamental or economic data inputs, it can be seen as “systematic macro” to the extent that its models capture major economic trends.
Though Bowmoor’s models only use technical price data, they are not super simple. The models, and various filters applied to them, can be quite intricate in some areas, though they are very much a “glass box” rather than a “black box”.
Abbot’s approach to model building is to, “Capture the overall structure of trends, which adapt and change, and avoid over-fitting or over-optimizing”. For instance, he dislikes some price derivatives – such as oscillators or MACD histograms – that are too far removed from the price action.
There are four trend signals, each applied over 4 timeframes, which works out at 16 models. “The primary trend, lasting at least a year, is the cornerstone and most important one, following macroeconomic and economic cycles. The secondary trend, over semi-annual or quarterly periods, also has some economic influences. The shortest trend is the minor trend, usually measured over a month, but ranging between a week and four weeks,” explains Abbot.
A smart pyramiding system allows the minor trend to either augment, or reduce, the primary trend, but the position will never trade against the trend. A minor trend recovery can rebuild a position that had been scaled back based on primary trend.
Abbot’s lookback periods for measuring trends do not go below a week. He finds that, “Most daily and intraday price action can be “noise”. We see inferior risk/reward in shorter term trading since transaction costs are a larger percentage of the moves”. A recent research project examined the value of price data during shortened holiday trading sessions., which tend to have lower volumes and liquidity.
The models are designed to enhance the natural positive skew of trend following. The four trend measures include well known moving average indicators such as EMA, swing price channels, Donchian price channels, and slope strength.
Abbot has strong views on the logical and intellectual consistency and coherence of models. He argues that, “If a model enters upon a moving average signal, it should also exit upon the same signal. In this context, exiting on a trailing stop, or a volatility level, is arbitrary. Fewer degrees of freedom are better to take less liberty around market strategy rules”.
Abbot admits that his trend indicators are not proprietary, rather the recipe for reading and combining them is: “A trading manual defines trading rules for portfolio allocation, market strategy, and risk management, and this is where the advanced proprietary maths comes into play”. Position sizes can be adjusted daily, based on risk, signal strength, volatility and correlation.
Average holding periods have been stable around 2-4 months, with profitable trades held for longer and losing trades cut. The average timeframe has not in fact grown longer during periods of stronger trends. “The reason is that there is always a balance between trending and non-trending markets and this ratio has been stable,” says Abbot. The stability of trend prevalence reflects both the models and the investment universe.
The current suite of models has traded a universe of 21 markets since 2013; between 2006 and 2013, a slightly simpler version of the strategy was applied to only 13 markets, in various vehicles and structures including joint accounts and an investment club.
The same models apply to all markets. The investment universe is determined based on an acronym that is also used to structure Bowmoor documentation, including an extensive DDQ: MECE (Mutually Exclusive and Collectively Exhaustive). “Every market needs to be different and uncorrelated, to achieve diversification with minimum correlation and without overlap. We cover 8 sectors of the global economy with minimal overlap. We only need one good position to make a good return, and that can be lost in hundreds of markets,” says Abbot.
Some 21 markets listed on CME Group and Eurex have been traded since 2013. They cover livestock, soft commodities, precious and base metals, energy, currencies, short term and long-term interest rates and equity indices. Bowmoor also monitor the largest 60 markets by volume.
Bowmoor would only cease trading a market if its liquidity dropped below acceptable levels. Abbot recognizes that markets such as lumber or cryptocurrencies can generate powerful trends but has concerns about liquidity and scalability: “We want to be able to flatten the book and go to 100% cash if necessary so we can provide daily liquidity”.
He also points out that the program can anyway generate strong returns from the existing markets it trades. For instance, in May 2023, both long cattle and short hogs were substantial winners.
Bowmoor is often exposed to all 21 markets but typically has positions in at least 18 of them with up to 3 switched off, though the lowest recently observed exposure has been to only 15 of the 21. “Markets are filtered out based on extreme volatility overshoots, which would need to be at least 3 sigma events, and this is after top slicing when the trend goes parabolic. Conversely, markets can also be turned off when a fractal efficiency filter finds the price action is trapped in a choppy range,” explains Abbot.
Correlation heat control uses proprietary measures introduced in 2013. “They work better than standard covariance,” says Abbot. The starting point is to assume that markets within each sector are correlated, and risk would not be increased even if correlations dropped. The objective is rather to find surprising correlations such as Euro Bund and cocoa, as an extra layer of risk control.
There are other filters as well. Abbot agrees with the common rule of thumb that a trend follower might capture one third of a move, but in some cases his filters have done much better and pinpointed certain moves between troughs and peaks.
Mulvany manages operations, systems, execution and volumes, based on his experience of technical analysis. Software and a suite of algorithms are used for execution and more will be added as assets grow. Mulvany aims to obtain better trade execution than some automated systems. We use Trading Technologies software for trade execution and for post trade allocations.
Britannia Global Markets in London is used as the current prime broker. The Guernsey fund will also add R.J. O’Brien as execution broker. EY are auditors and Northern Trust are administrator and custodian.
Panima Capital Management, who the Bowmoor team know well, are used for regulatory hosting, and some other support functions such as compliance monitoring, record keeping and best execution reporting.
Strategic hires have resulted in an impressive and experienced distribution team. Bowmoor’s Head of Distribution brings over 25 years of experience, with a particular focus on fundraising from family offices and institutional markets. The firm has also recently hired a Head of Investor Relations who has held senior hedge fund roles across a wide range of strategies, including as Head of Marketing at BH-DG Systematic Trading, and Head of Business Development at Kirkoswald Capital Partners.
Hiring Philip Salvesen, formerly of Withers LLP and Kingsley Napley LLP, as general counsel has expedited the setup of a Guernsey fund, which will offer a founders’ share class charging 0.75 and 15 or 0 and 20, versus normal fees of 1 and 20 and 0 and 30.
The Guernsey structure launched on 1 August 2023 is being seeded by the managers and London-based discretionary manager, Peregrine and Black, which is allocating a portion of its alternatives portfolio to the strategy.