This article was first published in The Hedge Fund Journal on 17th June 2025 here.
This article was first published in The Hedge Fund Journal on 17th June 2025 here.
Quantitative investing can occasionally deliver innovations that widen investor accessibility and unlock enhanced performance potential. CFM’s newly launched Cumulus program is just such an innovation, offering investors access to scalable strategies from the Stratus program, which is otherwise closed to new capital. Named in the house tradition initiated by founder Jean-Pierre Aguilar, who tragically passed away in a gliding accident in 2009, Cumulus represents an elegant synthesis of decades-long quant investing expertise, sophisticated data analysis, and adaptive risk management.
CFM’s award-winning flagship multi-strategy absolute return program Stratus (currently closed to new investments) is made up of hundreds of strategies, some pushed to capacity and some that are not. Raising assets further would upset that mix and affect the returns delivered. The more scalable signals, not being run at capacity, can now be accessed at attractive management fees (1%) and performance fees (15%) in a new multi-strategy vehicle, Cumulus, launched in January 2024. The program was seeded by external investors familiar with CFM, and current capacity is estimated at approximately USD 5 billion, which could add one third to CFM’s assets of USD 18.9 billion as of May 2025.
Jean-Pierre Aguilar, who features in The Hedge Fund Journal’s 50 Giants Across 5 Decades series, started the tradition of naming CFM programs after gliders, and Cumulus continues that tradition and represents the combined power of over three decades of experience in quant investing. It trades a subset of Stratus strategies, using the same teams, research process, portfolio construction and execution algorithms. It harnesses CFM’s scientific approach, cutting edge traditional and alternative data and advanced technology, which have been developed over more than 30 years in Discus, CFM’s oldest strategy run since 1991, and for over 20 years in Stratus.
Cumulus combines hundreds of strategies using pioneering techniques in portfolio construction and generating thousands of trades executed electronically by in-house-developed algorithms, minimizing execution costs. It also includes a defensive sleeve that was developed using insights into the convex properties of trend following and the Volatility Index (VIX) future.
Stratus trades strategies on a range of timescales and across a broad selection of assets including equities, fixed income, commodities, FX, credit and volatility executed through stocks, futures and options.
Cumulus accesses the scalable component of Stratus through daily strategies applied to futures and single stocks. Of approximately 300 equity market neutral models in Stratus, approximately half make their way algorithmically to Cumulus and of some 200 directional models in Stratus, around half are shared with Cumulus. “The directional parts of Cumulus are about 60% correlated to Stratus’ futures sleeve, and the equity market neutral strategy has a 60% correlation to Stratus’ equity market neutral Stat-Arb allocation. We project an overall correlation of approximately 50% between Cumulus and Stratus,” says Philip Seager, Head of Portfolio Strategy at CFM.
Cumulus also trades a small defensive sleeve that draws upon CFM’s research into convexity. The current Cumulus strategy split is 10% defensive, 45% directional and 45% equity market neutral. Of this over half is relative value: all 45% of the equity market neutral is relative value, as is approximately 20-25% of the 45% “directional” component.
Relative value trades could include going long one government bond market and short another in Europe, where there have been significant trends and reversals in recent years: in January 2025 Greek government debt traded at tighter spreads than French paper but traded at multiples of French yields a decade ago. Another example of a relative value trade could involve trading different mega-cap US tech stocks against each other, where Nvidia outperformed the others for most of 2024, but markedly lagged Broadcom Inc. in December 2024.
Cumulus’ defensive strategy draws on CFM’s research into a multi-angled approach to protecting against corrections and crises, which employs a short-term trend overlay and a long volatility strategy trading the very liquid CBOE future on the VIX index of implied US equity market volatility. These defensive strategies make different contributions to returns during various sorts of market environments, corrections and crises across multiple timeframes.
The defensive strategy has helped Cumulus to deliver attractive diversification properties. Remarkably, since January 2024 Cumulus has demonstrated a negative correlation to global equities even though both Cumulus and global equities have produced positive performance most of the time – because the pattern of performance has notably differed.
Beyond equities, Cumulus also aims to show low or no correlation to other asset-class betas such as interest rate duration, the USD or other common market factors. Cumulus can take long, short and relative value positions in asset classes and profit from bull, bear and rangebound markets.
Cumulus overall has no meaningful overlap with traditional quantitative styles and factors. For instance, new models added to Cumulus are only accepted if correlation to traditional trend following or equity factors is low.
Cumulus has seven broad strategy clusters that are designed to provide diversification: over time the different clusters will contribute to performance at different times and during different market regimes and environments.
However, Cumulus optimizes allocations in a much more granular way based on hundreds of individual models. Over 250 models are run in parallel on a daily basis and positions are adjusted according to models and changing market conditions. The average correlation across the 250+ models is low based on snapshots at any point in time, thanks to an eclectic mix of data and signal types.
Thematically, technical, fundamental, macroeconomic, sentiment and alternative data variables can all feed into signals and CFM describes the range of signals as “exhaustive”. The firm’s in-house data libraries and catalogues keep growing with terabytes of technical, fundamental and alternative data collected each day. The data budget has significantly increased over the past few years, with alternative data being sourced by a dedicated data scouting team. Data types can include news media, text, weather, satellite, and other unstructured data.
Philosophically, CFM is not wedded to a single machine learning paradigm. The manager uses a blend of techniques from linear regressions to deep learning and natural language processing to Large Language Models (LLMs). The choice in the application of such techniques depends on the context with complex, non-linear machine learning techniques more and more applied to high signal to noise environments such as execution and text analysis rather than traditional trading strategy development.
LLMs have become a key focus of research for CFM, enabled by migration to the cloud and vast GPU computing power. “We are now in a position to apply LLMs to the sizeable amount of text data we have collected over many years,” says Seager.
CFM has also been active in natural language processing (NLP) for many years, collecting text and extracting sentiment from it. Breakthroughs such as ChatGPT mean textual data can be analysed with the same scale and rigour as numerical data.
AI is only one example of ongoing evolution. CFM is constantly reinventing its models and signals and 70% of Cumulus risk is in models launched over the last five years. CFM has been adding predictors at an accelerating pace in recent years on the back of a significant increase in research spending, and many of the more scalable models naturally percolate into Cumulus.
Gargantuan innovative datasets, AI and LLMs are the most glamourous part of the process but CFM has always emphasised portfolio construction and execution as equally important elements of its process.
Optimization and portfolio construction in Cumulus are based on proprietary techniques. Futures allocations are optimized based on “Agnostic Risk Parity”, which CFM has published papers on, such as the 2016 paper Agnostic Risk Parity: Taming Known and Unknown-Unknowns, co-authored by Seager and six others at the firm. Out of sample Sharpe ratios have been improved by allocating somewhat more evenly across principal components.
Equity allocations are optimized using related technology. “The Principal Components Analysis (PCA) is a lens through which the portfolio research team can understand a set of positions. This diagonalization converts a portfolio positioned on instruments to a portfolio positioned on uncorrelated linear combinations of instruments. The different optimization schemes are better understood through this lens,” explains Seager, who has been with CFM for almost 25 years.
Execution and understanding transaction costs, mainly those due to market impact, are also topics that CFM has published extensively on for over two decades. CFM’s tests show that its proprietary algorithms have outperformed broker or vendor packages. Low latency, direct to broker and direct to exchange routes are used where appropriate. Cumulus trades at time scales typically longer than Stratus or Discus and as such is less sensitive to trading costs. Nevertheless, the program benefits immensely from the firm’s vast execution expertise.
In common with other CFM programs Cumulus has some leeway to vary volatility. The volatility target of the standard Cumulus program can range from 6-10% and CFM does have some degree of confidence in upsizing exposures to take advantage of a better overall opportunity set. “We do try to predict our own P&L strategy by strategy and allocate more to strategies when the environment is better. Overall, we aim to slightly beat a benchmark of equal allocations across strategies, even if we cannot forecast exactly where returns come from,” says Seager. A 1.5x leveraged version with volatility targets ranging between 9% and 15% is also available.
Stratus has a broader investment universe than some multi-strategy quant programs or typical CTAs. At the asset class level, it trades credit and volatility in addition to equities, fixed income, commodities and FX markets. The program can trade beyond the front month futures to optimize liquidity.
The universe of over 100 futures is the same as for Discus and Stratus with the exception of Chinese futures, which could be regarded as a foray into what might be dubbed as “alternative markets”. CFM has not embraced other “alternative markets” on the basis that their diversification benefits are not sufficient to compensate for concerns around liquidity, CFM arguing that less liquid assets are sensitive to autocorrelated asset flows in and out of these instruments, in particular in times of stressed market environments.
Within equities, CFM also trades a wide universe of single stocks: 14,000 stocks in 20 countries and 13 currency zones is towards the top end of the range for quantitative managers. It can include some small caps but no micro caps and relatively limited exposure to emerging or frontier markets, where market impact is greater. “We do not think we are necessarily trading a larger universe of equities compared with leading quantitative managers but do expect that we have achieved better cost control,” says Seager.
Cumulus can initially be accessed through offshore vehicles with monthly liquidity, though a UCITS vehicle may be launched with a partner soon and the program may also be accessible in 40 Act funds. “While Stratus has too many complex line items to contemplate a separately managed account, Cumulus could potentially cater for investors seeking such structures,” says Seager.
CFM has gone through a period of rapid growth with staff numbers up 30%, the data budget doubling, and the number of models tripling in five years, and the Cumulus launch underscores Seager’s strong confidence in continuing growth on all fronts: “We are continuing to grow and have made a big investment in infrastructure, to scale up the research process and continue to perform. We want to carry on investing in infrastructure and people and grow the business further”.
In a market with abundant generic quant offerings, Cumulus marries precision-driven alpha generation with thoughtful risk mitigation. By opening wider access to select proven strategies and harnessing cutting-edge technology – from machine learning and advanced natural language processing to innovative portfolio construction – CFM positions Cumulus not merely as another program but as a strategically refined investment strategy crafted for discerning investors in a challenging financial era. As CFM continues to innovate and expand its analytical arsenal, Cumulus promises to serve as both testament and tribute to a legacy of quantitative excellence.
Disclaimer:
Any statements regarding market events, future events or other similar statements constitute only subjective views, are based upon expectations or beliefs, involve inherent risks and uncertainties and should therefore not be relied on. Future evidence and actual results could differ materially from those set forth, contemplated by or underlying these statements. In light of these risks and uncertainties, there can be no assurance that these statements are or will prove to be accurate or complete in any way. All opinions and estimates included in this document constitute judgments of CFM as at the date of this document and are subject to change without notice. CFM accepts no liability for any inaccurate, incomplete or omitted information of any kind or any losses caused by using this information. CFM does not give any representation or warranty as to the reliability or accuracy of the information contained in this document. The information provided in this document is general information only and does not constitute investment or other advice. The content of this document does not constitute an offer or solicitation to subscribe for any security or interest”
* CFM paid a marketing fee to promote recent awards which includes this article written by THFJ.