Quantitative investment firms typically adopt one of two research management models:
1. The traditional Silo PM model,
2. The evolving Centralized Book model.
1. Silo PM
The Silo PM model, also known as the multi-manager platform, involves a firm hiring multiple Portfolio Managers (PMs) who operate in parallel. Each PM builds their own team and functions independently, creating an environment of internal competition. Prominent firms currently utilizing this model include Citadel, Millennium, and Point72.
2. Centralized Book
In the Centralized Book model, researchers collaborate through a division of labor, similar to an assembly line, rather than competing. Each researcher focuses on a specific link in the R&D chain, ensuring high specialization and accountability.
Under this model, you won't find a "quant superman" who masters programming, modeling, backtesting, and trading all at once. The advantage lies in deep specialization: the investment process is broken down into expert modules, each optimized and integrated to create a seamless pipeline from research to performance.
This structure is also highly inclusive. If you possess deep expertise in a specific research area, you can find a role tailored to your skills.
3. The Decline of Silo PM and the Rise of Centralized Book
North American hedge funds originated with the Silo PM model. In the 1980s and 90s, renowned quant firms often started with a few mathematicians or physicists developing their own models and trading. As they grew, they hired junior staff to assist with research, gradually formalizing the Silo PM structure.
Thirty years on, while quant methods have generated alpha, markets have matured, and competition among PM teams has become highly saturated and zero-sum.
Problems with the Silo PM Model
Problem 1: Strategy Homogenization
Different PM teams often spend their time independently developing similar strategies.
This is partly attributed to the maturity of the US market and the vast availability of research, leading to severe homogenization. As the industry saying goes: "If you've discovered a new signal, rest assured, at least ten thousand others have found it too."
Problem 2: The "Superman" Requirement
Researchers are expected to be jacks-of-all-trades, lacking deep specialization.
The workflow for a researcher in a Silo PM model typically involves everything: data procurement and cleaning, model development, backtesting, iteration... It demands a "superman" with speed, quality output, and immense pressure.
Problem 3: Limited Centralized Support
The firm acts like a horse racing track owner, providing the platform and letting the PMs (horses) compete. The winning PM gets the best rewards.
While this fosters competition, it leads to significant resource waste. Tasks like database access and backtesting tools often require each team to build and maintain its own infrastructure, resulting in massive duplication of effort.
Problem 4: High PM Turnover
In a classic Silo PM setup, successful PMs become highly independent and versatile, capable of handling research, development, and trading. Once they've built a strong track record, they have significant leverage and can easily leave to start their own firms or join competitors. Even Citadel has faced the dual challenge of top performers leaving and underperformers being let go.
In reality, with tougher market conditions and increasing difficulty in generating profits, the limitations of the Silo PM model have become evident.
According to our knowledge, firms like Citadel and Point72 are shifting towards the Centralized Book approach. Citadel, for instance, maintains the Silo PM model in its discretionary L/S department, but its quantitative arm operates entirely under the Centralized Book model.
Advantages of the Centralized Book Model
The Centralized Book model effectively addresses the issue of homogenized competition, allowing quant researchers to focus more independently on market dynamics. Successful adopters include Renaissance Technologies, Two Sigma, and DE Shaw.
Two Sigma's model is a notable example: each researcher focuses on a small component of a strategy. Some teams can even dedicate time to exploring highly esoteric alternative datasets – a clear advantage of having the resources and structure for deep specialization.
Of course, this model requires a firm to establish a standardized, overarching methodology covering the entire lifecycle from strategy research to execution and profit generation. While the work is collaborative, a core leadership role is essential to maintain a comprehensive view and ensure the organization runs smoothly.
From a career development standpoint, the Centralized Book model also offers significant advantages. Renowned quant investor Marcos Lopez de Prado, in his lecture "Escaping The Sisyphean Trap: How Quants Can Achieve Their Full Potential," analyzes the pros and cons of these two models. His conclusion is that in the era of machine learning, the Centralized Book model is superior to the often inefficient busywork found in Silo PM structures, making it a more suitable path for a quant's career growth.
4. Which Model Suits the Chinese Market?
The debate between Silo PM and Centralized Book is particularly lively in China, given its relatively younger quant industry which hasn't yet reached the maturity level of the US market.
However, a crucial factor is that the boom in China's quant industry coincides directly with the rise of AI and machine learning. Under these circumstances, the relative merits of each model warrant a more nuanced, long-term consideration.
top-tier Chinese quant firms have either completed the transition to the Centralized Book model or are in the process. In contrast, small and medium-sized firms, constrained by limited computing power and talent, largely continue using the Silo PM model.
This isn't to say the Silo PM model has no place. From the perspective of many domestic practitioners, the current focus is on developing strategies tailored to the unique characteristics of the Chinese market. Given the relative market inefficiency, alpha discovery is still comparatively easier, and the intense, "all-hands-on-deck" pressure seen in mature markets isn't yet the norm.
Furthermore, a major concern for practitioners is the "eating from the same big pot" mentality – where compensation doesn't reflect individual performance. Therefore, the Silo PM model remains popular domestically for its clear profit-sharing and protection of intellectual property .
What are your thoughts on choosing between these two models? Feel free to contact us for a discussion! We look forward to hearing your perspectives.
Welcome to our Blog!
If you are interested in career opportunities, please send your CV to: recruiter@alphamaker.link
If you would like to share your story, please contact us on WeChat: AMQuantcareer