You might have heard about quantitative funds and how they are taking over the financial markets. Although fascinating to follow and try to find out what they do, these types of funds are very secretive about their practices. The available information about them is scarce, incomplete, or even worse, incorrect.

In this article, I’ll answer the most common questions I get asked when it comes to quantitative funds. So, without any further ado, let’s get started!

How do Quantitative Hedge Funds invest?

In contrast to traditional firms, quantitative funds are famous for applying cutting-edge mathematical and statistical techniques to develop their trading strategies. These funds hire employees from highly technical backgrounds, such as computer science, physics, mathematics, electronic engineering, and to a lesser degree, economists with a strong background in econometrics.

The most famous example of a quantitative fund is Renaissance Technologies; the firm started by Jim Simons that only hires people with expertise in any technical field except finance. Although its legendary Medalion Fund is now closed and, as a consequence, its returns are not disclosed anymore, it yielded an average yearly return of 71% between 1993 and 2014.

Jim Simons giving a lecture at Berkley. Source

Quantitative Fund” is primarily an umbrella term for all funds with a mathematically rigorous approach towards investing, but they are not all similar to each other. We could further categorize them into arbitrage funds, Alternative Data Funds, Quantamental Funds, and Machine Learning Funds.

Without going into too much detail, we could describe each one of these quant funds as follows:

  • Arbitrage Funds: These funds specialize in creating very low latency algorithms to take advantage of minimal price discrepancies between exchanges or highly correlated assets. They mostly hire computer scientists and electronic engineers to develop custom software and hardware that minimize execution times. Additionally, they tend to colocate their algorithms next to the trading exchanges server to reduce latency further (check out my article on colocation).
  • Alternative Data Funds: these funds use different data sources to estimate and forecast the future price of assets. Whereas regular funds use balance sheets and pricing data, these funds create strategies from a broader set of data sources, such as satellite images, credit card transactions, mobile geolocation, or flight tracking. If you are interested in alternative data, check out my article.
  • Quantamental Funds: as the name suggests, these funds intersect with fundamental and quantitative investing. They employ large datasets of financial information and analyze them through complex computational models. In most cases, they do not engage in intraday trading, their algorithms are not fully automated, and human input is essential to their strategies.
  • Machine Learning Funds: these funds leverage today’s inexpensive processing power to train highly complex machine learning models with large amounts of data. By doing so, they find very subtle patterns in the data that are impossible to spot by any human. 

As you might have guessed, many quantitative funds could fall into more than one of the above categories. Especially popular are machine learning funds that use alternative data sources as inputs for creating their trading signals.

Advantages and Disadvantages of Quant Funds

One thing is for sure: quantitative funds are fascinating and one of the most mentally stimulating places to work if you are a researcher that constantly wants to iterate and test hypotheses.

But, despite having lots of advantages, they also have shown quite a few disadvantages and risks over the years. So, in the following section, I will briefly describe both!

Advantages of Quantitative Hedge Funds

A more rigorous approach to investing

Some metrics of a strategy that I’m currently testing.

Generally speaking, traditional portfolio managers use a more heuristic set of rules, resulting in loosely defined trading strategies that cannot be perfectly described and automated. Thus, it is not as straightforward to backtest and statistically analyze the performance of their strategies as it is when dealing with quantitative and clearly defined strategies.

Even if we keep in mind that backtests are never precise and do not guarantee that future performance to replicate the past, they are one of the most valuable tools for researching and testing automated strategies. Testing how a new idea would have performed historically enables quantitative researchers to apply the scientific method and iterate over hypotheses and theories in a more rigorous fashion.

More cost-efficient

Instead of sitting all day long at a trading desk executing orders, employees can focus on researching and implementing novel and innovative strategies. Quantitative funds have a tendency to have more researchers than traditional funds, in addition to requiring fewer traders. Quant shops benefit from being able to test more hypotheses, which in turn leads to a higher probability of finding profitable strategies (ceteris paribus). 

Having said that, even the safest and most reliable, and most tested automated strategies do require periodic monitoring to make sure that the trades get executed as expected.

More scalable than traditional funds

Some quantitative hedge funds rely heavily on automation. They can leverage their valuable trading ideas and apply them to hundreds of assets without linearly scaling the number of employees required in a traditional environment. It is common practice to have a single server handling multiple strategies across hundreds of assets. If required, it is also relatively inexpensive to install an additional server.

Having said that, ensuring that all algorithms work smoothly requires constant monitoring. Employees must also ensure that data sources used as inputs are always clean, homogeneous, and ready for ingestion.

Discover new sources of alpha

Unlike traditional investing, where everyone is looking at the same balance sheets and pricing data, quantitative hedge funds research and exploit novel datasets with the potential for untapping new sources of alpha.

Quantitative funds not only use traditional sources of information, such as income statements and historical time series of prices but also exploit the ever-growing availability of alternative sources of information. Alternative Data is an umbrella term that refers to datasets previously not used in financial markets, but that could be creatively applied for implementing trading strategies.

Although funds logically avoid revealing their alternative data sources, it is widely known some of them actively use satellite imagery, mobile geolocation, social media sentiment data, and credit card transactions, amongst others. I wrote an entire article on Alternative Data, which you can check out here.

Disadvantages of Quantitative Hedge Funds

There are quite a few more disadvantages of quant funds, especially those specializing in machine learning. It is also quite common to read news articles about one of them going bankrupt from time to time. As such, I find it worthwhile to go over the most common causes of their failure.

A majority of Quant Funds tend to use Black Box Models

Black Box Models have the characteristic of producing valuable and relevant outputs without revealing the formulas or logic executed for coming up with that solution. These models are put in place to generate signals based on the forecasted price of an asset, and automated algorithms use said signals to create buy and sell orders.

As such, not knowing the decision process of the models that drive the trading decisions poses an enormous risk that other types of strategies do not have. Moreover, these algorithms require monitoring, especially in the beginning, to prevent strange edge cases that could lead to erratic behavior and subsequent losses.

It is also important to note that not all machine algorithms are Black Box models, and even some straightforward models can be easily interpreted by any human trader with a basic background in statistics. Such is the case for logistic regression models and decision trees, which are commonly used as a benchmark in the context of classification problems (price up / price down, for example). On the other extreme, we have deep neural networks, ensemble models, and gradient-boosting algorithms that tend to provide high accuracy but at the expense of not being interpretable.

A common practice to address this issue is to implement fail-safe measures. 

It is worth mentioning that much of today’s academic efforts in machine learning are geared toward finding novel ways of making complex models somewhat easier for humans to interpret.

Pressure to come up with strategies

Although quantitative researchers use scientific approaches to testing their trading hypothesis, they are not exempt from feeling the pressure to deliver a trading strategy on a somewhat regular basis.

As such, they might be tempted to come up with a false positive and ship it to production instead of discarding it and finding a new idea. Highly qualified colleagues have confessed to doing so at least once during their careers.

It is a cause for suspicion when a researcher that focuses on intraday trading strategies suddenly delivers a strategy that trades monthly. They probably delivered it to buy themselves some precious time before enough actual trade data was collected to analyze the strategy’s performance. Also, keep in mind that quant salaries are high and are thus a great incentive to maximize the probability of remaining employed.

Based on a single strategy

Some quantitative hedge funds heavily rely on a single thesis and are highly specialized in some niche topic. In most cases, it is that precise specialization that enabled the firm to get its funding and good results during its initial years. As time goes by, more and more competitors develop similar approaches and strategies by using the same alternative datasets, and thus the alpha of the strategy tends to decrease. This phenomenon is known as alpha decay, and you can read the article I wrote on the topic here.

No proper risk management practices

Machine learning funds are oftentimes created by highly technical academics whose novel, different approach toward financial markets allows them to find previously undiscovered trading strategies. This can be verified by the emergence of a generation of funds created by physicists, mathematicians, statisticians, and computer scientists.

More often than not, these funds lack a severe lack of industry experience, which also has its disadvantages. One of the main ones is not having any proper risk management whatsoever. Consequently, machine learning funds are often either overleveraged, underdiversified, or both.


As you can see, quantitative funds are fascinating, and by no means are they just a temporary fad. In my humble opinion, these funds will continue to grow as a proportion of the overall active fund industry and eventually comprise most of the market.

I tried to answer as many questions as possible while also keeping the length of the article at a reasonable size. I prioritized the questions based on a mix of how popular, engaging, and insightful they were, but if you think I missed an important one, please let me know in the comments!



No responses yet

Leave a Reply

Your email address will not be published. Required fields are marked *