Time really matters: How do hedge funds rely on data science? | By Anastasia Nesterova | Capital | June 2021
The 20 most successful hedge funds in the world raised 63.5 billion U.S. dollars in 2020, setting a volatility record for a decade, as technology stocks rebounded sharply after the pandemic-induced sell-off. LCH Investment Report.
“The net income of $63.5 billion created by the top 20 fund managers for their investors is the highest in a decade. In this sense, 2020 is a year for hedge funds,”
-LCH Chairman Rick Sopher said in a statement.
It is generally believed that all technological innovations in the investment field were first tried by hedge funds. It is true: hedge funds manage investors’ capital in order to increase capital with minimal risk-this ratio of profitability to risk attracts their clients.
The main goal is to use the growth of other assets to offset the decline of certain assets. In order to achieve this goal, a well-thought strategy must be used and an excellent forecast of future trends. Technological innovation can be a tool to make these processes more accurate and efficient.
If hedge funds have to choose between lower but stable profitability and possibly higher roller coaster returns, most of them will choose the first option.
For example, the strategy of “stable and risk-free 200% per year” will be more advantageous than the strategy of “500% risk and greater loss per year”. This is because hedge funds are financially responsible for investors’ funds.
High returns are not enough, you need high risk adjusted returns. Most developers of hedge fund algorithmic trading robots started building software with this axiom in mind.
If you are confident in market analysis or ready to learn financial skills, then you can try to create your own algorithmic trading robot.There are special platforms on the market to create trading robots-such as Thinker swimming (Thinker swimming Group company) or MetaTrader 5 (Yuan quotation software company). Users can write their own robots or consultants, or download/rent/purchase any ready-made applications.
The most complex trading robot uses 3 moving parts:
[signal generator] -> [risk allocation] -> [execution]
…So the biggest challenge is your creativity and insight in signal selection and processing strategies.
Trading robots have taken root in hedge funds. If you ask why, the answer is time. In the financial world, even milliseconds can make a difference.
A good example is Spread the network case, It established a direct fiber optic cable between the Chicago Mercantile Exchange and the data centers of the Nasdaq Exchange in New Jersey. This is necessary in order to receive data a fraction of a second earlier than other market participants.
Algorithmic trading robots run much faster than human thinking time plus reaction time. Most of the computer algorithms used by hedge funds for trading mimic the behavior of human traders, but are more systematic, faster, and cheaper.
Priority technologies for hedge funds include artificial intelligence and machine learning. A 2018 BarclayHedge’s research It is found that more than half of market participants use these methods to make investment decisions, and two-thirds use these methods to generate trading ideas and optimize investment portfolios.
This interest in artificial intelligence is due to the need for hedge funds to quickly calculate trends, look for signals in the news, and predict that the price of a particular stock will change. The more accurate the forecast, the lower the risk and the higher the profit. One does not know the future, but artificial intelligence can predict where the market will go based on historical data and changing external factors.
The amount of data that artificial intelligence can analyze is much larger than that of humans, and the processing speed is much faster. The more data analyzed, the more accurate the forecast. Therefore, artificial intelligence has completed the tasks set by data scientists.
Data scientists initially build models based on data. Once the model is created, it will be tested on past events-this process is called backtesting. For example, there is a model that allows you to predict tomorrow based on a certain behavior of stock prices. This prediction forms the basis of the hedge fund’s actions—and shows whether to buy or sell specific stocks. In order to evaluate the accuracy of the prediction, historical data is applied to the model.
In theory, you can build a model without such tests, but for hedge funds, risk is no less important than profitability (remember, the way hedge funds manage investor capital is Increase capital with minimal risk).
So backtesting is the best choice. During the backtesting process, we will not buy stocks based on wrong predictions and cause financial losses, but we will check whether the predictions come true. For example, you can enter the data collected ten days ago into the system, make a prediction for the next day, and compare it with the stock price performance nine days ago to look into the future.
Most importantly, the prediction of artificial intelligence still needs human verification. Whether it’s done in random order or only testing the most unexpected predictions, the best choice is up to you.
The challenges that data science can solve in hedge funds are huge. Data analysts select information and classify it, but there may be various influencing factors-macroeconomic and political events, industry trends, individual company performance, its competitors, etc. Remember that everyone is talking about the ever-increasing amount of data big data-data lakes, even data oceans.
It is important to understand that the hedge fund industry and the financial industry as a whole value the effectiveness of strategies rather than complexity. You should always try the simplest models first, and then consider whether you need to complicate them. It is also important to understand the basic economic basis of a particular idea. So the use of artificial intelligence is not necessary, but if there are indicators that can be significantly improved using ML methods, then the technology should be implemented.