104 – Alpha Examples by Data Category: Part 1

Welcome to WorldQuant’s Learn to Quant – where we hope to demystify quantitative finance research and make it accessible to you – guiding you through different ideas and their implementation on WorldQuant BRAIN. WorldQuant BRAIN is a simulation platform that provides datasets and tools to test your own ideas and gain feedback in real time.

I’m Nitish Maini, the Chief Strategy Officer at WorldQuant. This series is hopefully your gateway to powering up your approach to quant research. So, let’s learn to quant!

Quant researchers tap into a variety of data to build predictive signals or alphas – which are defined by WorldQuant as mathematical models that seek to predict the future price movement of various financial instruments.

The data categories include Price Volume, Fundamental, Analyst, Sentiment, Options, Model, Insider Transactions, Short Interest, and more. Over the next few videos, we will introduce 4 data categories, and illustrate each with an alpha example, from idea to result. So log into WorldQuant BRAIN.

1. Price Volume Data

We’re going to start with the Price Volume data category. It includes data like stock prices – open, high, low, close – and other trading related information like volume of shares traded and market capitalization.

The change in a stock’s price between the open and close of a trading session can be a potential idea. Let’s see how this information can be used as a signal to generate an alpha.

1.1 Hypothesis

Here is one hypothesis: if a stock’s close price is lower than its open price, we may anticipate a reversion, expecting the stock to bounce back and outperform others in the near future. We take a long position, profiting if the price rises.

Conversely, if the close price is higher than the open price, we may expect a reversion to a lower price, prompting us to take a short position.

Shorting an instrument usually involves borrowing it from a broker and then selling it in the market. Later, we buy it back from the market and return it to the broker. We profit if the price falls after we have shorted.

1.2 Implementing the Alpha

Now let’s implement this idea on WorldQuant BRAIN, using the proprietary expression language. You can access all data categories discussed in this series on BRAIN.

• The group_rank function takes the close–open price difference as the first input and subindustry as the second. It ranks this difference among all stocks in the same subindustry and assigns values between zero and one. Higher values get better ranks, closer to one, and lower values get ranks closer to zero. Stocks with negative alpha values are shorted, positive ones are bought.
• Next, decide the data timing of the data for the alpha, referred to as Delay. Delay-1 uses yesterday’s prices, while Delay-0 uses today’s prices until a chosen time.
• Also, decide if you want to use Decay for your alpha. Decay considers a weighted sum of past alpha values to smooth out fluctuations or outliers. Decay helps if we want to use the previous day’s alpha values or to reduce the turnover.

In this simulation, we:

• Use Delay-1 data to prevent look-ahead bias,
• Neutralize alpha over subindustry,
• Apply a decay of 10 for signal smoothing,
• Restrict the maximum capital on any single stock to 1%.

The backtesting simulation runs for the previous five years, generating an alpha vector for each day.

Results: A consistent Sharpe of 1.7 across years with decent coverage across the selected universe of top 3000 US stocks on the basis of liquidity.

Potential improvements:

• Control turnover to trade only in conditions conducive to reversion.
• Trade only in higher volatility conditions and hold during low volatility.

2. Fundamental Data

Next up, the fundamental data category. Fundamentals capture the underlying business, financial and operational health of a company, usually reported every quarter. While the data fields are many, they can be summarized into three financial statements: Balance Sheet, Income Statement, and Cash Flow Statement.

• The Balance Sheet provides a snapshot of the company’s financial health, detailing its assets, liabilities, and equity.
• The Income Statement illustrates the company’s profitability, showing how revenue is transformed into income after accounting for various expenses.
• The Cash Flow Statement reveals the company’s liquidity by tracking the inflow and outflow of cash from activities like operations, investments, and financing.

2.1 Hypothesis

Now, let’s discuss an idea based on the changes in the cash flow from operations and market cap of a firm. Cash flow from operations is the cash earned through core business activities. Market capitalization represents the firm’s worth, calculated as the current market price of one share multiplied by the total number of the company’s outstanding shares.

The ratio of cash flow from operations to market cap indicates if a company is fairly valued in the market with respect to its cash generating abilities. One theory is that a high ratio suggests an undervalued firm which could provide higher future returns and vice versa.

The hypothesis of the alpha is that if this ratio is improving over time, then the company’s stock will outperform others; hence, we take a long position in the firm. Similarly, if the ratio declines, we expect underperformance and take a short position.

2.2 Implementing the Alpha

Now let’s examine how our idea gets implemented using BRAIN’s expression language.

-group_rank(close-open, subindustry)

• The time series z-score function shows how the cash flow to market cap ratio has changed over the last 60 trading days. In statistics, z-score describes how far a value is from the mean of a group, in terms of standard deviations.
• We simulate the alpha expression over the top 3000 most liquid US stocks and neutralize over the market, to make the alpha long-short neutral.

Results: A consistent Sharpe of about 2.03 across years, 9% returns, 29% turnover, and decent coverage across the selected universe.

Potential improvement:

• Consider analyst predictions for each stock’s operational cash flow, rather than the last actual value. Analysts’ estimates provide insights into expected company performance before the actual quarterly earnings reports are released. Using these estimates may enhance the accuracy of fundamental signals.

In this video, we explored price volume and fundamental data with alpha examples. Next, we’ll discuss options and sentiment data categories. Hopefully you can grow your knowledge and further your journey into quant finance research – as we explore alpha examples with other data categories. See you in there!