Research Blog Events
About Careers Newsroom Get Started

Value Strategies based on Machine Learning

Incorporating Profitability Measure and Sentiment Signals to Identify Winners and Losers

Following the popularity of their "Big Data and AI Strategies", the J.P. Morgan Global Quantitative & Derivatives Strategy Team demonstrates ideas on implementing machine learning algorithms to value strategies based on valuation.

Value Strategies

They attempt to predict the “fair value” of stocks using a large number of equity characteristics and quantify a “mispricing” signal where they buy undervalued stocks and short overvalued stocks.

Highlights of their Findings in Value Strategies:

  • Filtering mispriced stocks based on Gross Profits significantly improves their strategy returns, and results are better than using Return On Equity. They prefer to buy undervalued stocks with high profitability, and short overvalued stocks with poor profitability.
  • RavenPack’s news sentiment data is a useful overlay to their value strategy. Avoiding stocks with poor (good) sentiment in the long (short) portfolio helps to improve risk-adjusted returns.

Value Strategies

RavenPack Data as an Overlay

The figure below shows the results of their final strategy which filters the long and short portfolios based on sentiment levels and changes in sentiment, together with the thresholds that they choose. The Information Ratio (after transaction costs) improves from 1.46 to 1.55. This is mainly due to an increase in returns, while volatility is almost the same.

Value Strategies

Value Strategies

Value Strategies Methodology

For more details, our Chief Data Scientist gives an overview of the 5-step approach considered by J.P. Morgan when constructing their valuation strategy.

JP Morgan’s Global Quantitative & Derivatives Strategy team [JPM] has turned their attention to stock selection - using Machine Learning to come up with a valuation based strategy. In their new report, they attempt to predict the “fair value” of stocks using a large number of equity characteristics, and quantify a “mispricing” signal that either buys or shorts stocks that are under- or overvalued.

So why is this interesting to RavenPack? Well, the report shows that including a news sentiment overlay based on our data improves the overall performance of their mispricing signal - yielding an Information Ratio of 1.55 (and 1.85 without transaction costs).

Below, we provide an overview of the 5-step approach considered by JPM when constructing their valuation strategy. The backtesting universe consists of about 2000 stocks across 39 countries - all belonging to the MSCI AC World index.

Step 1- Feature Construction

As part of creating a mispricing signal, JPM considers over 37 stock characteristics that could potentially help to predict the price-to-book of a stock, with perspectives ranging from valuations, profitability, operational efficiency, quality, growth, sentiment, and risk. Table 1 lists the features that are included in their model.

Value Strategies

Step 2 - Machine Learning Models

Taking the above features as input, JPM fits a linear and elastic net regression as well as a boosted tree model (XGBoost). They find that the latter model provides the strongest performance with an IR of 0.89. A combination of all three models does similarly well with an IR of 0.88 (see Figure 18).

Value Strategies

Step 3 - Gross Profit Overlay

Intuitively, investors are interested in buying stocks at a reasonable price (i.e. value stocks) which also demonstrate their ability to make profits. By combining the mispricing model with Gross Profit to Assets (GP/A), JPM is able to improve strategy returns and increase the IR from 0.88 to 1.26 - also outperforming a combination strategy based on ROE as an alternative measure of Profit (see Figure 20).

Value Strategies

Step 4 - Combining vs. Filtering

In the above “combinative” approach, investors may end up with stocks that are either largely undervalued (but not really profitable), or with a very high profitability (but not necessarily undervalued). Instead, JPM tries to play the “mispricing” signal by first looking at stocks that are highly undervalued or overvalued, and then using the GP/A to filter stocks that they want to retain in the long and short portfolios - only keeping the top or bottom 40% based on GP/A, respectively. This approach brings further IR improvements, increasing it from 1.26 to 1.46, primarily driven by greater returns (see Figure 21).

Value Strategies

Step 5 - Including News Sentiment

As a final step, JPM shows how introducing RavenPack’s news sentiment signal as an overlay provides further improvements to their strategy. In particular, removing stocks with negative sentiment in the long basket (i.e. undervalued stocks with high profitability) and positive sentiment in the short portfolio helps to improve risk-adjusted returns. Figure 30 shows the results of their strategy, where the Information Ratio (after transaction costs) is increased from 1.46 to 1.55. This is mainly due to an increase in returns, while volatility is almost the same. Without transaction costs, this strategy would give an IR of 1.85.

Value Strategies

Value Strategies

Revisiting Value Strategies

Value strategies are looking to buy "cheap" stocks and sell "expensive" stocks. Whilst the logic is simple and intuitive, the actual strategy may not be that straight forward because we cannot observe the true value of a stock directly. Investors can use different proxies to estimate valuation from different perspectives. In equity space, the traditional way is to look at accounting items to infer the valuation of a company.

One could look at price-to-book ratios from the balance sheet, earnings yield from the income statement, or cash flow yield from the cash flow statement. As highlighted in “Sorting through the Trash” (Ma and Smith (2014)), combining valuations from different angles can lead to a more holistic measure, and ranking stocks with the "Holistic Value" factor, i.e. a combination of price-to-book (deep value), forward earnings yield (expected value) and price-to-cash-flow (intrinsic value) can drastically improve performance.

As many of the Value investors would recall, there have been numerous periods when Value has lost favor to other styles, particularly between 2010-2015 where a long/short strategy based on price-to-book in the Global Developed Market has lost 5.5% p.a.. In "Value Everywhere" (Hlavaty et al (2016)), we provide a comprehensive study on different value factors, and analyze their efficacies at the regional and sector levels.

Price-to-book: A rough path since 2010

Figure 1 shows the wealth curves of the quintile portfolios based on price-to-book in MSCI AC World. On the left, we show performance since 1994, where we observe an excellent period for value investors between 2000 and 2006. After the Global Financial Crisis in 2008, we have the well-known "junk" rally in 2009, but in general Value investing has been difficult since then. On the right of Figure 1, we zoom into the performance of price-to-book since 2010. Expensive stocks have been outperforming cheap stocks, except in the second half of 2016 where Value came back as a theme under the spotlight.

Value Strategies

What about other value factors?

Albeit being one of the most popular valuation measures, price-to-book is not the only value factor to look at. In Figure 2, we consider the performance of other value factors including dividend yield (a more defensive value), as well as some forwardlooking versions based on analyst forecasts.

As opposed to price-to-book and many other value factors, we find that shareholder yield (combining the effects of dividends, buy-backs and net issuance) performs quite well since 2010, as well as over the whole history from 1994 where it has delivered an Information Ratio of 0.92. As in "Value Everywhere" (Hlavaty et al (2016)), we notice that shareholder yield appears to be one of the best measures of Value.

This may due to the fact that shareholder yield better captures the reward to investors compared to dividend yield: Some stocks may prefer not to return cash to shareholders by way of dividends, and instead choose to use the cash for the purpose of M&A activity, share-buybacks or to decrease of the level of total debts (Quant Forensics: Volume 5, Replacing Dividend Yield with Shareholder Yield, Shaikh et al (2013)).

Click here to continue reading the "J.P Morgan: Value Strategies based on Machine Learning" White Paper

Request White Paper

Request White Paper

We will process your personal data with the purpose of managing your personal account on RavenPack and offering our services. You can exercise your rights of access, rectification, erasure, restriction of processing, data portability and objection by emailing us at For more information, you can check out our Privacy Policy.

Discover the RavenPack Platform

Choose from over 180 curated datasets covering a variety of different topics.

Explore Now