QuantDesk® Machine Learning Forecast

for the Week of November 27th, 2017

Protecting Your Long Only, Bottom Up, Deep Value Fundamental Portfolio

Erez Katz writes about Reducing Turnover Through Quantitative Research

by Erez Katz, CEO and Co-founder of Lucena Research.

We have been experiencing two very different types of incoming leads lately. One is centered around blockchain technology, and in particular cryptocurrency trading. The other stems from growing interest in quantitative analysis technology by traditional deep value fundamental analysts. Since we’ve yet to form a comprehensive research strategy for cryptocurrency trading, I’d like to defer this discussion to a later time. For today, I’d like to discuss how machine learning can be used effectively to enhance human intelligence in the context of fundamental buy and hold research.

What Is A Deep Value Fundamental Portfolio?

Value investing is a style of investment which focuses on specific stocks that are priced below their intrinsic value. Individual stocks are valued mainly based on their companies’ fundamental and financial performance. Most analysts have gathered domain expertise through many years of research and intimate knowledge of very few companies which in turn serve as the basis of highly concentrated portfolios. The concept of quantitative analysis has been mostly foreign to deep value fundamental analysts until recently.

What Has Changed?

First, we need to give credit where it’s due, as most deep value fundamental managers have been dead right for the past 9 years. Indeed, since the financial crisis there has been a constant flow of money away from active management to traditional buy and hold instruments. However, many portfolio managers recognize that markets can’t continue to climb forever and are growing nervous that our bull market is running out of steam. Many still hold a strong conviction in their portfolios but are seeking protection from an inevitable pullback. In addition, traditional analysts are finally recognizing that there is a big data evolution unfolding right in front of them and are looking for ways to incorporate data science into their traditional research.

Lucena’s Solution

Lucena’s Hedge Finder is geared to protect a portfolio from major pullbacks by introducing additional assets that together form an anti-correlation trend line against the portfolio’s core holdings’ trend line. The technology underneath Lucena’s Hedge Finder is a sophisticated machine learning pattern-matching technology, but QuantDesk makes it easy to deploy by simply following a few short wizard screens.


For today’s demonstration, I have arbitrarily constructed a portfolio of some of the most widely held stocks from the Dow 30.

Image 1: Bottom-up portfolio from the Dow -30.

Using Lucena’s Hedge Finder wizard we can now identify a set of additional constituents from the S&P 500 that when added to the portfolio will preserve its trend line while reducing its volatility.

Image 2: Hedge wizard summary screen. The wizard is an easy to follow set of screens that allows the user to configure the hedger with ease before execution.

What’s exciting about the Hedge Finder wizard (and for that matter, all of QuantDesk’s wizards) is that it gathers only the basic information and upon execution it forms a deep grid search in order to identify which configuration produces the best outcome. For example, rather than having the user select the look-back period (the period that the machine gathers historical information from in order to make its predictions, also called the training period), the machine tries a sequence of look-back periods in order to identify which works best.

Image 3: Hedge Finder wizard output screen. The blue line represents the original portfolio while the orange line represents the hedged portfolio. To the right of the vertical line is the before and after portfolio’s projected performance cone for the upcoming month. On the right, you can analyze the before and after statistics. As you can see the areas highlighted in orange are the metrics that the Hedge Finder was able to improve. Specifically, higher Sharpe, lower volatility and even a slightly better return during the look-back period.

Next, QuantDesk makes it easy to validate if the hedging method works out-of-sample on time periods that have not been analyzed in the past. Again, the Hedge Finder backtest wizard makes it extremely easy to interface with where the user can enter the absolute minimum information, such as the period of time he/she wishes the backtest to cover. The results are then displayed in a comprehensive performance report for additional analysis.

Image 4: Hedge Finder wizard backtest summary. The wizard is an easy to follow set of screens that capture basic backtest configurations before execution.
Image 5: Hedge Finder wizard backtest outcome. Past performance is no guarantee of future returns. The backtest simulates rolling back time and demonstrating how the Hedge Finder actually works over an extended timeframe in the past. QuantDesk features a comprehensive backtest performance report. In this example, you can see how our hedge portfolio performed against its benchmark $SPXTR (the SP total return benchmark). In addition, you can see in the scatter plot chart (marked in green on the bottom left) how the hedged portfolio indicates lower volatility for a higher return against the benchmark.

Finally, it takes only one click to convert the backtest into a forward-traded live strategy.

Image 6: Creating a live forward-traded strategy based on the same execution rules of a backtest.


QuantDesk® Hedge Finder enables the protection of any portfolio without disturbing its core holdings. The easy user interface can enable a non-technical portfolio manager with quantitative research for a fraction of the usual cost and time compared to other research alternatives.

Would you’d like to put your portfolio to the test? Feel free to reach out to us in confidence and we will gladly walk you through process. info@lucenaresearch.com