GB2590598A - System and method for optimal selection - Google Patents

Abstract

An investment portfolio, e.g. a quality-value (QV) portfolio, is constructed by filtering an initial selection of stocks, e.g. a benchmark equity index. A quality metric and valuation metric are identified for each stock within the selection. A first subset of stocks is identified by determining whether each stock’s quality metric passes a quality threshold. A second subset of stocks is then identified by determining whether the valuation metric for each stock within the first subset passes a valuation threshold. The second subset thus contains stocks which pass both the quality and valuation thresholds. Stocks may be equally weighted. The quality metric may be profitability, operating efficiency, leverage, return on equity, return on assets, return on invested capital, return on capital employed, gross margin, post-tax margin, pre-tax margin, profit margin, long-term debt to total capital ratio, interest coverage ratio, long-term debt to equity ratio, financial leverage ratio, total debt to equity ratio, or debt to assets ratio. Three quality metrics may be used, with the first subset of stocks being those stocks that meet all three metrics. The valuation metric may be enterprise value to gross profit ratio, free cash flow yield, price to earnings ratio, or price to sales ratio.

Description

Intellectual Property Office Application No G131916089.4 RTM Date:14 April 2021 The following terms are registered trade marks and should be read as such wherever they occur in this document:

BLOOMBERG FTSE

ISHARES

MORNINGSTAR MSCI

PYTHON

QUANTOPIAN S&P

YAHOO

Intellectual Property Office is an operating name of the Patent Office www.gov.uk/ipo System and Method for Optimal Selection The present invention relates to a computer implemented system and method for the identification of optimal elements within a larger set of elements, and particularly for constructing quality-value (QV) investment portfolios.

In everyday life, the best quality product or service is sought at the best possible price. However, this approach is not currently systematically applied to equity portfolio construction. Recent innovation in equity portfolio construction, such as factor-based 1() investing and fundamental-weighting, do not fully implement the idea of best quality at the best possible price.

"Smart Beta" exchange traded funds (ETFs) are products which provide investment portfolios that maximise exposure to identified factors that generate outperformance. For example, the iShares S&P 500 Value ETF provides a portfolio of stocks selected from the S&P 500 that benefit from low valuations. Mulfifactor ETFs, such as the iShares Edge MSCI Multifactor USA ETF, are optimised to provide maximum exposure to multiple factors (such as momentum, quality, and value). However, these approaches do not systematically select the stocks with the lowest valuation from a subset of quality stocks.

Current methods in the technical field of computer-based processing do not provide an effective way of systematically identifying a set of stocks with the lowest valuation from a set of quality stocks. There is thus a need for a computer-implemented method, whereby a processor carries out the technical steps for identifying an optimal set of elements from a larger set of elements, which will enable a systematic quality-at-areasonable-price equity portfolio construction approach.

According to a first aspect of the present invention, there is provided a method for constructing an investment portfolio from an initial selection of stocks, wherein the method is implemented on a computer in communication with a database of investment data for the initial selection of stocks, and the computer is programmed to execute the steps of the method, the method comprising: from the investment data, identifying one or more quality metrics and one or more valuation metrics for each stock in the initial selection of stocks; from the initial selection of stocks, identifying a first subset of stocks wherein for each stock a first quality metric of the one or more quality metrics passes a first threshold quality value; and from the first subset of stocks, identifying a second subset of stocks wherein for each stock one or more of the valuation metrics passes a threshold valuation value.

The invention provides a computer-implemented method which enables and provides for improvements to the capabilities and functioning of the computer. By utilisation of a two stage identification process in which from an initial selection of stocks, a first subset is identified in which each stock passes a first quality metric of the one or more quality metrics, and then from that first subset, a second subset is identified in which each stock passes a threshold valuation, a way of identifying a set is provided that is efficient in the use of computer resources since it is only from the first subset that the second subset of stocks is selected. This computer implemented method integrates into a practical application a broader method that provides for improvement of the functioning of the computer in this technical field. The size of data sets that are processed are controlled by the two stage selection process. The method thus provides a process that amounts to significantly more than simply a process of constructing an investment portfolio given the integration into a computer-implemented practical application that itself provides for improvement of the functioning of the computer in this technical field.

According to a second aspect of the present invention, there is provided a computing system for constructing an investment portfolio, the computing system comprising: a receiver configured to receive investment data for an initial selection of stocks from an investment database; a storage device configured to store the investment data; a processor programmed to execute the steps of: from the investment data, identifying one or more quality metrics and one or more valuation metrics for each stock in the initial selection of stocks; from the initial selection of stocks, identifying a first subset of stocks wherein for each stock a first quality metric of the one or more quality metrics passes a first threshold quality value; and from the first subset of stocks, identifying a second subset of stocks wherein for each stock one or more valuation metrics passes a threshold valuation value.

This method and system allow a computer to carry out the systematic identification of the best value stocks from a set of high quality stocks that have been systematically identified from a benchmark selection of stocks. By selecting the best valued stocks from the highest quality stocks, a well-performing investment portfolio can be created. Computer processing allows an investment portfolio to be quickly, easily and automatically constructed and therefore reduces the burden on the portfolio manager.

In one example, the one or more quality metrics comprise a first quality metric, a second quality metric and a third quality metric, and identifying the first subset of stocks further comprises: identifying a first set of stocks wherein the first quality metric for each stock passes a first threshold quality value; and identifying a second set of stocks wherein the second quality metric for each stock passes a second threshold quality value; and identifying a third set of stocks wherein the third quality metric for each stock passes a third threshold quality value.

In one example, identifying the first subset of stocks further comprises identifying the first, second and third sets of stocks from the initial selection of stocks; and identifying the stocks at the intersection of the first, second and third sets of stocks; wherein the first subset of stocks comprises the stocks at the intersection.

In one example, identifying the first subset of stocks further comprises identifying the second set of stocks from the first set of stocks and identifying the third set of stocks from the second set of stocks, wherein the first subset of stocks comprises the third set of stocks.

The first subset of stocks is therefore composed of stocks that have optimal values in three categories. This ensures that the first subset contains the highest quality stocks in the initial selection. Using three criteria to identify the highest quality stocks provides more accurate and reliable results than using just one or two criteria as it can provide a broader and more comprehensive view of the stocks' characteristics.

In an example, the one or more quality metrics are measures of one or more of profitability, operating efficiency, and leverage. Profitability, operating efficiency and leverage are good indicators of the quality of stocks. There are a number of metrics that can be used for each, depending on the preferences and goals of a portfolio manager.

In one example, the one or more quality metrics are one or more of the following: return on equity, return on assets, return on invested capital, or return on capital employed. In another example, the one or more quality metrics are one or more of the following: gross margin, post-tax margin, pre-tax margin, and profit margin. In a further example, the one or more quality metrics are one or more of the following ratios: longterm debt to total capital, interest coverage, long-term debt to equity, financial leverage, total debt to equity, and debt to assets.

In one example, the one or more valuation metrics are one or more of the following: enterprise value to gross profit ratio, free cash flow yield, price to earnings ratio, and price to sales ratio.

In one example, the stocks in the second subset are weighted equally. It is advantageous for the stocks in the investment portfolio to be equally weighted as this avoids any stocks dominating the portfolio and so the portfolio avoids being skewed.

In one example, the first subset contains no more than one-third of the number of stocks in the initial selection. In one example, the second subset contains no more than one-third of the number of stocks in the first subset. Reducing the initial selection and then the first subset by at least two-thirds ensures that a meaningful effect is achieved by filtering the stocks according to the quality and valuation metrics. Reductions by less than two-thirds may not exclude enough low quality or poor value stocks, resulting in a sub-optimal investment portfolio.

In one example, the second subset contains at least thirty stocks. This ensures that the investment portfolio is sufficiently diversified to be effective and profitable.

In one example, the method is repeated no more than monthly. Repeating the method ensures that the investment portfolio maintains an optimal selection of stocks. The desire to limit transaction costs means that it is preferable to not repeat the method too often. Monthly repetition provides a good balance between minimising transaction costs and optimising the portfolio.

Embodiments of the present invention will now be described in detail with reference to the accompanying drawings, in which: Figure 1 is a schematic diagram representing the reduction of the sets of stocks in the construction of a QV portfolio; Figure 2 is a flow chart representing steps in the QV portfolio construction method of the present invention; Figure 3 is a flow chart representing the method of the first step in the QV portfolio construction method of the present invention; Figure 4 is a schematic diagram representing the results of the first step in the QV portfolio construction method of the present invention; Figure 5 is a flow chart representing the method of the second step in the QV portfolio construction method of the present invention; and Figure 6 is a schematic diagram representing the results of the second step in the QV portfolio construction method of the present invention.

A method is provided for the identification of optimal elements. In embodiments, the method can be used for the selection of optimal stocks from a benchmark stock index in order to create investment portfolios from equities listed on global stock exchanges. In particular, the method can be used to create open-end funds, exchange traded funds (ETFs), and for managed accounts. The method enables the creation of portfolios comprised of high quality stocks selected at the most reasonable possible valuation, referred to as "quality-value" (QV) portfolios.

The method is implemented on a computing device with a processor programmed to carry out the steps of the method. In embodiments, the computing device comprises a personal computer, a smartphone, or a tablet computer.

The method does not just apply to the identification of optimal stocks, but rather has a range of technical applications, such as the identification of optimal materials for a manufacturing process. Identifying the optimal materials with which to manufacture a product requires the consideration of a large array of factors. This includes technical considerations, such as the suitability of the materials for the application of the product, and commercial considerations, such as the cost of obtaining the materials. Depending on the type of product to be manufactured and how the product is to be used, the suitability of materials can depend on a wide variety of factors, such as durability, strength, weight, hardness, elasticity, melting point, boiling point, and the time required to manufacture the product. The cost of obtaining materials can also depend on a wide variety of factors, such as availability, method of transport, distance of source, and scale of purchase.

The present method provides the ability to quickly and reliably compute which material, or combination of materials, is the most advantageous for the manufacture of a particular product. The method allows a plurality of factors to be considered and to contribute to the identification of optimal materials. By parameterising and optimising different aspects of a manufacturing process, the method can facilitate the efficient manufacture of well-made and high quality products.

For example, a processor programmed to carry out the steps of the method may be provided with a database of metrics for a range of materials, including information about the properties of the materials and other factors which contribute to materials' suitability and cost. Threshold values are specified for one or more material metrics and the processor executes the method steps to filter the range of materials using the threshold values to identify a selection of materials with optimum metric values.

Implementing this method on a computing device with a processor programmed to carry out the method steps allows the method to be versatile and easily adapted for various products, applications, and a varying importance of the considerations. Different products may require the use of different metrics in the filter; different applications may require different metric thresholds; and a changing importance of different considerations, such as needing to work to a tight budget, or needing a product to outperform competition, may require the metrics to be used in the filters in a different order. The method allows all of these adaptations to be performed quickly and easily by the processor, without the need to reprogram an amended method into the processor.

Embodiments of the method as it relates to the construction of QV portfolios will now be described in detail. As illustrated in Figure 1, in carrying out the QV portfolio construction method, the processor reduces a broad benchmark selection of stocks 2 into a subset of high quality stocks 4, referred to as the "quality subset". For the benchmark selection 2, a broad equity index such as the FTSE 350 or S&P 500 may be chosen. The quality subset 4 is then reduced by the processor into a subset of stocks with the best valuation 6, referred to as the "value subset'.

The stocks in the quality subset 4 are filtered from the benchmark selection 2 according to one or more stock metrics. In the embodiments described below, the processor is programmed to use at least three stock metrics to identify the quality subset 4. These metrics are measures of profitability, operating efficiency and leverage. In other embodiments, the processor is programmed to use just one or two of the stock metrics, or to use additional or alternative metrics.

Stock market indices for use as the benchmark selection of stocks 2 and stock metrics can be obtained from investment databases accessed through investment data providers such as Morningstar, Yahoo Finance and Bloomberg. Online investment algorithm platforms such as those offered by Quantopian and QuantConnect are able to provide access to the investment data.

In embodiments, the QV portfolio construction method is implemented on a computing device by programming the processor to execute the steps of the QV portfolio construction method using an appropriate programming language such as Python or C#.

The computing device has access to an investment database, preferably over a network such as the Internet. In embodiments, the investment database is accessed online through a live link for real-time provision of up-to-date stock information; in other embodiments, the whole database can be downloaded to the computing device for offline use in the QV portfolio construction method. Alternatively, an investment database may be provided and downloaded onto the computing device via a computer-readable non-transitory storage medium such as a USB drive or CD-ROM or via a cloud-based network connected to a remote server. In some embodiments, the data itself for the stocks in the benchmark selection 2 is stored on a storage device of the computing device, but in other embodiments only pointers or references to the stock data are stored on the storage device of the computing device. In further embodiments, the investment data is stored on a remote server on a cloud-based network and the investment data can be accessed on the cloud by the computing device.

In the embodiment demonstrated in Figure 2, the QV portfolio construction method is executed by the processor in two main steps 8, 10 and two auxiliary steps 12, 14. The first step 8 is the reduction of the benchmark selection 2 into the quality subset 4, and the second step 10 is the reduction of the quality subset 4 into the value subset 6 and basic QV portfolio 16. The auxiliary third step 12 is the application of a weighting to the stocks in the value subset 6 to provide an equal-weight QV portfolio 18. The auxiliary fourth step 14 is the repetition of the method in order to rebalance the QV portfolio 16, 18 to provide an updated QV portfolio 20. As demonstrated in Figure 2, the auxiliary fourth step 14 can be performed without performing the auxiliary third step 12.

In the first step 8, demonstrated in Figure 3, a subset of high quality stocks 4 is filtered out from the benchmark selection 2 by the processor in three stages according to profitability 22, efficiency 24, and leverage 26. In embodiments, the processor uses the profitability, efficiency and leverage filtering stages 22, 24, 26 to reduce the benchmark selection 2 to a quality subset 4 of at least around one-third of the initial size of the benchmark selection 2. Reducing the benchmark selection 2 by around two-thirds provides sufficient benefit from the filtering stages 22, 24, 26 without reducing the number of stocks to a point that would incorporate too much risk. Reducing the benchmark selection 2 by less than two-thirds may, in comparison, reduce the efficacy of the filtering stages 22, 24, 26 since fewer lower quality stocks would be discounted. At the same time, reducing the benchmark selection 2 by more than two-thirds may also reduce the efficacy of the filtering stages 22, 24, 26 since not enough higher quality stocks would be included, the quality subset 4 would be less diverse, and the investment risk may be higher. However, it may be appropriate or necessary in some situations to reduce the benchmark selection 2 by more than two-thirds, depending on the size of the benchmark selection 2. For very large benchmark selections 2 containing thousands of stocks, such as the Wilshire 5000 or MSCI World indexes, it may be necessary to reduce by more than two-thirds so that the final QV portfolio 16, 18, 20 is not too large to realise the selection effects. In such situations, the desired minimum size of the final QV portfolio would dictate the extent of the reduction in the first step 8.

Each filtering stage 22, 24, 26 is preferably independent of the other filtering stages 22, 24, 26, but may alternatively be dependent on one or both of the other filtering stages 22, 24, 26.

At the profitability filtering stage 22, to filter out stocks from the benchmark selection 2 by profitability, the processor may use one or more profitability metrics determined from the investment database to identify a filtered set 28 of the most profitable stocks in the benchmark selection 2.

Examples of profitability metrics that may be used to filter the benchmark selection 2 are return on equity (ROE), return on assets (ROA), return on invested capital (ROIC), or return on capital employed (ROCE). In some embodiments, the stocks are filtered using five-year averages of the profitability metrics. In other embodiments, the profitability metrics are averaged out over a period of time less than or more than five years. For example, three-year or one-year averages could be used. The processor uses a threshold value or percentage to delimit the most profitable stocks, such as by selecting the stocks that have a profitability metric above a certain threshold value, or a profitability metric that falls within a certain top percentage. The filtered set 28 of stocks is a representation of the most profitable stocks in the benchmark selection 2.

For example, the ROIC metric can be used to filter out all stocks in the benchmark selection 2 that have a five-year average ROIC within the top twenty-five percent of five-year average ROICs for all of the stocks in the benchmark selection 2.

The benchmark selection 2 may be filtered using just one profitability metric, or using a combination of two or more profitability metrics. When a combination of profitability metrics is used, each metric may be independent, or dependent on one or more of the other metrics. The metrics may be assessed in sequence, or averaged out.

For example, the profitable set of stocks 28 may be a combination of the stocks filtered out using ROIC in the previous example, in addition to stocks filtered out independently using a second profitability metric, such as stocks with a five-year average RCA within the top twenty-five percent of five-year average ROAs for all of the stocks in the benchmark selection 2. Alternatively, the second metric may be dependent on the first metric, so that the profitable set only contains stocks that meet both filtering criteria, that is, stocks with both a five-year average ROIC and a five-year average RCA within the respective top twenty-five percentages.

At the efficiency filtering stage 24, the processor may use one or more operating efficiency metrics determined from the investment database to identify a filtered set 30 of the stocks in the benchmark selection 2 that have the highest operating efficiency.

Operating margins such as gross margin, post-tax margin, pre-tax margin, and profit margin may be used as the operating efficiency metrics. In some embodiments, the stocks are filtered using five-year averages of the operating efficiency metrics. In other embodiments, the operating efficiency metrics are averaged out over a period of time less than five years. For example, three-year or one-year averages could be used. The processor uses a threshold value or percentage to delimit the most efficient stocks, such as by selecting the stocks that have an operating efficiency metric above a certain threshold value, or an operating efficiency metric that falls within a certain top percentage. This filtered set 30 of stocks is a representation of the most efficient stocks in the benchmark selection 2.

For example, the gross margin metric can be used to filter out all stocks in the benchmark selection 2 that have a five-year average gross margin within the top twenty-five percent of five-year average gross margins for all the stocks in the benchmark selection 2.

The benchmark selection 2 may be filtered using just one operating efficiency metric, or using a combination of two or more operating efficiency metrics. When a combination of operating efficiency metrics is used, each metric may be independent, or dependent on one or more of the other metrics. The metrics may be assessed in sequence, or averaged out.

For example, the efficient set of stocks 30 may be a combination of the stocks filtered out using gross margin in the previous example, in addition to stocks filtered out independently using a second operating efficiency metric, such as stocks with a five-year average profit margin within the top twenty-five percent of five-year average profit margins for all of the stocks in the benchmark selection 2. Alternatively, the second metric may be dependent on the first metric, so that the efficient set 30 only contains stocks that meet both filtering criteria, that is, stocks with both a five-year average gross margin and a five-year average profit margin within the respective top twenty-five percentages.

At the leverage filtering stage 26, the processor may use one or more leverage metrics determined from the investment database to identify a filtered set 32 of the least leveraged stocks in the benchmark selection 2. Ratios such as long-term debt to total capital, interest coverage, long-term debt to equity, financial leverage, total debt to equity, and debt to assets may be used as the leverage metrics. The processor uses a threshold value or percentage to delimit the least leveraged stocks, such as by selecting the stocks that have a leverage metric above or below a certain threshold value, or a leverage metric that falls within a certain top or best percentage. This filtered set 32 of stocks is a representation of the least leveraged stocks in the benchmark selection 2.

For example, the interest coverage metric can be used to filter out all stocks in the benchmark selection 2 that have an interest coverage ratio within the top twenty-five percent of interest coverage ratios for all of the stocks in the benchmark selection 2.

The benchmark selection 2 may be filtered using just one leverage metric, or using a combination of two or more leverage metrics. When a combination of leverage metrics is used, each metric may be independent, or dependent on one or more of the other metrics. The metrics may be assessed in sequence, or averaged out.

For example, the filtered set of least leveraged stocks 32 may be a combination of the stocks filtered out using interest coverage in the previous example, in addition to stocks filtered out independently using a second leverage metric, such as stocks with a debt to assets ratio within the best twenty-five percent of debt to assets ratios for all of the stocks in the benchmark selection 2. Alternatively, the second metric may be dependent on the first metric, so that the least leveraged set 32 only contains stocks that meet both filtering criteria, that is, stocks with both an interest coverage ratio and a debt to assets ratio within the respective best twenty-five percentages.

The processor determines the quality subset 4 by identifying the intersection 34 of the three filtered sets of stocks 28, 30, 32, i.e. the stocks that are identified at all three of the profitability, efficiency and leverage stages 22, 24, 26. Figure 4 represents the filtered sets of stocks 28, 30, 32 identified by the processor at each filtering stage 22, 24, 26 from the benchmark selection 2, and the quality subset of stocks 4 created by the intersection of each filtered set 28, 30, 32. The quality subset 4 is a representation of the highest quality stocks in the benchmark selection 2.

The processor may be programmed to perform the profitability, efficiency and leverage filtering stages 22, 24, 26 in any order. In the method of the first step 8 described above, each filtering stage 22, 24, 26 is independent and is performed on the whole benchmark selection 2. However, the first step 8 may be alternatively carried out so that only the first filtering stage 22 is performed on the benchmark selection 2. The second filtering stage 24 is then carried out only on the stocks identified in the first filtering stage 22 and the third filtering stage 26 is carried out only on the stocks identified in the second filtering stage 24. This alternative method of the first step 8 reduces the amount of computing that needs to be carried out. The quality subset 4 would in this case be comprised of the stocks identified at the third filtering stage 24 and the stage of identifying the intersection 34 would be unnecessary.

In the second step 10 of the QV portfolio construction method, demonstrated in Figure 5, at the valuation filtering stage 28 the processor filters the quality subset 4 using a valuation metric determined from the investment database to provide a subset of stocks 6 with the best valuation. Figure 6 illustrates the delimitation of the value subset 6 from the quality subset 4 of Figure 4. In embodiments, this value subset 6 is no more than one-third of the size of the quality subset 4. Reducing the quality subset 4 by around two-thirds provides sufficient benefit from the valuation filtering stage 36 without reducing the number of stocks to a point that would incorporate too much risk. Reducing the quality subset 4 by less than two-thirds may, in comparison, reduce the efficacy of the valuation filtering stage 36 since fewer poor value stocks would be discounted. At the same time, reducing the quality subset 4 by more than two-thirds may also reduce the efficacy of the valuation filtering stage 36 since not enough good value stocks would be included, the value subset 6 would be less diverse, and the investment risk may be higher. However, it may be appropriate or necessary in some situations to reduce the quality subset 4 by more than two-thirds, depending on the size of the benchmark selection 2 and by how much the benchmark selection 2 has been reduced in the first step 8. For very large benchmark selections 2 containing thousands of stocks, such as the Wilshire 5000 or MSCI World indexes, the quality subset 4 may also be large. Therefore, it may be necessary to reduce the quality subset 4 by more than two-thirds so that the final QV portfolio 16, 18, 20 is not too large to realise the selection effects. In such situations, the desired minimum size of the final QV portfolio would dictate the extent of the reduction in the second step 10.

In embodiments, the value subset 6 contains at least thirty stocks, however it is preferable that the value subset 6 contains at least forty stocks. Ensuring that the value subset 6 contains at least thirty stocks provides sufficient diversification within the QV portfolio 16, 18, 20. More optimum diversification is obtained with at least forty stocks.

In embodiments, the preferred minimum size of the value subset 6 is set as the starting point for the whole QV portfolio construction method and the filtering stages are calibrated to maximise the selection effects. For example, if the benchmark selection 2 is so large that thirty or forty stocks represent a very small proportion, such as the Wilshire 5000 or MSCI World indexes, the preferred minimum size of the value subset 6 may be larger than forty.

Examples of valuation metrics that may be used to enable the selection of the best valued stocks from the quality subset 4 are enterprise value to gross profit ratio, free cash flow yield, price to earnings ratio, and price to sales ratio. The processor uses a threshold value or percentage to delimit the best valued stocks, such as by selecting the stocks that have a valuation metric above a certain threshold value, or a valuation metric that falls within a certain top percentage. For example, the free cash flow yield metric can be used to filter out a predetermined number of stocks in the quality subset 4 with the highest free cash flow yields. The predetermined number of stocks to be filtered out may be the preferred size of the value subset 6, i.e. no more than one-third of the size of the quality subset 4 and at least thirty or forty stocks. The value subset 6 is a representation of the stocks in the quality subset 4 that have the best valuation, and a representation of the stocks in the benchmark selection 2 that have the highest quality combined with the best valuation.

The quality subset 4 may be filtered using just one valuation metric, or using a combination of two or more valuation metrics. When a combination of valuation metrics is used, each metric may be independent, or dependent on one or more of the other metrics. The metrics may be assessed in sequence, or averaged out. For example, the value subset 6 may be a combination of the stocks filtered out using free cash flow yield in the previous example, in addition to stocks filtered out independently using a second valuation metric, such as a predetermined number of stocks with the highest sales yields. Alternatively, the second metric may be dependent on the first metric, so that the value subset 6 only contains stocks that meet both filtering criteria, i.e. a predetermined number of stocks with both the highest free cash flow yields and sales yields.

Since the value subset 6 is a representation of the stocks in the benchmark selection 2 that have the highest quality and best valuation, the value subset 6 is a basic QV portfolio 16. The value subset 6, or basic QV portfolio 16, can be optimised in the auxiliary third step 12.

In the auxiliary third step 12 of the QV portfolio construction method, the processor applies an equal weighting to the stocks in the value subset 6 to provide an equal-weight QV portfolio 18. The third step 12 is supplementary to the main method and can be optionally included for a more optimal QV portfolio 18. It is advantageous for the stocks to be weighted equally as this avoids any stocks dominating the portfolio and so the portfolio avoids being skewed. Equally weighting the stocks allows the full benefit of the QV portfolio construction method to be realised.

In the auxiliary fourth step 14 of the QV portfolio construction method, the processor periodically maintains the basic or equal-weight QV portfolio 16, 18 by repeating the first step 8 and second step 10, and optionally the third step 12, to rebalance the portfolio 16, 18 for an updated QV portfolio 20. The auxiliary fourth step 14 may optionally include reinvesting the received dividends to gain further benefit from the QV portfolio construction method. As with the third step 12, the fourth step 14 is supplementary to the main method and can be optionally included to keep the QV portfolio 18, 16, 20 up-to-date for reduced risk. The fourth step 14 would not be carried out on the first occasion that the QV portfolio construction method is used, but it is preferable to carry out the fourth step 14 at least annually to maintain the QV portfolio 16, 18, 20. Repeating the steps of the method to maintain the QV portfolio 16, 18,20 incurs transaction costs, so the periodic maintenance is preferably performed at most on a monthly basis. In other embodiments, the periodic maintenance is performed no more than quarterly.

Using the S&P 500 as benchmark, a QV portfolio 20 constructed using the QV portfolio construction method, including all four steps 8, 10, 12, 14, was backtested over a full 18-year period from 1 January 1999 to 31 December 2016. The backtest results are provided in Table 1 with the S&P 500 used as the overall universe and comparable benchmark.

Table 1

Portfolio Total Annualised Maximum Sharpe Standard Return Return Drawdown Ratio Deviation QV 903.17% 13.67% -41.28% 0.75 16.65% S&P 500 154.10% 5.32% -55.42% 0.29 15.13% Table 1 shows that over the 18-year period the total return for a QV portfolio 20 constructed using the method of the present invention is over five and a half times better than the total return for the S&P 500 index. The annualised return is also more than two and a half times better for the QV portfolio 20 than the S&P 500. These results therefore show that the test performance of the QV portfolio 20 was significantly better than the S&P 500.

The maximum drawdown for the QV portfolio 20 is also better than that for the S&P 500, showing that the QV portfolio 20 has a lower financial risk. The higher Sharpe ratio of the QV portfolio 20 shows that the QV portfolio 20 has the better risk-adjusted return, and is thus the better portfolio. The standard deviation column shows that the QV portfolio 20 is slightly more volatile than the S&P 500, suggesting a slightly higher risk. However, the higher standard deviation is offset by the other metrics, such as total return and maximum drawdown, and so a valid conclusion can be drawn that the QV portfolio 20 performed better than the S&P 500 in the backtest. Therefore, the QV portfolio construction method has been shown in the backtest to provide valuable investment advantages.

The following is an example pseudocode listing to show how the QV portfolio construction method could be programmed onto a computing device. Each step of the process, for execution by the processor of the computing device, will be explained in detail below.

Function QI/bBenchmarkl- # Set global variables Threshold = 100% IndexSize = Size[Benchmark] IntersectSize = IndexSize FinalPortfolioSize = 40 it Step 1: select the quality subset While IntersectSize > 1/3 IndexSize do: it Filter by profitability Set[Profitable] = All stocks in [Benchmark] where AVG5YrsROIC is in top (Threshold) AVG5YrsROIC # Filter by operating efficiency Set[Efficient] = All stocks in [Benchmark] where GrossMargin5YrAvg is in top (Threshold) GrossMargin5YrAvg # Filter by leverage Set[Leverage] -All stocks in [Benchmark] where InterestCoverage is in top (Threshold) InterestCoverage # Collect stocks at the intersection of all filtered sets Set[QualitySubset] = [Profitable] 11 [Efficient] 11 [Leverage] IntersectSize = Number of stocks in [QualitySubset] Threshold = Threshold -1% End While; # Step 2: select the value subset from the quality subset If FinalPortfolioSize > 1/3 Size[QualitySubset] then: FinalPortfolio Size = 1/3 Size[QualitySubset] End If.

Set[FinalPortfolio] = FinalPorffolioSize stocks in [QualitySubset] having highest FCFYield # Step 3: construct the portfolio in equally weighted positions for each stock and # re-run this process at most on a monthly basis (with rebalancing and # reinvesting of received dividends) The following is a repetition of each step in the above pseudocode accompanied by a detailed description: Function QV[Benchmark] The process is set out in a function for constructing a QV portfolio 16, 18, 20 from a selected broad benchmark selection of stocks 2. The Benchmark variable represents the chosen benchmark selection 2. For example, QV(FTSE350) would construct a QV portfolio 16, 18, 20 from the FTSE350.

Threshold = 100% The Threshold variable represents the percentage used to filter the benchmark selection of stocks 2. Threshold is used to define the threshold values of the metrics for the filtering stages 22, 24, 26. A lower Threshold value means fewer stocks in the quality subset 4. Initialising this variable at 100% means that all of the stocks in the benchmark selection 2 will be selected for the quality subset 4 in the first pass of the while loop. The Threshold value can be gradually decremented with each pass of the while loop to reduce the size of the quality subset 4 (IntersectSize) until the desired size is reached. The pseudocode is set up so that one Threshold value is set for all three filtering stages 22, 24, 26, however the process can alternatively use individual variables to specify separate percentages for each filtering stage 22, 24, 26.

IndexSize = Size/Benchmark] The IndexSize variable represents the number of stocks in the selected benchmark selection of stocks 2 (Benchmark). Index-Size is used to provide the reference value for the condition of the while loop.

IntersectSize = IndexSize The IntersectSize variable represents the number of stocks in the quality subset 4 formed by the intersection of the profitable, efficiency and leverage sets 28, 30, 32 of stocks selected by the filtering stages 22, 24, 26. IntersectSize is used as the test value for the condition of the while loop and is initialised as the IndexSize in order to satisfy the condition to enter the while loop.

FinalPortfolioSize = 40 The FinalPortfolioSize variable is used to define the desired target number of stocks in the final QV portfolio 16, 18, 20. For sufficient diversification within the QV portfolio 16, 18, 20 this should be set to at least thirty, but is optimally set to at least forty. FinalPortfolioSize is used to determine the number of stocks selected to form the value subset 6 in Step 2.

# Step 1: select the quality subset 4 While IntersectSize > 1/3 IndexSize do: End While; This while loop causes the filtering stages 22, 24,26 in Step 1 to repeat if the quality subset 4 created by the intersection of the profitable, efficiency and leverage sets 28, 30, 32 is too large, that is, if IntersectSize is too high. In embodiments, the number of stocks in the quality subset 4 (lntersectSize) is no more than one-third of the number of stocks in the benchmark selection 2 (IndexSize). As mentioned above, for very large benchmark selections 2 it may be necessary to reduce the number of stocks by more than two-thirds. Therefore, in other embodiments, a fraction smaller than one-third can be used to provide the condition of the while loop. Alternatively, for the desired minimum size of the final QV portfolio 16, 18,20 to dictate the extent of the reduction, 1/3 IndexSize can be replaced by a function of the target number of stocks in the final QV portfolio 16, 18,20 (FinalPortfolioSize), for example 3*FinalPortfolioSize, or another multiple of FinalPortfolioSize With each iteration of the while loop, the filtering stages 22, 24,26 are repeated with a decreased Threshold value in order to decrease the number of stocks filtered out at each stage. The while loop terminates once the number of stocks in the quality subset 4 reaches or falls below one-third (or other fraction as necessary) of the number of stocks in the benchmark selection 2, or below a value that is a function of the target number of stocks in the final QV portfolio 16, 18, 20, that is, once IntersectSize reaches or falls below the specified fraction of IndexSize or the value of a function of FinalPortfolioSize.

# Filter by profitability Set[Profitable] = All stocks in [Benchmark] where AVG5YrsROIC is in top (Threshold) AVG5YrsROIC The most profitable stocks are selected from the benchmark selection 2. [Profitable] represents the filtered set 28 of stocks in the benchmark selection 2 (Benchmark) with the highest five-year average returns on invested capital (AVG5YrsROIC). This filtered set 28 represents the most profitable stocks in the benchmark selection 2. Threshold specifies the cut-off value relative to the AVG5YrsROIC values for all of the stocks in the benchmark selection 2. For example, if Threshold equals 75%, the [Profitable] set 28 will contain all of the stocks with a five-year average ROIC in the top 75% of five-year average ROICs for the benchmark selection of stocks 2. Other profitability metrics, such as those described above, can be used in place of the five-year average ROIC.

# Filter by operating efficiency Set[Efficient] = All stocks in [Benchmark] where GrossMargin5YrAvg is in top (Threshold) GrossMargin5YrAvg The most efficient stocks are selected from the benchmark selection 2. [Efficient] represents the filtered set 30 of stocks in the benchmark selection 2 (Benchmark) with the highest five-year average gross margins (GrossMargin5YrAvg). This filtered set 30 represents the most efficient stocks in the benchmark selection 2. Threshold specifies the cut-off value relative to the GrossMargin5YrAvg values for all of the stocks in the benchmark selection 2. For example, if Threshold equals 75%, the [Efficient] set 30 will contain all of the stocks with a five-year average gross margin in the top 75% of five-year average gross margins for the benchmark selection of stocks 2. Other operating efficiency metrics, such as those described above, can be used in place of the five-year average gross margin.

# Filter by leverage Set[Leverage] -All stocks in [Benchmark] where InterestCoverage is in top (Threshold) InterestCoverage The least leveraged stocks are selected from the benchmark selection 2.

[Leverage] represents the filtered set 32 of stocks in the benchmark selection 2 (Benchmark) with the highest interest coverage ratios (InterestCoverage). This filtered set 32 represents the least leveraged stocks in the benchmark selection 2. Threshold specifies the cut-off value relative to the InterestCoverage values for all of the stocks in the benchmark selection 2. For example, if Threshold equals 75%, the [Leverage] set 32 will contain all of the stocks with an interest coverage ratio in the top 75% of interest coverage ratios for the benchmark selection of stocks 2. Other leverage metrics, such as those described above, can be used in place of the interest coverage ratio.

# Collect stocks at the intersection of all filtered sets Set[QualitySubset] = [Profitable] fl [Efficient] Fl [Leverage] QualitySubset represents the quality subset 4 of stocks at the intersection of the Profitable, Efficient and Leverage sets 28, 30, 32 of stocks. The stocks in the QualitySubset 4 are the stocks in the benchmark selection 2 which have been selected by all three of the profitability, efficiency and leverage filtering stages 22, 24, 26.

IntersectSize = Number of stocks in [QualitySubset] After the benchmark selection of stocks 2 has been filtered to produce the quality subset 4, the IntersectSize variable, initially set as the number of stocks in the benchmark selection 2 (IndexSize) for the first pass of the while loop, is re-evaluated as the number of stocks in the newly-created quality subset 4. The new IntersectSize value is used in the while loop condition to determine if the desired size of the quality subset 4 has been reached, or if the while loop is to be re-entered for the repetition of the filtering stages 22, 24, 26. With each pass of the while loop, the IntersectSize variable is reevaluated as the number of stocks in the most recently created quality subset 4.

Threshold = Threshold -1% The final step of the while loop is to adjust the Threshold value in preparation for the next iteration. The while loop iterates if the quality subset 4 contains too many stocks, so the size of the quality subset 4 (IntersectSize) is reduced by reducing the number of stocks filtered out of the benchmark selection 2 at each filtering stage 22, 24, 26. To reduce the number of stocks selected by the filtering stages 22, 24, 26, the Threshold value is reduced. In this pseudocode example, the Threshold value is decremented by 1% for every pass of the while loop. For example, if a Threshold value of 25% results in too many stocks in the quality subset 4, the while loop is repeated with a reduced Threshold value of 24%. If the adjusted Threshold value of 24% also results in too many stocks in the quality subset 4, the while loop is repeated with a Threshold value of 23%.

# Step 2: select the value subset from the quality subset If FinalPortfolioSize > 1/3 Size[QualitySubset] then: FinalPortfolioSize = 1/3 Size[QualitySubset] End If.

In Step 2, before the quality subset 4 is filtered to create the value subset 6, the value of FinalPortfolioSize is compared with the size of the set QualitySubset 4, and FinalPortfolioSize is reduced if it is too large. If the target number of stocks in the final QV portfolio 16, 18,20 (FinalPortfolioSize) is greater than one-third of the number of stocks in the quality subset 4 (QualitySubset), then the target number of stocks in the QV portfolio 16, 18, 20 is reduced to the value of one-third of the number of stocks in the quality subset 4. It is necessary for FinalPortfolioSize to be less than the size of QualitySubset because FinalPortfolioSize is the number of the best valued stocks that will be chosen from the quality subset 4 to form the value subset 6.

Set[FinalPortfolio] = FinalPortfolioSize stocks in [QualitySubset] having highest FCFYield The quality subset 4 is reduced using a valuation metric. FinalPortforto represents the set of stocks in the quality subset 4 with the highest free cash flow (FCF) yield. The FinalPortfolio set of stocks is the value subset 6 of stocks, which contains all of the stocks that will form the final QV portfolio 16, 18, 20. A method for selecting the stocks with the highest FCF yield is arranging the stocks of the quality subset 4 in order by FCF yield, and, starting from the highest FCF yield, inserting the stocks into the value subset 6 (FinalPortfolio) until the number of stocks in the value subset 6 reaches the value of FinalPortfolioSize. Other valuation metrics, such as those described above, can be used in place of the FCF yield. The FinalPortfolio subset 6 represents the stocks in the quality subset 4 that have the best valuation and the stocks in the benchmark selection 2 that have the best combination of quality and value. The FinalPortfolio subset 6 is the basic QV portfolio 16.

# Step 3: construct the portfolio in equally weighted positions for each stock and # re-run this process at most on a monthly basis (with rebalancing and # reinvesting of received dividends) The equal-weight QV portfolio 18 is constructed from the value subset 6 by equally weighting all of the stocks. The QV portfolio constructing function is executed periodically, but no more than monthly, in order to maintain the equal-weight portfolio 18 as an updated portfolio 20 by rebalancing and optionally reinvesting the received dividends.

Embodiments of the present invention have been described with particular reference to the examples illustrated. However, it will be appreciated that variations and modifications may be made to the examples described within the scope of the present invention.

Claims (28)

1. Claims 1. A method for constructing an investment portfolio from an initial selection of stocks, wherein the method is implemented on a computer in communication with a database of investment data for the initial selection of stocks, and the computer is programmed to execute the steps of the method, the method comprising: from the investment data, identifying one or more quality metrics and one or more valuation metrics for each stock in the initial selection of stocks; from the initial selection of stocks, identifying a first subset of stocks wherein for each stock each of the one or more quality metrics passes threshold quality values; and from the first subset of stocks, identifying a second subset of stocks wherein for each stock each of the one or more valuation metrics passes a threshold valuation value.
2. 2. The method of claim 1, wherein the one or more quality metrics comprise a first quality metric, a second quality metric and a third quality metric; and wherein identifying the first subset of stocks further comprises: identifying a first set of stocks wherein the first quality metric for each stock passes a first threshold quality value; identifying a second set of stocks wherein the second quality metric for each stock passes a second threshold quality value; and identifying a third set of stocks wherein the third quality metric for each stock passes a third threshold quality value.
3. 3. The method of claim 2, wherein identifying the first subset of stocks further comprises: identifying the first, second and third sets of stocks from the initial selection of stocks; and identifying the stocks at the intersection of the first, second and third sets of stocks; wherein the first subset of stocks comprises the stocks at the intersection.
4. 4. The method of claim 2, wherein identifying the first subset of stocks further comprises: identifying the second set of stocks from the first set of stocks; and identifying the third set of stocks from the second set of stocks; wherein the first subset of stocks comprises the third set of stocks.
5. 5. The method of any preceding claim, wherein the one or more quality metrics are measures of one or more of profitability, operating efficiency, and leverage.
6. 6. The method of any preceding claim, wherein the one or more quality metrics are one or more of the following: return on equity, return on assets, return on invested capital, or return on capital employed.
7. 7. The method of any preceding claim, wherein the one or more quality metrics are one or more of the following: gross margin, post-tax margin, pre-tax margin, and profit margin.
8. 8. The method of any preceding claim, wherein the one or more quality metrics are one or more of the following ratios: long-term debt to total capital, interest coverage, long-term debt to equity, financial leverage, total debt to equity, and debt to assets.
9. 9. The method of any preceding claim, wherein the one or more valuation metrics are one or more of the following: enterprise value to gross profit ratio, free cash flow yield, price to earnings ratio, and price to sales ratio.
10. 10. The method of any preceding claim, wherein the stocks in the second subset are weighted equally.
11. 11. The method of any preceding claim, wherein the first subset contains no more than one-third of the number of stocks in the initial selection.
12. 12. The method of any preceding claim, wherein the second subset contains no more than one-third of the number of stocks in the first subset.
13. 13. The method of any preceding claim, wherein the second subset contains at least thirty stocks.
14. 14. The method of any preceding claims, wherein the method is repeated no more than monthly.
15. 15. A computing system for constructing an investment portfolio, the computing system comprising: a receiver configured to receive investment data for an initial selection of stocks from an investment database; a storage device configured to store the investment data; a processor programmed to execute the steps of: from the investment data, identifying one or more quality metrics and one or more valuation metrics for each stock in the initial selection of stocks; from the initial selection of stocks, identifying a first subset of stocks wherein for each stock a first quality metric of the one or more quality metrics passes a first threshold quality value; and from the first subset of stocks, identifying a second subset of stocks wherein for each stock one or more valuation metrics passes a threshold valuation value.
16. 16. The system of claim 15, wherein the one or more quality metrics comprise a first quality metric, a second quality metric and a third quality metric; and wherein, in being programmed to execute the step of identifying a first subset of stocks, the processor is further programmed to execute the steps of: from the initial selection of stocks, identifying a first set of stocks wherein the first quality metric for each stock passes a first threshold quality value; from the initial selection of stocks, identifying a second set of stocks wherein the second quality metric for each stock passes a second threshold quality value; and from the initial selection of stocks, identifying a third set of stocks wherein the third quality metric for each stock passes a third threshold quality value.
17. 17. The system of claim 16, wherein, in being programmed to execute the step of identifying a first subset of stocks, the processor is further programmed to execute the steps of: identifying the first, second and third sets of stocks from the initial selection of stocks; and identifying the stocks at the intersection of the first, second and third sets of stocks; wherein the first subset of stocks comprises the stocks at the intersection.
18. 18. The system of claim 16, wherein, in being programmed to execute the step of identifying a first subset of stocks, the processor is further programmed to execute the steps of: identifying the second set of stocks from the first set of stocks; and identifying the third set of stocks from the second set of stocks; wherein the first subset of stocks comprises the third set of stocks.
19. 19. The system of any of claims 15 to 18, wherein the one or more quality metrics are measures of one or more of profitability, operating efficiency, and leverage.
20. 20. The system of any of claims 15 to 19, wherein the one or more quality metrics are one or more of the following: return on equity, return on assets, return on invested capital, or return on capital employed.
21. 21. The system of any of claims 15 to 20, wherein the one or more quality metrics are one or more of the following: gross margin, post-tax margin, pre-tax margin, and profit margin.
22. 22. The system of any of claims 15 to 21, wherein the one or more quality metrics are one or more of the following ratios: long-term debt to total capital, interest coverage, long-term debt to equity, financial leverage, total debt to equity, and debt to assets.
23. 23. The system of any of claims 15 to 22, wherein the one or more valuation metrics are one or more of the following: enterprise value to gross profit ratio, free cash flow yield price to earnings ratio, and price to sales ratio.
24. 24. The system of any of claims 15 to 23, wherein the stocks in the second subset are weighted equally.
25. 25. The system of any of claims 15 to 24, wherein the first subset contains no more than one-third of the number of stocks in the initial selection.
26. 26. The system of any of claims 15 to 25, wherein the second subset contains no more than one-third of the number of stocks in the first subset.
27. 27. The system of any of claims 15 to 26, wherein the second subset contains at least thirty stocks.
28. 28. The system of any of claims 15 to 27, wherein the processor is programmed to execute the steps no more than monthly.