BE1031671B1 - Method and system of data processing for the selection of mutual funds - Google Patents

Abstract

A computer-implemented method for determining a mutual fund selection S from a set of mutual funds, and a data processing system comprising means for implementing said method, said method comprising: - acquiring market information of each of the mutual funds over a plurality of past periods; - calculating a performance measurement result of each of the mutual funds for each past period; - for each past period, determining a subset comprising the T % of mutual funds with the best performance measurement results; - determining the mutual fund selection S belonging to at least P % of the subsets of the plurality of past periods.

Description

+ BE2023/5461

Method and system of data processing for the selection of mutual funds

Technical field

[0001] According to a first aspect, the invention relates to a computer-implemented method for determining a mutual fund selection from a set of mutual funds. The invention also provides a data processing system comprising means for implementing said method.

State of the art

[0002] In the financial field, there is a need for a method and means for selecting mutual funds by machines. There is further a need for such a method and means for selecting mutual funds that can select the best or worst funds from a set of mutual funds, based on one or more different performance evaluation methods.

[0003] The invention aims to meet the above needs and proposes for this purpose a computer-implemented method for determining a selection of mutual funds from a set of mutual funds, as well as a data processing system comprising means for implementing said method.

Summary of the invention

[0004] One object of the present invention is to provide a computer-implemented method for selecting mutual funds in an automated manner.

[0005] According to a first aspect, the invention relates to a computer-implemented method for determining a selection S of mutual funds from a set of mutual funds and comprising steps of:

2 BE2023/5461 - acquiring market information of each of the mutual funds of the set of mutual funds over a plurality of past periods; - based on the market information, calculating a performance measurement result of each of the mutual funds of the set for each period of the plurality of past periods; - for each period of the plurality of past periods, determining a subset comprising the T % of mutual funds of the set with the best (respectively worst) performance measurement results; - determining the selection S of mutual funds belonging to at least P % of the subsets of the plurality of past periods.

[0006] The steps of the method define a probability of occurrence

P, a threshold T, and a function F(P,T) to determine the selection S=F(P,T) based on the probability of occurrence P and the threshold T.

[0007] Such a computer-executed method makes it possible to automate the selection of mutual funds from among all mutual funds.

[0008] In a preferred version, the method further comprises steps of: - for each of the mutual funds of the selection S, and for a given probability P, determining a minimum threshold Ti such that the mutual fund belongs to the selection S=F(P, Ti); - classifying the mutual funds of the selection S according to their respective thresholds Ti.

[0009] Such a method makes it possible to establish a classification of mutual funds, from best to worst, and this in an automated manner.

[0010] In a preferred version of the method: - the threshold T is between 1% and 50%, preferably equal to 5%, 10%, 20%, 25%; - the probability of occurrence P is between 50% and 99%, preferably equal to 75%, 80%, 85% or 90%.

3 BE2023/5461

[0011] In a preferred version of the method according to the invention, the performance measurement results of each of the mutual funds in the group are calculated over the plurality of past periods by: - a single calculation method, or - aggregation of a plurality of intermediate performance measurement results calculated according to different performance measurement methods.

[0012] Preferably, the performance measurement results of each of the mutual funds in the set are calculated over the plurality of past periods by aggregating a plurality of intermediate performance measurement results calculated using different performance measurement methods.

[0013] Such a method makes it possible to select and establish a classification of mutual funds on the basis of a plurality of different measurement methods.

[0014] In a preferred version of the method, the aggregation of the plurality of intermediate performance measurement results is a weighted linear combination of the intermediate performance measurement results.

[0015] Such a method makes it possible to vary the weight associated with each of the performance measurement methods in the performance measurement results of mutual funds.

[0016] In a preferred version of the method, the market information includes price histories of the mutual funds, histories of economic variables such as the inflation rate, category information of the mutual funds, and/or performance histories of at least one stock index.

[0017] According to a second aspect, the invention relates to a data processing system for determining a selection S of mutual funds from a set of mutual funds, and comprising means for implementing the steps of the method according to the invention.

[0018] Preferably, the data processing system according to the invention comprises:

* BE2023/5461 - a computing unit for receiving information on investment criteria, probability of occurrence P and threshold T; - a database server in communication with the computing unit and configured to store market information of a group of mutual funds comprising at least price history information and category information of each mutual fund of the group over a plurality of past periods;

[0019] The calculation unit is configured to: - receive the market information from the database server; - select a set of mutual funds from the group of mutual funds based on the investment criteria information and the category information; - based on the market information, calculate performance measurement results of each of the mutual funds of the set over the plurality of past periods; - for each period of the plurality of past periods, determine a subset comprising the T % of mutual funds of the set with the best (respectively worst) performance measurement results; - determine the selection S of mutual funds belonging to at least P % of the subsets of the plurality of past periods.

[0020] Such a data processing system makes it possible to automate the selection of funds from a set of mutual funds, based on a user's investment criteria. In particular, such a system makes it possible to select the funds that have the best (worst) performance measurement results in the past.

[0021] In a preferred embodiment, the data processing system further comprises a first user terminal for encoding investment criteria information by a user and transmitting the investment criteria information to the computing unit, the first user terminal preferably being a computer or a portable device such as a smartphone.

> BE2023/5461

[0022] In such a system, the user can encode the information himself according to his investment criteria.

[0023] Preferably, the computing unit is configured to communicate with the first user terminal and/or with the database server via the Internet.

[0024] Preferably, the calculation unit is configured to transmit output information representative of the selection S to the first user terminal.

[0025] Preferably, the first user terminal comprises: - a screen to allow the user to view said output information; - manual encoding means, such as a touch screen or a keyboard, to encode the investment criteria information by the user.

[0026] Preferably, the data processing system comprises a second user terminal for encoding the probability of occurrence P and threshold T information by a user and for transmitting the probability of occurrence P and threshold T information to the computing unit. Preferably, the second user terminal is a computer or a portable device.

Detailed description of certain embodiments of the invention

[0027] The invention relates to a computer-implemented method for determining a selection S of mutual funds from a set of mutual funds. For example, the method can be executed by a server, a personal computer, or by several machines forming a data processing system, exchanging information and/or operating in a network. The method makes it possible to determine a selection S of mutual funds from a set of mutual funds. For example, the selection S of mutual funds consists of the funds with the best or worst past performances.

[0028] The method comprises a step of acquiring market information. The market information is acquired for each of the funds

° BE2023/5461 mutual fund of the whole, and for each period of the plurality of past periods. Preferably, the past periods are successive and do not overlap. For example, the plurality of past periods consists of 10 periods of a duration of 6 months, successive, and spanning the last 5 years. It is also possible for the past periods of the plurality of past periods to overlap. For example, the past periods of the plurality of past periods each have a duration of 1 year and are offset by 6 months from each other. Market information over a plurality of past periods is acquired. For example, the market information includes price histories of the whole mutual funds, histories of economic variables such as the inflation rate, category information of the whole mutual funds, and/or performance histories of at least one stock index.

[0029] The method comprises a step in which a performance measurement result is determined for each of the mutual funds in the set, and for each period of the plurality of past periods. The performance measurement results are calculated on the basis of the acquired market information. For each period of the plurality of past periods, it is therefore possible to establish a ranking of the mutual funds according to their respective performance measurement results.

This helps determine the mutual funds with the best (worst) performance measurement results over each period.

[0030] An example of such a ranking is illustrated in Table 1 below. In Table 1, the 6 funds with the best performance results over each of the periods are presented. The 6 funds are selected from a set of 100 funds. The rank of each fund over a period is determined based on its performance measurement result over the period.

7 BE2023/5461 [Rank Year 1 Year 2] Year s[year 4

FondB |FondB [FondT |FondF 2 [Fonda [Fond [Fond [Fond 3 [Fond [Fond A [Fond |FonaH [Fona X [Fong [Fora [Fonda 5 Fond & [Fond [Fond W Fond

Table 1: Example of subsets of funds with the best performance measurement results, over 4 successive annual periods.

[0031] For each period of the plurality of past periods, the 5 Method comprises a step of determining a subset comprising the

T % of mutual funds in the set with the best (or worst) performance measurement results. In the example in Table 1, the threshold T is set to 6%. This means that the subset for each period includes the 6 funds among the 100 funds in the set with the best performance measurement results. For example, in Table 1, the 6 best-performing funds in year 1 are funds B, A, F,

C, X, R.

[0032] Next, the method comprises the step of determining the selection S of mutual funds belonging to at least P% of the subsets of the plurality of past periods. In the example of Table 1, the probability of occurrence P is set at 75%. This means that the selection S corresponds to the funds that appear in the top 6% of funds, and this for at least 75% of the past periods. In the example of Table 1, the funds that satisfy these criteria and are therefore part of the selection S are funds A, B, C, and F. These funds appear in fact in the top 6% of funds over 3 of the 4 years, or for 75% of the time. In other words, the probability of occurrence P of these funds in the top 6% of all funds over all past periods is greater than 75%.

[0033] Similarly, the method allows selecting the worst funds, instead of the best funds, from among the set of mutual funds.

3 BE2023/5461 placement. To do this, it is sufficient, when determining the subset for each period of the plurality of past periods, to select the T % of mutual funds from the set with the worst performance measurement results.

[0034] An example of such a method of selecting the worst mutual funds is illustrated in Table 2. [Rank [Year 1 [Year 2 [Year 3| Year 4 6 |FordH |FundL [Funds |FundQ as [Font [Fund [rounds Fund 99 |Fund@ Fund! [FundP |FundU 100 [Fong [Fona® [Fund A [Fund

Table 2: Example of subsets of funds with the worst performance measurement results, over 4 successive annual periods

[0035] In the example of Table 2, the set of mutual funds includes 100 funds, most of which are not represented. The threshold

T is set to 5%. This means that the subset of each of the periods (years 1, 2, 3, and 4) includes the 5 funds among the 100 funds in the set whose performance measurement results over the period are the worst. In Table 2, the 5 worst performing funds over year 1 are funds P, Q, , J, H.

[0036] Next, the method comprises the step of determining the selection S of mutual funds belonging to at least P% of the subsets of the plurality of past periods. In the example of Table 2, the probability of occurrence P is set at 75%. This means that the selection S corresponds to the funds that appear in the 5% of funds with the worst performance measurement results, and this for at least 75% of the past periods. In the example of Table 2, the funds that satisfy these criteria and are therefore part of the selection S are the funds P, Q, |, and J. These funds indeed appear in the 5% of the best funds over 3 of the 4 years, or for 75% of the time. In other words, the probability of occurrence P of

? BE2023/5461 these funds in the worst 5% of all funds over all past periods is greater than 75%.

[0037] In a preferred version of the method according to the invention, the mutual funds belonging to the selection S are classified according to their respective minimum threshold Ti. For each of the mutual funds of the selection S, and for a given probability P, the minimum threshold Ti of a fund corresponds to the minimum value of Ti such that the mutual fund belongs to the selection S=F(P, Ti).

[0038] For example, in the case shown in Table 1 and for a given probability P equal to 75%, fund B has a minimum threshold Ti equal to 2%. Indeed, the probability of occurrence of fund B in the best 2% of all funds over all past periods is greater than 75%.

Similarly, for the given probability P equal to 75%, fund F has a minimum threshold Ti equal to 3%, fund C has a threshold Ti equal to 4%, and fund A has a threshold Ti equal to 5%. This allows us to classify the funds of the selection S=F(P=75%,

T=6%) depending on their respective threshold Ti. In the example in Table 1, the order is: B, F, C and then A.

[0039] The skills of fund management teams (in terms of market timing or asset allocation) can be assessed using a wide range of performance measures. To assess fund performance accurately, it is advantageous to consider different methods for measuring mutual fund performance and compare them.

In one example of the method, the performance measurement results of each of the mutual funds in the set are calculated over the plurality of past periods by aggregating a plurality of interim performance measurement results calculated using different performance measurement methods. Each interim performance measurement result may thus be calculated using a different performance measurement method. This allows for different aspects to be taken into account in evaluating the performance of the funds.

[0040] For example, the aggregation of the plurality of intermediate performance measurement results is a weighted linear combination of the intermediate performance measurement results. The weighting coefficients of the

10 BE2023/5461 linear combinations can be chosen to give different weights depending on the relevance of each of the performance measurement methods. This allows to select the measures that best explain the variation in performance between the funds in the set, for example.

[0041] It is also possible to select performance measurement methods from a set of performance measurement methods, particularly when the latter includes redundant and/or correlated information. It is then possible to isolate or select the most informative performance measures. This reduced set represents what best explains the management skills of the funds over time. The selection of performance evaluation methods improves the subsequent ranking of funds by taking into account the information that best explains the cross-variation of returns. The selection can be carried out by combining the following tasks: - defining the set of measures appropriate to the sample of funds; - removing correlations; - performing a principal component analysis.

[0042] The invention makes it possible to evaluate the stability of a performance measure by: - ranking the performances of all funds over several past periods; - counting the occurrence of each fund among the best (and worst) performing funds during these periods.

The more stable the measure, the more it repeatedly detects a large number of the best (and worst) performing funds.

[0043] In summary, the stability model depends on two main parameters: - a threshold defining the fraction of the best (and worst) performing funds (e.g., the top 10%); - a probability defining the number of occurrences required to be considered the best (or worst) performing for a fund (e.g., being the best performing 50% of the time).

11 BE2023/5461

[0044] Another advantage of the stability approach is that it allows the best (and worst) performing funds to be ranked. Indeed, all positions or ranks of funds in previous rankings are recorded and compared to those of all peers. This comparison provides information on the management's ability to outperform (or underperform) its peers.

[0045] The invention also relates to a data processing system for determining a selection S of mutual funds from a set of mutual funds, and comprising means for implementing the steps of the method according to the invention.

[0046] The data processing system comprises a computing unit for receiving investment criteria information from a user, for example a financial analyst or an investor. For example, a financial analyst provides probability of occurrence P and threshold T values to the computing unit. Market information from which mutual fund performance measurement results can be calculated is stored on a database server in communication with the computing unit.

[0047] The computing unit may be configured to, in an automated manner - receive market information from the database server; - select a set of mutual funds from the group of mutual funds based on the investment criteria information and the category information.

For example, fund category information includes information about mutual funds such as: the type of financial assets under management, the industry sector and investment style, and the geographic distribution of investments.

[0048] The calculation unit can be configured to, in an automated manner, determine the selection S of mutual funds with the best or worst performances among all the mutual funds.

Claims (13)

12 BE2023/5461 Claims 1. A computer-implemented method for determining a selection S of mutual funds from a set of mutual funds and comprising steps of: - acquiring market information of each of the mutual funds of the set of mutual funds over a plurality of past periods; - based on the market information, calculating a performance measurement result of each of the mutual funds of the set for each period of the plurality of past periods; - for each period of the plurality of past periods, determining a subset comprising the T % of mutual funds of the set with the best (respectively worst) performance measurement results; - determining the selection S of mutual funds belonging to at least P % of the subsets of the plurality of past periods; the steps of the method defining a probability of occurrence P, a threshold T, and a function F(P,T) for determining the selection S=F(P,T) as a function of the probability of occurrence P and the threshold T. 2. Method according to claim 1, further comprising steps of: - for each of the mutual funds of the selection S, and for a given probability P, determining a minimum threshold Ti such that the mutual fund belongs to the selection S=F(P, Ti); - ranking the mutual funds of the selection S according to their respective thresholds Ti. 3. Method according to any one of the preceding claims, and wherein: 13 BE2023/5461 - the threshold T is between 1% and 50%, preferably equal to 5%, 10%, 20%, 25%; - the probability of occurrence P is between 50% and 99%, preferably equal to 75%, 80%, 85% or 90%. 4. A method according to any preceding claim, and wherein the performance measurement results of each of the mutual funds in the set are calculated over the plurality of past periods by: - a single calculation method, or - aggregation of a plurality of intermediate performance measurement results calculated according to different performance measurement methods. 5. A method according to the preceding claim, and wherein the aggregation of the plurality of intermediate performance measurement results is a weighted linear combination of the intermediate performance measurement results. 6. A method according to any preceding claim, and wherein the market information comprises price histories of the mutual funds, histories of economic variables such as the inflation rate, category information of the mutual funds, and/or performance histories of at least one stock index. 7. Data processing system for determining a selection S of mutual funds from a set of mutual funds, and comprising means for implementing the steps of the method according to any one of claims 1 to 6. 8. Data processing system according to claim 7 and comprising: - a calculation unit for receiving information on investment criteria, probability of occurrence P and threshold T; 14 BE2023/5461 - a database server in communication with the computing unit and configured to store market information of a group of mutual funds comprising at least price history information and category information of each mutual fund of the group over a plurality of past periods; the computing unit being configured to: - receive the market information from the database server; - select a set of mutual funds from the group of mutual funds based on the investment criteria information and the category information; - based on the market information, calculate performance measurement results of each of the mutual funds of the set over the plurality of past periods; - for each period of the plurality of past periods, determine a subset comprising the T % of mutual funds of the set with the best (respectively worst) performance measurement results; - determine the selection S of mutual funds belonging to at least P % of the subsets of the plurality of past periods. 9. A data processing system according to the preceding claim, further comprising a first user terminal for encoding investment criteria information by a user and transmitting the investment criteria information to the computing unit, the first user terminal preferably being a computer or a portable device such as a smartphone. 15 BE2023/5461 10. Data processing system according to the preceding claim, and where the computing unit is configured to communicate with the first user terminal and/or with the database server via the internet. 11. Data processing system according to one of claims 9 or 10, and where the calculation unit is configured to transmit output information representative of the selection S to the first user terminal. 12. Data processing system according to the preceding claim, wherein the first user terminal comprises: - a screen to allow the user to view said output information; - manual encoding means, such as a touch screen or a keyboard, to encode the investment criteria information by the user. 13. Data processing system according to the preceding claim, comprising a second user terminal for encoding the probability of occurrence P and threshold T information by a user and for transmitting the probability of occurrence P and threshold T information to the calculation unit, the second user terminal preferably being a computer or a portable device.