Traditional wealth managements, be they funds, open-end collective investments or personalized asset management lines, usually have as a primary goal to follow the benchmark
performance (although aiming at far better results), which can be easily achieved by a private investor through the mere holding of an ETF replicating the index itself, with significantly lower costs.
This happens because in increasingly dynamic and reactive markets, it is very difficult to develop a sufficiently stable medium/long term investment strategy. Investors prefer not to deviate from the benchmark in order to eliminate the operational risk, but in this way the market risk remains unchanged and remains high.
Up-Down Forecast instead is aimed at counteracting market risk through a methodology that systematically controls it through specific models designed to contain it at predefined intervals. One of the Up-Down Forecast parameters that controls the risk is the maximum number of securities that can be held at a given time by the system. An example: In the Italian stock market, with an actually operating model, it is possible to contain the risk within the benchmark risk by moving only five securities simultaneously in the period from 2012 to today (June 2017).
Up-Down Forecast uses decision-making techniques that are completely different from those inspired by traditional technical analysis. Being inspired by the theory of topos-theoretic bridges, it is forced to search for invariants in the seemingly chaotic context of the stock market and to use them to build “bridges” between different stocks that at different times have the same invariant. The management is based on switching through these “bridges” so that a stock displaying an invariant until a certain moment is replaced by another stock displaying the same invariant from that exact moment on. As in the case of the theory of topos, many relevant invariants identified by the system through statistical analysis on historical series have a geometric/topological nature. They are provisionally effective as they are able to capture some essential aspects of a stock at a given time in a qualitative way.
Let us use a metaphor to understand the meaning of the paradigm change that allowed us to watch the markets in a completely new way. Let us assume we are observing the movement of the planet Mercury from the Earth: we will see a very strange path, very difficult to understand. Now let us observe the same movement from the center of the Sun: magically the path of Mercury seems to be very regular and precisely an elliptic path. A simple change in the point of view allowed us to extract knowledge from the apparent chaos. Simplifying we can say that reality often appears to us as a 'dress' that hides its features. Likewise, the Stock market is a meeting place for multiple, often confusing and contradictory instances and it appears to us as a chaotic, non-regular phenomenon. However, if we change the point of view, to identify the relational complexity of the phenomenon, we can see the invariants that allow us to 'strip' reality so that we can see its features. Up-Down Forecast comes from a concrete implementation of this philosophy.
The expert system can be useful to anyone who has to manage mainly stock equity. EOS is aimed at institutional companies (Banks, IMCOs, Investment intermediaries, etc.) or companies interested in optimizing their financial investments as well as individuals with significant assets. Small investors can benefit from Up-Down Forecast indirectly through companies that have a partnership with EOS.
Analysts usually try to identify the potential of a given stock or of a group of stocks through an evaluation based on methodologies that can be very different from one another. Unfortunately, the accuracy of such evaluation can be seen with large delay because the interest of investors moves quickly form one stock to another or from a group of stocks to another group, often leaving the once promising stock dormant. From our analysis, it is apparent that the significant movements in stocks take place in very rapid and concentrated periods. We could say that a stock stays dormant 90% of the time attracting investors’ attention only in short periods that are hard to foresee. Waiting for the relevant performance event becomes therefore unworkable. On the contrary, Up-Down Forecast watches thoroughly the market and switches from one stock to another trying to identify those stocks investors are interested in. This way, there can be errors (up to 50% of the times), but it’s ready to promptly let go of stocks that aren’t ready yet, collecting very limited amount of losses with respect to the significant gain due to confirmed signals. The amount of earnings often exceeds 4 or 5 times the loss rate, and this guarantees relevant profits over time, even in a very variable context.
The system does not aim at duly comparing with benchmark performances, and in certain periods, it can diverge significantly (both in positive and in negative). This is due to the choice of holding few stocks simultaneously and the willingness to reduce handling. The primary goal is to exceed the benchmark performance, even significantly, when the markets allow it, but not in short periods of time, when strong deviations can also occur.
The system intends to optimize the performance/risk ratio, within a pattern that sets the maximum risk the client is willing to take. If this risk is equal to the stock benchmark, the optimal choice is to operate simultaneously on 5 stocks. If the client prefers to reduce the risk by accepting lower expected performance, a more 'open' template can be used, for example by operating simultaneously on 10 or more stocks.
The first assumption to seize a trend on a stock is that this trend exists. Sometimes there are periods of political or economic uncertainty, which bring investors to wait on the side-lines. In these situations, stocks move in a turmoil zone, with no significant trend, and the system accumulates losses and earnings, without having any chance to follow non-existent trends. In summary, we learned from experience that the system has a significant advantage over the medium/long term benchmark, both in euphoric and depressed markets, while it just swings around the benchmark in trendless times.
Up-Down Forecast makes use of two methods to evaluate the effectiveness of its models relative to the risk. The first method is based on the calculation of the standard
deviation of its equity line. The goal is not to deviate too far from the values of that parameter relative to the benchmark; deviations can only be tolerated with specific models intended for
those who agree to risk potential capital losses in short periods in order to obtain higher returns in the long run. The second method is more empirical, thus more understandable, and it is based
on the maximum drawdown. This is the measure of the maximum losses suffered over time by a hypothetical investor who follows the system. Conventionally, we deal with the maximum five-year drawdowns,
although it may be useful to have the absolute maximum draw in the long run. For example, in the back test period (1997-2011) the model applied on the big/mid capped stocks of the Italian market the
maximum drawdown of the equity line was 33% as opposed to the maximum drawdown of the FTSE All-Shares index, which was 69,71%. The advantage is still significant during the operating period
(2012- June 2017): about 23% as opposed to 35% of the index.
A useful tool that helped us to fully understand the risk exposure of the operating models used by the expert system are the statistic tables reporting the probability of exposure to a gain and that of exposure to a loss of a defined percentage within one year of the investment. These tables answer questions like these ones: what are the odds of losing 10% of the asset within a year? Or gaining 10%? As of today the answer to the first question regarding the Italian market in the period starting in 2012 is 4.36% (as opposed to 18.94% of the benchmark); while the answer to the second question would be 73.57% as opposed to 44.04% of the benchmark.
The major variables are as follows:
Up-Down Forecast is a quantitative system, and as such, it seems incompatible with the management that is based on fundamental analysis. However, its extreme generality inspired
by the theory of topos-theoretic bridges and its modularity allow to build alternative models, which can be used by managers who base their work on fundamental analysis. The compatibility is enhanced
by the average number of operations, which is perfectly compatible with that of fundamental analysts. A first approach in building such a model, which we can define 'mixed', could be based on letting
the managers decide what and when to purchase and let the sales be decided by the system. It can evaluate in a more objective way the outcome of the choice made by the operators, supporting it or
interrupting it in the event of manifest lack of interest by the market on the selected stocks. Another model could use the input provided by the fundamental managers through a questionnaire, to
quantify their guidelines and thus provide the system with an objective choice criterion while respecting the autonomy of the managers. In both cases, the traditional managers would have a useful
tool capable of controlling the risk of their choices.
In these cases, we could use the term 'semi-quantitative' models.
So far, EOS has never tested such hypothetical models, especially because it is impossible to evaluate the effectiveness in the past. However, it is likely that the human management experience together with the versatility of the Up-Down models could lead to high profitability.
The Up-Down Forecast system was created in an abstract way, thus it is adjustable to any market and it is capable to perform tests even on very long periods. Its operation is programmed through 'styles' that define the decisional rules based on the size and characteristics of the reference basket (the set of stocks on which the system operates its choices), on the desired level of risk and on operational procedures including those regarding the money management to control the risk of concentration on managed stocks, as well as other things.
The setting of the parameters is the most difficult phase in optimizing the system. We face an immense space of dependent and/or independent variables looking for the parameters that are sufficiently abstract to maintain adequate profitability even outside the test period. In doing this, we have always been guided by the intuition and the experience gained in years of research spent to optimize the previous versions of the system. However, recently, we realized that this activity was becoming extremely costly; EOS has therefore devised a tool that completely automates this research by delegating the choices about the parameters to a sophisticated mathematical model that operates autonomously using artificial intelligence techniques. The complexity of this tool is much greater than that of the decision-making process that is the basis of Up-Down Forecast and it is essential to substantially reduce the subjective impact of the tester. The results soon became apparent: even at the initial stage of experimentation, we are getting very significant results that encourage us to continue in the automation of the entire decision-making process, including the evaluation of the criteria used for optimization.
No, especially when considering models with a maximum number of stocks held very low (<7). In these cases, in particular moments, a larger number of stocks would satisfy the requirements to be purchased, but only a fraction of them will be accepted into the system. The other ones could be part of alternative models, as profitable as the main one. Being able to have alternative models can become a strategic element to reduce the simultaneous impact on the market given by too many purchases/sales on the same day.
Up-Down Forecast has always been used and optimized for the Italian stock market but it is applicable to any market. In order to verify the robustness of the models in different geographical and temporal contexts, numerous tests have been conducted over time on the French and American markets obtaining very satisfactory results.
Yes, Up-Down Forecast would be applicable also to funds. As of today, however, we have not yet conducted analysis on investment funds, because these financial instruments lack of open price, as well as high and low prices, and this lack of data would adversely affect the reliability of the operating strategies used by the system. We tested the system on ETFs with very good results, not as good as those we obtained with the stocks, but always higher than the benchmark. This happens because ETFs are usually built as an average of different items, and this way the underlying trends are dulled and thus difficult to identify.
Yes, of course. EOS has set up a complex software platform that enables the system to operate simultaneously in different contexts as if it were working on a single market. No test has been conducted yet, but certainly the results will only confirm and probably outperform those achieved on individual markets, as with a greater basket in which the system chooses the stocks, the greater are the chances of doing well.
Yes, but with bond markets it is not possible to catch the peculiarities of identifying short-term trends, that are a crucial factor to ensure a high prospective profitability. To date, EOS has not experienced any pattern on bond markets.
Yes, of course. To date, it is sufficient to schedule the system to manage the equity component according to the model chosen, leaving the traditional operator to force the system to operate on the bonds he chooses, according to a semi-quantitative model. In the near future, it is certainly desirable to create a wholly quantitative model, which is completely autonomous, which simultaneously manages the stock and the bond.
No. The starting point is always very delicate and it can affect the overall return on investment for several months. However, Up-Down Forecast is conceived to recover as early as possible any initial drops due to the market or withdrawals due to the natural outflows of the system from the benchmark. The investments based on our system must necessarily be medium to long-term (3-5 years) so that it can best adapt to the volatility of stock markets by absorbing the effects of less fortunate times.
Not necessarily, though timeliness has a primary role. In any case, two to three days delay is still compatible with adequate profitability.
Usually, we wait for the official closing of markets in order to process the definitive data of the day. If necessary, temporary processing can be obtained at any time of the day, but in that case, the signals could suggest, at closing time, different investment choices also incompatible with those resulting from the provisional data.
It depends on the pattern applied to the model. Patterns that suggest very few operations (less than 10 per year) can cause a very variable profitability, thus they must be considered riskier than others; while patterns with many operations (more than 40 per year) are less risky but they must suffer a very significant impact on the market and this could also reduce considerably the real profitability. Experience tells us that a compromise is needed and that this choice depends largely on the reference market as well.
Up-Down Forecast only uses opening, closing, high and low prices as well as daily volumes, and sometimes provisional prices when spot simulations are required. However, the system is modularly open to receive any kind of information about the stocks being studied and to consider them within the mathematical model adopted. Each stock is associated with a specific table, which collects relevant information about it, which can be discretionally used, if needed. In this sense, a semi-quantitative model could effectively use this support.
Yes, but only in emergencies and they are almost never used. The system is able to detect the decreasing interest on a stock long before any collapse occurs.
The first tested models implemented, for simplicity, a feature involving the total disinvestment of a single stock in the presence of a sale signal. Recently new models were tested, that increase or decrease the amount invested in order to optimize the performance through money management. To this end, some of these models use a modified form of Optimal-f, which has proven to be capable of controlling very well the risk of centrally capitalizing on a few stocks.
It depends on the level of confidence the user has in the ability of the system to continue to grow its assets. Disinvesting without an appropriate and specific return strategy
doen not make sense. Moreover, anyone with such a strategy would probably have better predictive or managerial capabilities than the system itself and could therefore do better without it.
A great profitability achieved by the system should not be seen as if it had entered a bubble that would justify the total or at least a significant disinvestment. Bubbles are characterized by abnormal growth in static contexts. Up-Down Forecast investments, on the other hand, are not static because they systematically seek new profitable opportunities that the statistic proves to be frequently present with relative continuity in any market.