Investment / Trading systems are computer programs that send
orders in the capital market. It's technology architecture is designed to
support a decision system that sends signals to enter and exit on asset positions
(buy and / or sell).
What ensures investment efficiency is the
ability of these systems on making profit in the long run with the maximum
possible security. The big question when we consider these systems is the
ability to evaluate the performance and rank it as sufficient to ensure
efficient management.
That is, put another way, how to separate
systems that lack credibility on signals generating of which represent a
significant result.
The line between this models classification can be tenuous,
several models with varying degrees of advancement, may present near this
classification.
While many models do not have a significant
degree of robustness, and in fact many do not have, yet can be managed with
relative efficiency. Many systems maintain a good performance in certain
periods, even being composed of models with poor complexity in decision making
or optimization methodologies that may not necessarily increase the actual
performance [Backtesting: good guy or bad guy?].
The capital market is an environment where
competitiveness and the search for implicit information is what motivates the
interaction of speculative capital. There is no mathematical way to determine
whether an investment system will always win, ie, it is impossible to say formally
that a particular strategy is profitable in the long run.
However it is possible to evaluate the
performance on different ways. You can change the operating conditions to
assess the sensitivity and behavior in adverse situations and stressful
situations. It is possible to evaluate* in depth the patterns of prices and
financial returns (* evaluate means "to understand" or model the
past). It is possible to classify certain time intervals and the scale of
certain asset price as repetitive with some significant degree.
You can monitor a change of state. However the change of state (the state classified by the model, like specific volatility or return level) may be apparent, masked by noise, characteristic of financial data.
You can monitor a change of state. However the change of state (the state classified by the model, like specific volatility or return level) may be apparent, masked by noise, characteristic of financial data.
After performing all these steps and analyze
the behavior of the system, a more realistic notion of risk is built. That is, we can win on knowing about repetitive
patterns that present possibility of gain - buy cheap and sell expensive, such
as arbitration and trend patterns.
These patterns, of course, are not repeated
regularly. This characteristic (the fact of the lack of regularity /
unpredictability with which the patterns are repeated) highlights that
financial movements have some degree of deterministic or chaotic properties.
Finally, you can monitor several models
simultaneously and manage them, looking if their behavior remains consistent
with that expected, looking if the indicators used to measure their performance
remains stable and finally determine your own limits of performance and risk.
Read related articles:
