An automated speculative trading system proposal / Uma proposta para um sistema automatizado de tomada de decisão financeira especulativa

AUTOR(ES)
DATA DE PUBLICAÇÃO

2009

RESUMO

In this work we present the development of an automatic trading system based upon simplified time series models that, when applied on prices of liquid financial assets, derivatives and indexes, and with parameters adjusted by extensive back testing and reality conditions checks (Hansens SPA test), seeks to reach out financial returns that exceed the traditional Brazilian local benchmark, the CDI. The algorithm was applied upon 65 Brazilian stocks, 18 worldwide stock indexes, 9 currencies and 16 commodities (CRB Index), with time series ranging from may 1996 until may 2009, a 13 years approximated time stamp. With the aim of return maximization and not forgetting risk control, some of its parameters were calibrated, such as leveraging and stop-loss. Reality tests were applied in a way to identify and correct data mining and data snooping effects. The system was build to include 3 different ways to extract the trend component of the time series, 1) A simple moving average, 2) an exponential moving average and 3) a HP filter. The stop-loss usage showed to be an effective way to move asymmetrically to the right the distribution of daily returns, showing some evidence of criticism to the weak, strong and semi-strong forms of market efficiency. At the end, what we search is to identify the best tendency component parameters for each risk factor and allocate the investment decision for each asset in a way to profit maximize, with limitation of returns volatility and risk, always seeking data-mining control

ASSUNTO(S)

artificial inteligence análise técnica technical analysis tendência médias móveis economia trading systems artificial intelligence trends trading systems moving average

ACESSO AO ARTIGO

Documentos Relacionados