Zusammenfassung
The objective in this paper is to develop and
implement FGP-2 (Financial Genetic Programming) on
intra daily tick data for stock index options and
futures arbitrage in a manner that is suitable for
online trading when windows of profitable arbitrage
opportunities exist for short periods from one to ten
minutes. Our benchmark for FGP-2 is the textbook rule
for detecting arbitrage profits. This rule has the
drawback that it awaits a contemporaneous profitable
signal to implement an arbitrage in the same direction.
A novel methodology of randomised sampling is used to
train FGP-2 to pick up the fundamental arbitrage
patterns. Care is taken to fine tune weights in the
fitness function to enhance performance. As arbitrage
opportunities are few, missed opportunities can be as
costly as wrong recommendations to trade. Unlike
conventional genetic programs, FGP-2 has a constraint
satisfaction feature supplementing the fitness function
that enables the user to train the FGP to specify a
minimum and a maximum number of profitable arbitrage
opportunities that are being sought. Historical sample
data on arbitrage opportunities enables the user to set
these minimum and maximum bounds. Good FGP rules for
arbitrage are found to make a 3-fold improvement in
profitability over the textbook rule. This application
demonstrates the success of FGP-2 in its interactive
capacity that allows experts to channel their knowledge
into machine discovery
Nutzer