The project we have been working on is about creating an automaton which is able to trade on the shares, index stock markets and Forex.
The automaton uses High-frequency trading (HFT) in the range of 1 minute.
The strategy is based on a data-mining process using a multidimensional decision tree. It uses four distinct parameters: the direction of the asset on the previous state, the correlation with another quotation, the volume of the asset trade on the interval that we are studying, and the volatility of the asset based on several intervals.
Each exit has an unique identifying number, a number of rise and fall, and the average payoff and loss. After we collect those data, we can calculate the expected payoff. If this payoff is high enough, that is to say after a certain value that we call “buying threshold”, the automaton will buy the asset, and the quantity will be determinate with the Kelly’s Formula.
Also, special automatons can trade in particular conditions: they can specialize on time period trading, on the volume when it reachs a high quantity of exchange, when the volality become high or low, and when the asset is correlated or decorrelated from its reference market.
Several parameters of the software’s interface are set by the user. First, the number of branches of the decision regarding the three parameters (the volatily,volume and the correlation) can be choosen to resize the tree at will.
After that the user has to inject some historical data from the market he wants to trade on. Then, the number of calculation's periods of the asset’s volatility and finally the buy threshold ( which is represented in percentage) that allows to determinate the expected payoff.
Once the automaton has loaded the historical data, the user has just to choose the period where he wants to trade on, then the automaton will make orders.
The objective is not only to beat the market, but also to have the best performance, meaning that if the market is down, the portfolio won’t be affected even if he doesn’t trade during the period.