In the first part of the paper, a universal method of algorithmic trading in stock markets, which provides the asymptotically highest profit as compared to any trading strategy that is not extremely complex, is proposed. The universal strategy is compared with the class of strategies that compute the amount of bought or sold units of a financial instrument made in each step of trading by means of continuous functions of input information. The universal strategy makes decisions based on the forecasts of future prices of financial instruments with the help of the randomized algorithm for computing well-calibrated forecasts in Dawid’s sense. The given algorithm is the development of the Kakade and Foster theory and is combined with the similar forecasting method suggested by V.G. Vovk. In the second part, the results of numerical experiments performed with the proposed universal algorithm and historical data on financial series are presented.
|Number of pages||15|
|Journal||Journal of Communications Technology and Electronics|
|Publication status||Published - 23 Jun 2015|
- algorithmic trading
- stock price
- well-calibrated forecasts