Transformations  in
Business & Economics

AN ALGORITHM-BASED STATISTICAL ARBITRAGE HIGH FREQUENCY TRADING SYSTEM TO FORECAST PRICES OF NATURAL GAS FUTURES
Kestutis Driaunys, Saulius Masteika, Virgilijus Sakalauskas, Mantas Vaitonis

ABSTRACT.Professional fund managers, investment banks and regulatory authorities raise the question about the impact of algorithmic trading on trading businesses, economy and market efficiency. The questions rise if high frequency trading (HFT) provides more efficient, liquid markets and is economically beneficial. In this paper an algorithm based on statistical arbitrage is tested. Statistical arbitrage is a well-known trading strategy where profit arises from pricing inefficiencies between correlated financial instruments. An algorithm of statistical arbitrage tries to find a pair of correlated instruments that move together and take long/short positions when they diverge abnormally, hoping that the prices will converge in the near future. Recent computational expansion in financial modeling and the ever increasing demand for order execution speed is moving market making and price discovery strategies into high frequency trading (HFT) or the milliseconds realm, where HFT already generates nearly 2/3 of the overall trading volume. In this report we apply high frequency data from NYMEX exchange to test a trading system based on statistical arbitrage in one of the most liquid futures market, i.e. natural gas futures. The overall results suggest that statistical arbitrage in HFT environment significantly outperforms traditional trading strategies, provides liquidity to the markets and denies the efficient market hypothesis.

KEYWORDS: efficient market hypothesis, high frequency trading, statistical arbitrage, pairs trading, futures market, algorithmic trading.

JEL classification: G14, G17, C63, B26.