The Application Of Statistical Arbitrage Based On Garch Model On Futures

In the1980s, Morgan Stanley innovated pair trading stocks, from then on, researchers started to use various of quantitative methods in securities trading, and this caused glory times of statistical arbitrage However as time goes to21th century, statistical arbitrage faced huge challenges since it brings less and less profits, then people started to treat statistical arbitrage skeptically, even though it is undoubtedly that statistical arbitrage worth its existence.It is not a quite long time since statistical arbitrage was used and researched in our homeland, this essay is based on co-integration model and GARCH model analyzed soybean oil futures’and palm oil futures’ price series which start from January to October this year According to the results, the arbitrage based on equilibrium spread differences which derived from co-integration model and error correction model is better than the one based on market neutrality; the statistical arbitrage based on conditionally heteroscedastic variance which derived from GARCH model is better than the one based on constant variance on stability and dynamics, the arbitrage based on constant standard variance may bring high profits in the sampling series, but in the long run it will cause risks and increase uncertainties.The summary is that using co-integration model、 error correction model and GARCH model on two strongly related futures to practice statistical arbitrage will gain stable and low-risk profits. Using different standard variance as the signal of trading brings different profit. The statistical arbitrage based on a big data of standard variance brings more profits and decrease the times of arbitrage; instead, statistical arbitrage base on a small standard variance brings less profit and increase the times of arbitrage.Investor can adjust the transaction signal to gain stable profits according to different requires.