Today highly powerful statistical software are available that can be used to analyze and crunch any time series and make future predictions in the short term. Financial time series are all stochastic in nature. If we can successfully develop statistical models that can predict prices for pairs like EURUSD, GBPUSD, USDJPY, GBPNZD, USDCAD, USDCHF, NZDUSD, AUDUSD, GBPJPY etc, we can combine that statistical prediction with our technical analysis skills and see if it helps in improving our trading results. In this post we are going to discuss what are the most suitable models for the financial time series analysis and what have been the reported results in practically applying these models. We will be using ARIMA plus GARCH model. ARIMA stands for Auto-regressive Integrated Moving Average and GARCH stands for Generalized AutoRegressive Conditional Heteroskedasticity
There are many statistical software available in the market. Some are very expensive. But a few are FREE. One such software is R. R is a very powerful statistical analysis software that can be used easily in currency market analysis. You can download R from its site: The R Project For Statistical Computing. If we can use R to predict EURUSD H4 close prices for the next 10 candles with a high degree of accuracy, we can use these predictions with our trading system and see if this helps in taking the winrate above 90%. This is what we will be doing is downloading EURUSD H4 data and uploading it on R to make predictions for the next 20 candles. This post is going to explain in detail how we are going to do it. Downloading EURUSD H1 data from MT4 is easy. Click on Tools > History Center > EURUSD > H4. But more on this in later posts.
You should go through this PDF that explains in detail how the ARIMA plus GARCH method was applied to Apple Stock AAPL price. ARIMA plus GARCH modelling was able to predict AAPL price with 95% confidence level. Due to a negative earnings report released, stock price fell from $600 per share to $574 per share which was close to the lower limit predicted by this model. So you can see ARIMA plus GARCH modelling can predict pretty good results.
This is another good article that explains how to practically apply the GARCH model to the market data. According this post: A Practical introduction to GARCH modelling, the most suitable timeframe for this modelling is the daily timeframe.
According to this article, GARCH modelling on intraday timeframes may not be suitable due the presence of intraday seasonality in addition to the trend component. However we believe GARCH modelling can be done on H4 timeframes as we have sufficient time to do the calculations and remove the seasonality. The above article says the GARCH(1,1) is the best model as it gives fairly good results. Higher models require a lot more computations and maybe time consuming to implement.
This post claims that it successfully applied the ARMA plus GARCH model to S&P 500 index and got fantastic results. This is something interesting. We have decided to delve deeply on the topic of statistically modelling of the currency market data. If you can connect R software with MT4, you have a very powerful combination that you can use to do statistical modelling of currency price data in real time. This forex factory forum thread on how to connect R with MT4 should be helpful as well as useful.