Forward-looking Moving Average Trading Model:Theory And Evidence

Moving average trading strategy is one of the most common analytical methods in technical analysis.The classic moving average trading strategy has gradually lost efficiency in recent years as financial markets become efficiently.This paper proposes a forward-looking moving average trading model by combining the forecasting model with the moving average trading strategy:ARMA-Moving Average Trading Model and BP Neural Network-Moving Average Trading Model.By comparing the price predicted by time series model or neural network model with the forward-looking price index,we can get the trading signal and buy or sell in advance,which can help us lock the loss of the income caused by the lag of transaction in classic moving average trading model.This paper uses 10-year government bond future market 1 minute high frequency data as sample.Then using ARMA(1,1)model and 5-5-1BP neural network model to perform one-step forward static prediction on the data sample.By comparing the predicted value with the forward-looking price index to obtain the trading signal in advance.This paper calculates the evaluation index of each trading model,such as cumulative yield,winning rate,maximum withdrawal rate and Sharpe ratio.Then compare the forward-looking moving average trading model with the classical moving average trading model to analyze the pros and cons.At the same time,this paper also discusses the effectiveness of the forward-looking moving average trading model in bull market,bear market and shock market.Empirical results indicate that the forward-looking moving average trading model can significantly improve the efficiency of the classic moving average trading model,wherever in bull market,bear market or shock market.Specifically,the ARMA-moving average trading model and the BP neural network-moving average trading model have higher cumulative yields,winning rates,Sharpe ratios,and lower maximum withdrawal rates.Secondly,as a whole,the BP neural network-moving average trading model performs better than the ARMA-moving average trading model.Since in a short period,the correlation between financial time series is significantly affected by nonlinear correlations such as white noise,and the BP neural network-moving average trading model can describe nonlinear relationship better when predicting prices,so it's more suitable for constructing high frequency trading strategies.