Author of the idea: Vladimir, mq5 code author: barabashkakvn.
The Expert Advisor uses Burg's method for linear prediction. Linear prediction is based on finding future values as linear functions of previous values. Suppose we have the x..x[n-1] price range where the higher index corresponds more recent prices. The prediction of the x[n] future price is calculated as follows:
here a[i=1..p] are model ratios, and p is the model order. Burg's method finds the a ratios by decreasing a mean-square error on the last training n-p bars.
- MaxRisk - maximum risk of all simultaneously performed deals.
- ntmax - maximum number of deals in one direction.
- MinProfit - minimal predicted profit at which positions will be opened.
- MaxLoss - maximal predicted loss at which positions will be closed.
- TakeProfit - Take Profit value.
- StopLoss - Stop Loss value.
- TrailingStop - the Trailing Stop function.
- PastBars - the number of previous bars used to predict future values.
- ModelOrder - the order of Burg's model as a fraction of the number of past bars (0..1).
- UseMOM - enables detrend of input data: mom(i)=log[p(i)/p(i-1)].
- UseROC - enables detrend of input data: roc=100*(p(i)/p(i-1)-1).
Only one of UseMOM and UseROC variables can be true, i.e. UseMOM=true AND UseROC=true is not allowed.
Like most of optimized Expert Advisors, Burg Extrapolator only works well on training bars. The Expert Advisor will steadily lose money without a constant re-optimization.