|
New York, NY
Posts: 5 since Feb 2017
Thanks Given: 0
Thanks Received: 7
|
here is a code snippet that implements the Hurst Exponent in Mathematica with 2 different implementations
rets = Differences[Log[FinancialData["^GSPC",All,"Value"]]];
RescaledRange[List_,n_] :=
Mean[Map[Sqrt[n] (Max[Accumulate[#]]-Min[Accumulate[#]]) / Sqrt[Total[#^2]]&,
Map[#-Mean[#] &, Table[List[i,,i+n]], {i,1,Length[List]-n}]]]];
Hurst[data_,lag_] := h/.FindFit[Table[RescaledRange[data,n],{n,2,lag}], {c*x^h}, {c,h}, x];
Hurst[rets, 16]
Hursttq[data_,lag_] := Fit[Log@Table[RescaldRange[data,n], {n,2,lag}], {1,Log[x]}, x] [[2]] [[1]];
Hursttq[rets, 32]
anybody want to implement this code into C++ for SierraChart?
The idea is to calc Hust at every closing price for a range of lags (i think 16-->128 would be ideal), which yields 112 results of Hurst Exponents for each closing price. Then take the avg of those Hust exponents to give you an AvgHustStrength study.
Another option is to output all 112 results, each as a colored pixel on a vertical line. (so the study has a vertical line for every price bar...where each vertial line is a stack of colored pixels, based on the Hurst Exponents)
0 = Blue
0.5 =white
1 = Red
and in-between results are progressively lighter shades of blue or red.
So, a price series with high concentrations of Red = trend is becoming unstable and should be ending (signal to fade)...while a series with a high concentration of blue should be preparing to break out of the recent range (signal to go with a breakout).
|
