ALGLIB – Numerical Analysis Library – library MetaTrader 4

Real author:

Sergey Bochkanov. ALGLIB project website – The library dates back to 1999.

ALGLIB is one of the largest and most complete mathematical libraries

Do you need to make a fast Fourier transform or to solve a differential equation system? Do you perform a complex data analysis trying to gather all methods in one place as a source code? Then ALGLIB library of numerical methods is for you!

ALGLIB is currently one of the best libraries of the multi-language algorithms. Below are ALGLIB features mentioned on the official website:

ALGLIB library is being constantly enhanced, new functions and improvements according to users’ comments are implemented regularly. The latest version is 3.6.0.

Besides, the library contains the large collection of test cases covering the major part of the proposed methods’ functionality. That will allow you to carry the tests and report detected errors to the project’s authors.

CAlglib class static functions should be used to work with the library – all library functions are moved to CAlglib system class as static functions.

testclasses.mq4 and testinterfaces.mq5 test cases scripts are attached together with a simple usealglib.mq4 demo script. Include files of the same name (testclasses.mqh and testinterfaces.mqh) are used to launch test cases. They must be placed to \MQL4\Scripts\Alglib\Testcases\.

Below is more detailed information about ALGLIB MQL4 ported library packages:

Code:

Library functions have detailed comments on their usage.

//+------------------------------------------------------------------+
//| Calculation of the distribution moments: mean, variance,         |
//| skewness, kurtosis.                                              |
//| INPUT PARAMETERS:                                                |
//|     X       -   sample                                           |
//|     N       -   N>=0, sample size:                               |
//|                 * if given, only leading N elements of X are     |
//|                   processed                                      |
//|                 * if not given, automatically determined from    |
//|                   size of X                                      |
//| OUTPUT PARAMETERS                                                |
//|     Mean    -   mean.                                            |
//|     Variance-   variance.                                        |
//|     Skewness-   skewness (if variance<>0; zero otherwise).       |
//|     Kurtosis-   kurtosis (if variance<>0; zero otherwise).       |
//+------------------------------------------------------------------+
static bool CBaseStat::SampleMoments(const double &cx[],const int n,double &mean,
                                     double &variance,double &skewness,double &kurtosis)
  {
//--- check
   if(!CAp::Assert(n>=0,__FUNCTION__+": the error variable"))
      return(false);
//--- check
   if(!CAp::Assert(CAp::Len(cx)>=n,__FUNCTION__+": length(x)<n"))
      return(false);
//--- check
   if(!CAp::Assert(CApServ::IsFiniteVector(cx,n),__FUNCTION__+": x is not finite vector"))
      return(false);
//--- create variables
   double v=0;
   double v1=0;
   double v2=0;
   double stddev=0;
//--- Init, special case 'N=0'
   mean=0;
   variance=0;
   skewness=0;
   kurtosis=0;
//--- check
   if(n<=0)
      return(true);
//--- Mean
   for(int i=0;i<n;i++)
      mean+=cx[i];
   mean/=n;
//--- Variance (using corrected two-pass algorithm)
   if(n!=1)
     {
      //--- calculation
      for(int i=0;i<n;i++)
         v1+=CMath::Sqr(cx[i]-mean);
      for(int i=0;i<n;i++)
         v2+=cx[i]-mean;
      v2=CMath::Sqr(v2)/n;
      variance=(v1-v2)/(n-1);
      //--- calculation
      stddev=MathSqrt(variance);
     }
   else
      variance=EMPTY_VALUE;
//--- Skewness and kurtosis
   if(stddev!=0)
     {
      //--- calculation
      for(int i=0;i<n;i++)
        {
         v=(cx[i]-mean)/stddev;
         v2=CMath::Sqr(v);
         skewness+=v2*v;
         kurtosis+=CMath::Sqr(v2);
        }
      //--- change values
      skewness=skewness/n;
      kurtosis=kurtosis/n-3;
     }
//--- successful execution
   return(true);
  }

CAlglib class static functions should be used to work with the library. Below is usealglib.mq5 function script’s source code for calculation of some statistical trading parameters:

//+------------------------------------------------------------------+
//|                                                    UseAlglib.mq4 |
//|                        Copyright 2012, MetaQuotes Software Corp. |
//|                                               |
//+------------------------------------------------------------------+
#property copyright "Copyright 2012, MetaQuotes Software Corp."
#property link      ""
#property version   "1.00"
#property strict
//+------------------------------------------------------------------+
//| Include libraries                                                |
//+------------------------------------------------------------------+
#include <Math\Alglib\alglib.mqh>
#include <Arrays\ArrayDouble.mqh>
//+------------------------------------------------------------------+
//| Script program start function                                    |
//+------------------------------------------------------------------+
void OnStart()
  {
//--- object for storage of profit/loss data for each order
   CArrayDouble *profit=new CArrayDouble;
//--- object for storage of balance data
   CArrayDouble *balance_total=new CArrayDouble;
//--- initial balance
   double        balance=0;
//--- historical data request
//--- get total amount ot orders
   int orders_total=OrdersHistoryTotal();
//--- get profit data
   for(int i=0;i<orders_total;i++)
     {
      if(OrderSelect(i,SELECT_BY_POS,MODE_HISTORY)==false)
        {
         Print("Error in OrderSelect (",GetLastError(),")");
         return;
        }
      int order_type=OrderType();
      //--- initial balance
      if(order_type==6) // OP_BALANCE=6
        {
         if(NormalizeDouble(OrderProfit()+OrderSwap(),2)>=0.0)
            if(balance==0.0)
               balance=OrderProfit();
        }
      //--- add order profit
      if(order_type==OP_BUY || order_type==OP_SELL)
        {
         double order_profit=OrderProfit()+OrderSwap()+OrderCommission();
         profit.Add(NormalizeDouble(order_profit,2));
         balance_total.Add(balance);
         balance=balance+NormalizeDouble(order_profit,2);
        }
     }
//---
   balance_total.Add(balance_total.At(balance_total.Total()-1)+profit.At(balance_total.Total()-1));
//--- copy balance data to the array of double type
   double arr_balance[];
   ArrayResize(arr_balance,balance_total.Total());
   for(int i=0;i<balance_total.Total();i++)
      arr_balance[i]=balance_total.At(i);
//--- copy profit data to the array of double type
   double arr_profit[];
   ArrayResize(arr_profit,profit.Total());
   for(int i=0;i<profit.Total();i++)
      arr_profit[i]=profit.At(i);
//--- linear regression
//--- number of independent variables
   int nvars=1;
//--- number of points
   int npoints=balance_total.Total();
//--- creating parameters matrix, used for linear regression
   CMatrixDouble xy(npoints,nvars+1);
   for(int i=0;i<npoints;i++)
     {
      xy[i].Set(0,i);
      xy[i].Set(1,arr_balance[i]);
     }
//--- variable for detecting calculations result (successful, unsuccessful)
   int info;
//--- class objects necessary for storing data on calculations
   CLinearModelShell lm;
   CLRReportShell    ar;
//--- arrays for storing regression results
   double lr_coeff[];
   double lr_values[];
   ArrayResize(lr_values,npoints);
//--- calculate linear regression coefficients
   CAlglib::LRBuild(xy,npoints,nvars,info,lm,ar);
//--- get linear regression coefficients
   CAlglib::LRUnpack(lm,lr_coeff,nvars);
//--- get the recovered linear regression values
   for(int i=0;i<npoints;i++)
      lr_values[i]=lr_coeff[0]*i+lr_coeff[1];
//--- calculation of Expected Payoff
   double exp_payoff,tmp1,tmp2,tmp3;
   CAlglib::SampleMoments(arr_profit,exp_payoff,tmp1,tmp2,tmp3);
//--- calculation of HPR array
   double HPR[];
   ArrayResize(HPR,balance_total.Total()-1);
   for(int i=0;i<balance_total.Total()-1;i++)
      HPR[i]=balance_total.At(i+1)/balance_total.At(i);
//--- calculation of the standard deviation and mean of the HPR
   double AHPR,SD;
   CAlglib::SampleMoments(HPR,AHPR,SD,tmp2,tmp3);
   SD=MathSqrt(SD);
//--- calculation of LR Correlation
   double lr_corr=CAlglib::PearsonCorr2(arr_balance,lr_values);
//--- calculation of LR Standard Error
   double lr_stand_err=0;
   for(int i=0;i<npoints;i++)
     {
      double delta=MathAbs(arr_balance[i]-lr_values[i]);
      lr_stand_err=lr_stand_err+delta*delta;
     }
   lr_stand_err=MathSqrt(lr_stand_err/(npoints-2));
//--- calculation of Sharpe Ratio
   double sharpe_ratio=(AHPR-1)/SD;
//--- print
   PrintFormat("-----------------------------------------------");
   PrintFormat("Linear regression: y = %.2fx + %.2f",lr_coeff[0],lr_coeff[1]);
//--- parameters
   PrintFormat("Expected Payoff = %.2f",exp_payoff);
   PrintFormat("AHPR = %.4f",AHPR);
   PrintFormat("Sharpe Ratio = %.2f",sharpe_ratio);
   PrintFormat("LR Correlation = %.2f",lr_corr);
   PrintFormat("LR Standard Error = %.2f",lr_stand_err);
   PrintFormat("-----------------------------------------------");
//--- delete objects
   delete profit;
   delete balance_total;
  }
//+------------------------------------------------------------------+

We will receive the following result (depening on results of your trading):