Overlay Positive Hedging EA

Quoting charvo

Disliked

DB, i think that the direction you head into is promising.

this EA seems more like a semi-auto EA.

i imagine a scenario like this:

i get up in the morning, check the market, decide EG is more likely to go up, then i turn on the EA to do "buy eu sell gu" = true;

leave it alone, by maybe US market close, the EA stops trading and closes all positions.

next day, i do the same shit.

as i guess, this is not a time-consuming thing to spend a couple of minutes everyday to turn on the EA. as i put 'red font'...

Ignored

Quoting charvo

Disliked

DB, i think that the direction you head into is promising.

this EA seems more like a semi-auto EA.

i imagine a scenario like this:

i get up in the morning, check the market, decide EG is more likely to go up, then i turn on the EA to do "buy eu sell gu" = true;

leave it alone, by maybe US market close, the EA stops trading and closes all positions.

next day, i do the same shit.

as i guess, this is not a time-consuming thing to spend a couple of minutes everyday to turn on the EA. as i put 'red font'...

Ignored

// format a string that presents the hedge ratios
   // to the user and that will be displayed in the plot
   // it will also multiply them with base_units ...
   ratios = "hedge ratios [multiples of Lots]n";
   for (i=0; i<pairs; i++){
      ratios = ratios + symb[i] 
      + " " + DoubleToStr(MathRound(base_units * coef[i]), 0) 
      + " (" +  DoubleToStr(GlobalVariableGet(GLOBALNAME+Symbol()+Period()+"_"+symb[i]),0) + ")n";
   }
   Comment(ratios);
   plot();   
}

Quoting iwjw

Disliked

checking the market...isn't that something the EA could do for you if the EA knew what to look for?
trading only in a specified time window should also be doable
so semi-automated isn't neccessarily where we have to end up with this ea

Ignored

void onTick(){
   int units, units1; 
   int i; 
   
   // update the last row
   if (!trend){
      for (i=0; i<pairs; i++){
         if (symb[i] != Symbol()){
            prices[i] = iClose(symb[i], 0, 0);
         }else{
            prices[i] = 0;
         }
      }
      Rv("current_others", prices);
      Rd("current_this", Close[0]);
      Rx("regressors[1,] <- current_others");
      Rx("regressand[1] <- current_this");
   }else{
      for (i=0; i<pairs; i++){
         prices[i] = iClose(symb[i], 0, 0);
      }
      Rv("current_all", prices);
      Rx("regressors[1,] <- current_all");
   }
   
   plot();

void onOpen(){
   int i, ii, j;
   int ishift;
   
   // if any pair has less than back bars in the history
   // then adjust back accordingly.
   back = back_bars;
   
   if (ObjectGet("back", OBJPROP_TIME1) != 0){
      back = iBarShift(NULL, 0, ObjectGet("back", OBJPROP_TIME1));
   }
   
   for (i=0; i<pairs; i++){
      if (iBars(symb[i], 0) < back){
         back = iBars(symb[i], 0) - 2; // use the third last bar.
      }
   }
   
   ArrayResize(coef, pairs);
   ArrayResize(prices, pairs);
   ArrayResize(regressors, back * pairs);
   ArrayResize(pred, back);
   Ri("back", back);
   Ri("pairs", pairs);
   
   // fill the matrix of regressors
   // and then copy it over to R
   for (i=0; i<back; i++){
      for (j=0; j<pairs; j++){
         ishift = iBarShift(symb[j], 0, Time[i]);
         regressors[i * pairs + j] = iClose(symb[j], 0, ishift);
      }
   }
   Rm("regressors", regressors, back, pairs);
   
      
   // do the regression
   // first we need a regressand
   if (trend){
      // we simply use a straight line that will be our ideal trend
      // note that it points downward since the history is ordered backwards
      Rd("trendslope", 0.01);
      Rx("regressand <- trendslope - trendslope * seq(1, back) / back"); 
   
   }else{
      Ri("cthis", this + 1);                       // counting starts with 1
      Rx("regressand <- regressors[, cthis]");     // use this column as regressand
      Rx("regressors[, cthis] <- rep(0, back)");   // set the column to zero in the matrix
   }
   Rx("y <- regressand");                          // stupid R will remember the variable names so we  
   Rx("x <- regressors");                          // have to be careful how we name them in the formula
   Rx("model <- lm(y ~ x)");                       // fit the model
   Rp("summary(model)");
   
   
   // get the coefficients
   Rgv("coef(model)[-1]", coef);   // remove the first one (the constant term)
   Rx("stddev <- sd(resid(model))");
   stddev = Rgd("stddev");
 
   // convert the coefficients to usable hege ratios by multiplying
   // xxx/usd pairs with their quote. The results can then be
   // conveniently interpreted as multiples of needed Lots or Units.
   // also take care of the special case when fitting a spread 
   // instead a trend
   for (i=0; i<pairs; i++){
      // if we fit a spread then all pairs except this one are on the other 
      // side (negative) and this one (the regressand) is 1 by definition
      if (!trend){
         if (i == this){
            coef[i] = 1;
         }else{
            coef[i] = -coef[i];
         }
      }
      
      // convert to units
      if (StringSubstr(symb[i], 0, 3) == "USD"){ 
         coef[i] = coef[i] * iClose(symb[i], 0, 0);
      }
      
      
      // The following makes sure that if the first pair is an USD/XXX pair
      // it is normalized to 1 again and the lot sizes of the other ones 
      // instead made smaller by the same factor.
      if (!trend){
         if (StringSubstr(Symbol(), 0, 3) == "USD"){
            coef[i] = coef[i] / Close[0];
         }
      }
   }
   
   // format a string that presents the hedge ratios
   // to the user and that will be displayed in the plot
   // it will also multiply them with base_units so you
   // have some reasonable numbers for your oanda account 
   ratios = "hedge ratios [multiples of Lots]\n";
   for (i=0; i<pairs; i++){
      ratios = ratios + symb[i] 
      + " " + DoubleToStr(MathRound(base_units * coef[i]), 0) 
      + " (" +  DoubleToStr(GlobalVariableGet(GLOBALNAME+Symbol()+Period()+"_"+symb[i]),0) + ")\n";
   }
   Comment(ratios);
   plot();   
}
void plot(){
   static int last_back;
   // predict and plot from the model
   Rx("pred <- as.vector(predict(model, newdata=data.frame(x=I(regressors))))");
   
   if (!trend){
      if (last_back != back){
         plotRemove("others");
         plotRemove("spread");
         last_back = back;
      }
      // plot into the chart
      Rgv("pred", pred);
      plotPrice("others", pred, clr_below, clr_above);
      Rgv("regressand-pred", pred);
      plotOsc("spread", pred, 1, 0, clr_spreadline, clr_spreadline);
      label("spread_cur", 10, 70, 1, DoubleToStr(pred[0] / Point / 10, 1), Lime);
   }   
   
   // make the R plot (optional)
   if (Rplot){   
      Rs("descr1", Period() + " minute close prices");
      Rs("descr2", "begin: " + TimeToStr(Time[back-1]) + " -- end: " + TimeToStr(Time[0]));
      Rs("ratios", ratios);
      
      Rx("options(device='windows')");
      if(trend){
         Rx("curve <- rev(pred)");  // it is still ordered backwards, so we reverse it now
         Rs("lbly", "combined returns");
         Rx("linea <- trendslope/back");
      }else{
         Rx("curve <- rev(regressand - pred)");
         Rs("lbly", "spread");
         Rx("linea <- 0");
      }
      Rx("plot(curve, type='l', ylab=lbly, xlab=descr1, main='Arb-O-Mat', sub=descr2, col='cornflowerblue')");
      Rx("abline(0, linea, col='cornflowerblue', lty='dashed')");
      Rx("abline(stddev, linea, col='green', lty='dashed')");
      Rx("abline(2*stddev, linea, col='green', lty='dashed')");
      Rx("abline(-stddev, linea, col='green', lty='dashed')");
      Rx("abline(-2*stddev, linea, col='green', lty='dashed')");
      Rx("text(0, range(curve)[2], ratios, adj=c(0,1), col='black', font=2, family='mono')");
   }
}

Quoting iwjw

Disliked

checking the market...isn't that something the EA could do for you if the EA knew what to look for?
trading only in a specified time window should also be doable
so semi-automated isn't neccessarily where we have to end up with this ea

Ignored

Quoting Dirtybrown

Disliked

Hopefully we'll go from overlay hedge EA users to hedge funds managers getting laid everyday we'll see!

Ignored

Quoting charvo

Disliked

ok, 1 lot eu = 100000 eur that you paid 127000 usd to buy (or sell).

1 lot eu = 100000 gbp that you paid 158000 usd to sell(or buy)

as you see, 158000 > 127000, this means if a trader buy 1 lot eu and sell 1 lot gu, the position is not really hedged (assuming correlation exits). it is in fact you are MORE on sell side of gu.

in order to be hedged (on position size), it should be:

127000 usd vs 127000 usd;

127000 usd can get you 100000 eur$, that's 1 lot;

how much lot 127000 usd can get you gbp$?

127000/1.58(gu price)...

Ignored

Quoting jajano

Disliked

hello, im gonna forward test this ea next week , first using mini account 500$. is there any recommendation setting for this ea into the mini account ?

Ignored

Quoting Dirtybrown

Disliked

http://www.mathworks.com/matlabcentr...exchange/13171

Ignored

Quoting KingHigh

Disliked

Dirty Brown,

You should go to the site below to get a really good thread on CoIntegration. It was started and then has mysteriously stopped with any substance. I believe the originator, Mediator, and some others have went off to develop on there own. Nevertheless, it is a GREAT read about using CoIntegration, R, and Forex.

Regards


http://www.stevehopwoodforex.com/php....php?f=5&t=113

Ignored

Quoting abal1000

Disliked

I think we should take one step back for a second with this EA and dos some fundamental analysis.

The issue we have not discussed is why is there is large drawdowns?

My simple answer is you can't have everything, for instance broader analysis of EURUSD and GBPUSD (old pair) vs another example like NZDUSD AUDUSD (new more highly correlated pair) shows that the reason for the large drawdowns in the old pair is the large gapping that occurs, this causes profit to increase as it gaps and meets and so on.

Whereas with the new pair there is less...

Ignored

Quoting fx4money

Disliked

Buy EU Sell GU = buy eur/gbp

this system will make money very fast when eur/gbp is uptrend on sideway, and will fail terribly when eur/gbp is downtrend.

see the eur/gbp chart for april it was downtrend until 15th may, many trader lost their money during that period. eur/gbp start to rise on 16th may, and of course traders are making profit using this system during this period.

Ignored

Quoting Thepipster

Disliked

My demo now has $3350 profit, the extra layers works well, just have to make sure the lot size is small to handle DD, because about 4 of these layers have stayed open since I started the EA...which make me think it may be best to start EA when the correlation spread is 0 on almost every TF, instead of at a deviation.

Ignored