Too few bars when backtesting fixed timeframes

void start()
{
//----
//    Print daily EMAs of the last days
    for(int i = 4; i >= 2; i--)
    {
        double ema5  = iMA(NULL,PERIOD_D1,5 ,0,MODE_EMA,PRICE_CLOSE,i),
                 ema20 = iMA(NULL,PERIOD_D1,20,0,MODE_EMA,PRICE_CLOSE,i),
                 ema50 = iMA(NULL,PERIOD_D1,50,0,MODE_EMA,PRICE_CLOSE,i),
                 stoch20= iStochastic(NULL,PERIOD_D1,25,3,3,MODE_SMA,0,0,i);

        Print(TimeToStr(iTime(NULL,PERIOD_D1,i)),
            ": EMA5 = "+DoubleToStr(ema5,Digits),
            ", EMA20 = "+DoubleToStr(ema20,Digits),
            ", EMA50 = "+DoubleToStr(ema50,Digits),
            ", Stoch20= "+DoubleToStr(stoch20,Digits));
    }

    Print("Bars = "+iBars(NULL,PERIOD_D1));
} 

Quoting iya

Disliked

When backtesting with indicators from different timeframes, the results can be off, because StrategyTester simply does not provide enough history. Here's a simple EA for testing, which displays the values of a few EMAs from the daily timeframe:

PHP Code

void start()
{
//----
//    Print daily EMAs of the last days
    for(int i = 4; i >= 2; i--)
    {
        double ema5  = iMA(NULL,PERIOD_D1,5 ,0,MODE_EMA,PRICE_CLOSE,i),
                 ema20 = iMA(NULL,PERIOD_D1,20,0,MODE_EMA,PRICE_CLOSE,i),
                 ema50 = iMA(NULL,PERIOD_D1,50,0,MODE_EMA,PRICE_CLOSE,i),
                 stoch20= iStochastic(NULL,PERIOD_D1,25,3,3,MODE_SMA,0,0,i);

        Print(TimeToStr(iTime(NULL,PERIOD_D1,i)),
            ": EMA5 = "+DoubleToStr(ema5,Digits),
            ", EMA20 = "+DoubleToStr(ema20,Digits),
            ", EMA50 = "+DoubleToStr(ema50,Digits),
            ", Stoch20= "+DoubleToStr(stoch20,Digits));
    }

    Print("Bars = "+iBars(NULL,PERIOD_D1));
} 

...and it's results: the error gets bigger, the longer the EMA period. The Stochastic is correct, as long as it's period does not exceed the available bars.

http://img65.imageshack.us/img65/265...problemdv0.png

I guess it's one of those MetaTrader quirks, but it would be nice if somebody has a solution. It does get somewhat better, by using an earlier start date, but that's inconvenient when testing a certain period and outright impossible at the beginning of available data.
Is there maybe some magic LoadHistory() command that I've missed?

Ignored

Quoting sergiu

Disliked

Well, compared to you i am a noob when it comes to MQL but i am quite proficient in a different language for a different platform. That platform even though more robust than MT4 it is also a bit more limited than MT4. One of the limitations is that in order for me to reference other TFs i would have to have those TFs inserted in the chart as a second data set. Sometimes however i would need to avoid that so i would create virtual bars inside the code, store them in an array and use them to calculate whatever indicators i would need from that TF. So for instance if you have an EA for the 15 M and you need to use an EMA from the 1H you would store the open of the 15 M candle, the highest high and lowest low of the last 4 candles and the close of the fourth candle. That would become your HLOC for an hourly candle. I guess you can see where i am going this. Coding this could perhaps be a bit intense but it should eliminate the errors of referencing other TFs and you would only need one data set for your testing. Just an idea...

Ignored

Quoting smjones

Disliked

LOL, You are an AMI BROKER programmer... I know exactly what you are talking about, because it all has to be done with arrays.

Ignored

Quoting sergiu

Disliked

Not quite. I use Tradestation. The arrays just come in handy when it comes to that kind of coding...

Ignored

Quoting sergiu

Disliked

Well, compared to you i am a noob when it comes to MQL but i am quite proficient in a different language for a different platform. That platform even though more robust than MT4 it is also a bit more limited than MT4. One of the limitations is that in order for me to reference other TFs i would have to have those TFs inserted in the chart as a second data set. Sometimes however i would need to avoid that so i would create virtual bars inside the code, store them in an array and use them to calculate whatever indicators i would need from that TF. So for instance if you have an EA for the 15 M and you need to use an EMA from the 1H you would store the open of the 15 M candle, the highest high and lowest low of the last 4 candles and the close of the fourth candle. That would become your HLOC for an hourly candle. I guess you can see where i am going this. Coding this could perhaps be a bit intense but it should eliminate the errors of referencing other TFs and you would only need one data set for your testing. Just an idea...

Ignored