Period in Strategy Tester
Perhaps someone else can explain this more eloquently than I, but it has to do with the price data being used. Using M1 data is going to be more accurate that H1 data. So if you have your timeframes hardcaoded, you should just use M1 as the timeframe. Or you could use a higher timeframe and set the method to "Every Tick". This method uses the lowest timeframe available for the date time-line specified when you run the test. So if you set period to Daily, and run "every tick" model it will use all timeframe-data down to and including M1 data assuming you have historical data that far back for M1.
The point is, get yourself some M1 data as far back as you can, always use "every tick" model, and don't run test on time lines that you don't have data for.
Alpari offers historical M1 data, maybe someone has the link handy.
The point is, get yourself some M1 data as far back as you can, always use "every tick" model, and don't run test on time lines that you don't have data for.
Alpari offers historical M1 data, maybe someone has the link handy.
So according to what you say, when using 'every tick' model, the Period field is insignificant because the tester uses all data it has anyway? (that is if the time-frame is hardcoded as in my expert...)
That is my understanding.
Alpari historical data at http://www.alpari-idc.com/en/dc/databank.php
That is my understanding.
I have done some checking and it appears that this is not the case. I get different data with different Period values even when 'every tick' is selected. Actually, I tested it with an expert that does nothing except printing the tick time and prices -
int start()
{
Print(TimeToStr(CurTime(),TIME_DATE|TIME_MINUTES|TIME_SECONDS)," ",
Ask, " ", Bid, " ", Open[0], " ", Close[0]);
return(0);
}
I also made sure that I'm using dates that I know I have full historical data for them in all time-frames. I can't explain the results. Sometime it seems that when using D1 I get more accurate and more ticks and sometimes it seems true for M1. If I test it with a 'real' expert I get different trading results with different profitability.
Its very nice to have a tool like the strategy tester but using a tool not in the right way might bring damage rather than benefit. It doesn't seem right to try all combinations on my strategy and pick the most profitable one :-)
Can anyone explain?
I begining to think that accurate backtesting is actually the holy grail (read: unachievable goal).
I hope someone here could clarify on this:
I have an expert that uses hard coded timeframe (PERIOD_H1). Still I get different results when using different values in the Period field in the Strategy Tester. What is the significance of this field besides determining the value returned by the function Period(), so that it affects the results in the tester?
Still in the same issue, the reports in the journal of the tester show timeframe different than the one I select in the Period field (e.g. I select M15 and in the journal I see "...EURUSD, M30: ..."). What's the deal?
Thanks,
Joe