Historical data differences?
Hmmm.. Wondering if my TP's are too small.... I was thinking that the data variances
between the different data providers (seems to average around 3-5 pips on a quote)
wouldn't be a problem for a trading system as they would even themselves out (i.
e. sometimes positive and sometimes negative towards a trading system). Maybe I'm
wrong on that assumption.... Anyone else done battle with this sort of thing?
I have found that the better the data, the worse the system performs. If I were
you I would not go lower the 1 hour timeframe with 1 min data.
Also you have to make sure you are not doing anything daft like using the close of the current bar, or trading off the interpolated ticks in bars.
If you are really serious about backtesting, I would get tick data from an independant source and import it into MetaTrader.
I finally did this after getting sick of the wild divergences in results you mentioned, I guess you can't expect much for free...
Also you have to make sure you are not doing anything daft like using the close of the current bar, or trading off the interpolated ticks in bars.
If you are really serious about backtesting, I would get tick data from an independant source and import it into MetaTrader.
I finally did this after getting sick of the wild divergences in results you mentioned, I guess you can't expect much for free...
It seems there was a lot of people having trouble with that data. Someone from metaquotes
was explaining that they combined histories from many sources, while dealers use
only a few sources. That seems to create a vast difference in MQ data and others
such as Alpari.
I wonder would the data improve if they try to emulate history from one such dealer.
In any case if the system gives you some fantastic results, it most likely soemthing is wrong with it. Always run system on demo for a time (like a week or a month) than compare with results from tester for the same period. They should be pretty close.
I wonder would the data improve if they try to emulate history from one such dealer.
In any case if the system gives you some fantastic results, it most likely soemthing is wrong with it. Always run system on demo for a time (like a week or a month) than compare with results from tester for the same period. They should be pretty close.
Hi Guys,
Thanks for your comments. In answer to Craig's question, I place all my trades based on current bar -1 (i.e. - only fully formed bars) so there can't be any problems with incorrect indicators etc. And with regards to irusoh1's comments - if MetaQuotes are providing combined histories from many sources (i.e. indicative, rather than real data), then it's probably not a hell of a lot of use to anyone, unless you're trading the bigger time frames (i.e. H4 and above). Ah well, that's life. As you said Craig, you can't expect too much when you get it for free....
Might have to splash out and get some tick data..... But again, that's another hornets nest. Because each broker makes their own markets internally (i.e. no centralised exchange and therefore no "official" prices), each broker's tick data will be slightly different to every other broker's... I see that data providers have wised up to this and now provide tick data from a number of the big brokers. God, it's not easy! But if it was, every bugger would be rich and life would get pretty boring!
Thanks for your comments. In answer to Craig's question, I place all my trades based on current bar -1 (i.e. - only fully formed bars) so there can't be any problems with incorrect indicators etc. And with regards to irusoh1's comments - if MetaQuotes are providing combined histories from many sources (i.e. indicative, rather than real data), then it's probably not a hell of a lot of use to anyone, unless you're trading the bigger time frames (i.e. H4 and above). Ah well, that's life. As you said Craig, you can't expect too much when you get it for free....
Might have to splash out and get some tick data..... But again, that's another hornets nest. Because each broker makes their own markets internally (i.e. no centralised exchange and therefore no "official" prices), each broker's tick data will be slightly different to every other broker's... I see that data providers have wised up to this and now provide tick data from a number of the big brokers. God, it's not easy! But if it was, every bugger would be rich and life would get pretty boring!
I find it difficult to believe that Metaquotes could not have done a better job
with the data download, they are, after all good buddies with a number of brokers
who all must have reasonable data sets. In my more cynical moments I could almost
start to think that things like this are done deliberately to get more people to
take the plunge with poor EA's.
But I'm sure that's just my raging paranoia :) Cw1, you are correct, there will be differences in tick data from different sources, but it's still about 100000 better than testing at 90%.
But I'm sure that's just my raging paranoia :) Cw1, you are correct, there will be differences in tick data from different sources, but it's still about 100000 better than testing at 90%.
Only one note: many times it is only a misunderstanding that the so called 90% accurate
simulated ticks (this number is arbitrary, and doesn't reflect reality) works fine
using M1 data for generating ticks for H1 charts. Bars are only compression of
price data and different periods mean only different amount of compression.
-If your EA uses only bar data (Open, close, high, or low), and ONLY these data, you will get realistic results with the simulated ticks. In this case H1 data is enough for H1 period, M1 data won't have an effect, since only the 4 points of an H1 bar will be used for any calculations.
-If your EA uses price data (bid or ask) for any calculations, you will need accurate price data, no matter if you use H1, or M1 periods otherwise. Note! Price is price: no matter which point you arbitrarily set for bar close/open.
The more accurate price the EA needs (eg.: tighter stops than average hourly movement needs pretty accurate data) the more unreliable your results will be. There can be large movements within one minute very often. If your EA uses anything (indicator, stops, anything) that needs price data from within that movement, the result will be totally messed up.
cw1: I can be reached via forexzapATgmailDOTcom for further tick data info.
-If your EA uses only bar data (Open, close, high, or low), and ONLY these data, you will get realistic results with the simulated ticks. In this case H1 data is enough for H1 period, M1 data won't have an effect, since only the 4 points of an H1 bar will be used for any calculations.
-If your EA uses price data (bid or ask) for any calculations, you will need accurate price data, no matter if you use H1, or M1 periods otherwise. Note! Price is price: no matter which point you arbitrarily set for bar close/open.
The more accurate price the EA needs (eg.: tighter stops than average hourly movement needs pretty accurate data) the more unreliable your results will be. There can be large movements within one minute very often. If your EA uses anything (indicator, stops, anything) that needs price data from within that movement, the result will be totally messed up.
cw1: I can be reached via forexzapATgmailDOTcom for further tick data info.
You are missing trading opportunities:
- Free trading apps
- Over 8,000 signals for copying
- Economic news for exploring financial markets
Registration
Log in
You agree to website policy and terms of use
If you do not have an account, please register
Just wondering if anyone else seems to obtain totally different results for specific trading systems when simulation trading those trading systems using MetaQuote data or data from different brokers? I constantly come across this problem. If I use data from say, MetaQuotes (obtained via the history function), my trading system would produce say, 24% profit for a specific year. If I then try the same thing again using data from a different broker (say, Alpari or FXDD), I then end up with a loss of 80% for that same year! And this is using indicators based on an hourly chart, and using 1 minute data to determine the actual trades (i.e. - it can't really get any better than that unless tick data is used).
Has anyone had these huge variances when using data sets from different brokers? The only reason I can think of for this happening is that because each broker is in effect a market maker, you would really need to use a quote data set from the broker that you intend to trade with. Anybody have any other possible reasons for this or comments on the above?
Any help would be greatly appreciated!
Cheers,
Coen Willemse