Back testing... or testing backwards... or has Alice slipped some potion into my tea, not sure of anything anymore !!! - page 2

 
tovan wrote >>

If you ever want to trade for real then you've got to use the broker's data at some point. I say bite the bullet now and start using it in your backtesting. Once it's there it doesn't change (that's my experience with IBFX anyway). It's not always perfect, but that's life. I prefer to test in an environment that's as close to real trading as possible so I know that I can expect something close when I get to the forward test phase. In my opinion, the only reason to go out and download someone else's data is if you want to test over a period where your broker doesn't have the data available.

These are my opionions - take them for what they're worth.

- Tovan

So,

In anyone's experience - which broker has the most back test data available ?...

If there's no point back testing "other" data (which makes sense) then who's got the most ?...

Seriously - Igiven that its 2009 and we've had electronic trading for a LONG time there must be a broker out there with back data for ALL the major pairs for at least the last 4-5 years. My experience is that some of the pairs only come with data for a few months...

Who has the best ?

 
TheForexDevMan:

So,

In anyone's experience - which broker has the most back test data available ?...

If there's no point back testing "other" data (which makes sense) then who's got the most ?...

Seriously - Igiven that its 2009 and we've had electronic trading for a LONG time there must be a broker out there with back data for ALL the major pairs for at least the last 4-5 years. My experience is that some of the pairs only come with data for a few months...

Who has the best ?

However, as much as I wanted to import this data the tester won't let me use it in a reliable way.


I don't mind using the broker's data as far as it is complete. For back testing purposes, any good quality data should be adequate, and backtest results shouldn't change much. I registered my account with InterbankFX and this broker's historical data doesn't seem to be complete. The question remains however, what is the value of the download functionality in the history center if this data gets corrupted during the test?


IMHO the tester should be a separate program, not part of the terminal. The tester should let user select any historical data source.


I find myself wasting time on technical issues instead of working on and testing Forex strategies.

 
lukasz74nj wrote >>

However, as much as I wanted to import this data the tester won't let me use it in a reliable way.

I don't mind using the broker's data as far as it is complete. For back testing purposes, any good quality data should be adequate, and backtest results shouldn't change much. I registered my account with InterbankFX and this broker's historical data doesn't seem to be complete. The question remains however, what is the value of the download functionality in the history center if this data gets corrupted during the test?

IMHO the tester should be a separate program, not part of the terminal. The tester should let user select any historical data source.

I find myself wasting time on technical issues instead of working on and testing Forex strategies.

I totally agree. I must spend a lot of time double checking test results. Its arbitary as to what happens. No alert or message box if you don't have the data that covers the date range you've selected. The test runs through only for you to see the first "trade" was last week as that's the only data set it has - peculiar to say the least.

I've been a software developer for 28+ years now. I understand software development and in the big scheme of things this bizaare handling of history data has left me gob-smacked. Why are there even unmatched data errors ?... Its not rocket science - its DATA in a flat file - Its not difficult to read it without corrupting it.

This really needs addressing. Theres not much point to back testing if the history isn't even reliable or un-corrupted...

 
TheForexDevMan:

I totally agree. I must spend a lot of time double checking test results. Its arbitary as to what happens. No alert or message box if you don't have the data that covers the date range you've selected. The test runs through only for you to see the first "trade" was last week as that's the only data set it has - peculiar to say the least.

I've been a software developer for 28+ years now. I understand software development and in the big scheme of things this bizaare handling of history data has left me gob-smacked. Why are there even unmatched data errors ?... Its not rocket science - its DATA in a flat file - Its not difficult to read it without corrupting it.

This really needs addressing. Theres not much point to back testing if the history isn't even reliable or un-corrupted...


I finally made this work and can enjoy 90% modeling quality with no corrupted data! Here are the steps:


- open MT4 terminal

- delete the current account using the terminal interface (don't delete any broker config files!), as well as all .hst files

- open the history center and import your data - make sure to select M1 before importing (I downloaded data mentioned in the above link, then post-processed using AWK to make it compatible with the .csv format which can be imported into MT4)

- now you have to close MT4 terminal - if you try to open an offline chart before closing the terminal you won't see any M1 data - that's OK

- open terminal again and CANCEL dialog boxes asking to open an account - you want to be in an offline mode all the time, which also implies that you need a dedicated terminal instance for testing

- use the period converter script to generate all higher timeframes

- now as long as you're in an offline mode MT4 tester won't attempt do download any historical data from the broker which would otherwise corrupt your own data


Hope this helps.

 
May be this is because the MetaTrader hostoric data is simulated and not real data? Any ideas on that?
 
jay_forex wrote >>
May be this is because the MetaTrader hostoric data is simulated and not real data? Any ideas on that?

I read this forim as I do myself backtest an average of 10 to 20 EA by week. I can not understand no one from the beautiful compagny of metaquote can give us any answers on all the questions regarding backtesting with their tool.

How can you work with a tool every users have doubt on ?

don't know for you, but I do not understand, how metaquote want us to work in partnership with us, hands on hands, without giving us the answers to the questions we ask ? . We invest alot of time in study market, learning their programing language, and at the end, we have to loose time on the result to double check and be sure it is not the strategie that is not good but the back test system itself ?

What a partnership with this compagnie, if the only way to get answers is to ask on a forum a wait until a members (who we dont know anything on at all) answers ? anybody can answer anything, that is why I continue backtesting even if some of the members say on the forum that the backtest prog and reliabylity may be to discuss but when I see SO MANY people saying how they dont trust the backtest, this tell me that they might have some good reson, but metaquote never answered anything on the subject in the forum and never tell it was true or not. Ok, they prevent us it may be different from fowardtesting, i agree and do understand, but is it SOOO diffuclt to reproduce what happened in the past only to check if the strategie or the programming is good ?

what is so different to send live data

and

record live data to send it back to the terminal to see what would have happen if the program would have react in an other way on this event or another

please excuse my bad english, (for me, more then learning metaquote language, studdying markets, I have to learn english to)