You know what I love...

 

Spending endless hours with the brilliant MQL debugger, coding an new idea. Watching endless visual backtests to see that everything is working ok. Finally, I think I have a winner (I'm sure you all know this feeling), only to find out that my history data has once again become 'corrupted' and that if I wipe the history and re-download the data (for what seems like the 1000th time) the idea does not work. In case you have not worked it out, I'm being sarcastic, WHEN WILL METAQUOTES FIX THIS F**KIN BULLS**T!!!

 

I have some experience in backtesting:

- I am discunnecting of Metatrader from the broker;

- deleting all history files.

- I am downloading Alpari data which is back to 2001.

- and backtesting in MetaTrader's offline mode.

Why offline mode?

Because I'm affraid that Meatrader will download some uknown data from somewhere and good data will be mixed with bad one.

Why Alpari data?

many people spent a lot of time during the long period of time creating and checking this data so this data is ok. Besides it is free (not for money as many other unknown data).

Just hope that somebody will do the same with other broker's data but it is not really easy.

I am not recommending to backtest EAs in Metatrader's data which everybody can download in build 200 for example.

 

Thank you for your relpy, how does one set this 'offline mode' for the tester?

 
Craig:
Thank you for your relpy, how does one set this 'offline mode' for the tester?

What is one set?

I am doing it exactly as Codersguru article:

http://www.metatrader.info/node/67

https://www.mql5.com/en/forum

Only differencies that I am doing it in offline mode from the beginning: i mean I disconnet from broker's server. Otherwise Metaquotes may download some data and I should start from the beginning. But good things are the following: their data is always on the different files so I can simple delete it.

Build 200 is much more difficult for us concerning backtesting (i mean this "improvement" that we can download their data from nowhere). But anyway we will survive.

 
Craig:

Spending endless hours with the brilliant MQL debugger, coding an new idea. Watching endless visual backtests to see that everything is working ok. Finally, I think I have a winner (I'm sure you all know this feeling), only to find out that my history data has once again become 'corrupted' and that if I wipe the history and re-download the data (for what seems like the 1000th time) the idea does not work. In case you have not worked it out, I'm being sarcastic, WHEN WILL METAQUOTES FIX THIS F**KIN BULLS**T!!!

Personally I think the quality of backtesting is a secondary issue - too many variables, not least of which is test data freely available, which sucks! However, until there is a proper debugger introduced, Metatrader will never be taken seriously by anyone that matters. A platform is only as good as the software available and this would also include indicators and EAs which at the end of the day is what 'feeds the beast'. There is a limit to the level of complexity that can be achieved without a debugger and I think that MT4 development of indicators/EAs has plateaued as a result. The introduction of the Visual Backtester was hugely welcome and would be doubly useful with even rudimentary debugging features - I haven't seen anyone mention that there is no 'Expert' tab available in Visual mode and unless someone can tell me how to enable it, is beyond belief. This means that the only real debugging tool we have (the humble Print command) cannot even be used. I have come to the conclusion that Metaquotes really doesn't want to see any development - remember, they are trying to sell their system to brokers who more often then not are taking the other side of your trade. Would you implement a system that you knew would support numerous PROFITABLE EA's etc.??? I know I wouldn't and you can be pretty sure that Metaquotes knows this too. It would appear that not only is there little incentive for them to include a proper debugger, there are major reasons why they shouldn't. Unfortunately, since only about 10% of people probably do any programming, most have no clue as to just how bad Metatrader is when it comes to developing stuff. All they focus on is the pretty colours having no idea as to how difficult it may have been to produce them.

Please excuse the rant but I just spent about 2 days trying to debug an indicator that would probably have taken 30min with a proper debugger.......

 

I feel your pain brother, I guess we get what we deserve for free.

BTW, the print stuff comes up on the 'Journal' tab, and you can also use the comment function to print stuff out.

 

agreed. ..Metatrader is very tough. It would be nice to have customized scripts to add to the entirety instead of starting from scratch. Rsi + Stoch + MAC + various time frames + alert = trade!

Just a simple macd alert cross is hard enough

 

It's a tough learning curve if you don't know C. Eventually you get the hang of writing this crossing that type stuff, and you find out none of it works! then stuff get's really difficult.

 
Craig:
I feel your pain brother, I guess we get what we deserve for free. BTW, the print stuff comes up on the 'Journal' tab, and you can also use the comment function to print stuff out.

Wow, I must be losing the plot! You're right of course about the Journal (just checked!...). I would have bet diamonds that the Print function didn't working in it, so much so that I was was using additional display buffers to monitor debug info. Comment function has it uses too of course.

 
omelette:
Wow, I must be losing the plot!

Don't worry, I lost it long ago!