Data-Feed: This is Ridiculous !!

Have you tried using the method of importing reliable tick data to strategy tester from birts tick data blog ? I havent got my EA debugged to where I can test it yet but when I do this is how I am going to do it http://eareview.net/tick-data

MT4 doesn't save the spread, it get it from the last server value. So if you're online, then one test run my be very different from another.

you can save/restore the current spread by making a copy of trader\history\x\symbols.sel

SDC: Have you tried using the method of importing reliable tick data to strategy tester from birts tick data blog?

I'm almost certain mt4 don't have real tick back-testing. That site more than like have 1 minute data (without gaps). My strategy that got hurt the most was the Gap concept.

WHRoeder: MT4 doesn't save the spread, it get it from the last server value.

My broker have a fixed spread of 2. I have Indicators that shows the broker current spreads and it was 2 back-then as it is 2 now.

MT4 normaly would create simulated ticks from its one minute chart, it creates those simulated ticks and saves them in a strategy tester file called an fxt file, if you replace that fxt file with one created from real ticks, strategy tester then uses those real ticks in backtesting, this is where that site i gave you comes in, it shows you how to download real tick data and create an fxt file from them and how to prevent mt4 from overwriting it with its simulated ticks.

So yes you are right mt4 doesnt normaly have real tick back testing, but you can force it to use real ticks by following birts instructions on his site.

Also if you download the tick data from the source he recommends, it does include gaps, the files contain every tick, 7 days a week 365 days of the year, for non market trading times like the weekend each tick has a zero value but they all are there, i know this because i studied some of the files to see what they contain when I downloaded them.

I think anyone that has attempted to write an EA has come across this. you need to understand if your EA is based on indicators, the values will differ from broker to broker. So something you may have optimised for one broker, may not be tuned for the next. The simple rule I follow is to tune it on the brokers data your going to use it on.

After going through this kind of mayhem, I decided that I would review my EA's and base them more on price as much as possible, after looking at hundreds of wierd and wonderful indicators, I decided that you can't beat simple price action and basic indicators.

On another note, I had the same problem with thinking historical data changed too! I couldn't figure it out until I realised that windows vista and windows7 uses Virtualstore. This prevents you from saving directly to programfilesx86 folder. You need to do ALL your EA editing from: \Users\yourlogin\AppData\Local\VirtualStore\Program Files (x86)\MetaTrader\

Latest OS don't let you write to files in your native programfiles folder.. This caused me a lot of pain.

For consistent results in the Tester u need identical History Data and identical Market Properties across all tests (Market Properties can be seen by pressing 'Symbol properties' in the Tester) . If there are differences in either, then results will differ.

For testing it is recommended to use a separate Terminal that is disconnected (you can find a method to do this here -> https://www.mql4.com/search/bogus%20proxy).

Thanks everyone gordon,Genma,SDC and WHRoeder. So here's the plan. I have 2 Meta-trader on my windows xp. The broker's I intend to use version (currently running forward tests on); and the default metatrader4.com version. I'm gonna download and install the tick data on the default version and give it a bogus proxy to use for my testing.

However, I can't help but agree with Genma from an overall stand-point. When gordon said "justification should first come from a real theoretical basis as a foundation for the strategy". The first thing that came to mind was ah-ha gaps lol. But not just your every weekend gap (which are not really gaps but missing activity) but rather supply/demand forcing brokers to fulfill orders (I mean thats how they make their money).

To me, using more accurate data in back-testing which would not have some of these broker specific gaps (among other things), and then using that information to forward test in the less-then-perfect environment brings me back to the same dilemma I opened this topic with. Besides news releases,gaps and maybe divergence (which is next on my list). I cannot for the life for me conceive anything else which fits gordon's description. Why should any Indicator/Oscillator work, why should 3 day moving average crossover 14 day ma be any better than 5 day moving average. Do the bulls or bears lose their appetite on the third day.

I know, I know, the point of back-tests is not to validate what WILL work in the future but to give a theoretically indication of what would have worked in the past. If that's the case then I should consider my systems which are losing some performance but still profitable a success. The Market Properties changed and the system still worked, guess this is the best anyone can ask for.

All these historical data feeds and demo trade and back testings and forward testings are all bull-crap!

Now-a-days one can trade as small as 0.01 lot with a 10 bucks margin why don't you guys just test your strategy or EA on real time live trading and feel the ups and downs of real heart-throbbing market.

Although that method works, it's not straight forward, it's extremely consuming in system resources (the Tester would be waaay slower, it would take u days to download the data with the script and the uncompressed files are ~GB's in size), it's not your broker's data (and if it's Dukascopy - it's not even from an MT4 platform). So I recommend u skip this method for the time-being cause it's likely to cause u more problems than benefits. There are only rare occasions when it's worth going to all that trouble.

Regarding my "theoretical basis" comment. Firstly, that's just my own opinion (I am sure many don't agree). Secondly, I noticed u followed my discussion with 7bit (https://www.mql5.com/en/forum/126410/page2#330808) - well, that's an excellent example of trying to build a "theoretical basis" for the strategy. I am not saying that I think that strategy would work or that the basis is 'solid' enough (IMHO, it isn't and there is still much theoretical work to be done), but I'm willing to bet that anyone who uses that sort of approach to algorithm development is much much more likely to succeed than the average joe who notices that when indicator x crosses indicator y something happens... but can't answer the simple question - "why would it work?".

(IMHO).

EuroTrader: All these historical data feeds and demo trade and back testings and forward testings are all bull-crap!Now-a-days one can trade as small as 0.01 lot with a 10 bucks margin why don't you guys just test your strategy or EA on real time live trading and feel the ups and downs of real heart-throbbing market.

Thanks but, No thanks, My experience would just not allow it. Thats like someone saying why don't you learn to count cards on the $2 blackjack tables and forget those simulators. mini lots on 10 bucks margin is the same as learning to play poker at penny tables on-line, no heart throbbing there, everyone pretty much All-in-All day. Forward testing is not the way to built a strategy, profit trading for 3 months of forward testing could be mere coincidence. Here's another angle, what if a strategy you read in a book or got from a professional trader seem promising but after 6 month of forward testing fails. Do you abandon the strategy or give it more time? That would be way too time consuming compared to any of the mention methods above.All that's going to do is allow the trader to form bad habits, same habits thats going to bind him surely as steel when he attempts to break them later.

gordon: [...]well, that's an excellent example of trying to build a "theoretical basis" for the strategy.[...]I'm willing to bet that anyone who uses that sort of approach to algorithm development is much much more likely to succeed than the average joe who notices that when indicator x crosses indicator y something happens[...]

Your point is made :) gordon. As you already know, I value your Opinions. If you're correct then my fear of someone have to be proficient at Statistic/Mathematics,Programming and Psychologist to be a successful trader would be realized. I better dust off the pre-Algebra book and start ordering statistic /calculus materials.