Brokers and Strategy Tester Accuracy? - page 2

In order to answer that question let's start with the data. Whose data are you using and for what length of time. Ex 10 years or 1 week?

I was using minute data that I downloaded a while ago. It was minute data that dated back to 1999.

Then you're using meta-quotes data. That data is missing gaps as long as a week sometimes. Also, it's Not Broker A nor Broker B's Data. Their data's would have entirely different prices than what you see in the History download form mt4.

Second question: what's the current spread for Broker A and for Broker B. Also, do they have fixed spreads or do their spreads move around? 

Ohh I see, so I would need to ask them directly for their own data. I remember asking other brokers before, and they said that the metaquotes data would be good enough since pips vary but not that significant. For this purpose, should I not use metaquotes data?

Also, the current spread for EUR/USD was 6.4 pips for the unprofitable one; but 2.0 pips for the profitable broker (which was ok because I increased the target profit another 5 pips just to see what would happen; and it still was profitable). They both are variable.

Well see, brokers can say anything they want. The reality is very different as you can already see with your back.tests. Yeah sure it does not make much difference when testing a system with stop.loss and take.profit of over 50pips but the smaller the tp and sl, the more un.stable the results.

 

Even if you had their data. They'll also need to give you the spreads at the time of your trades in the Back.tester to make it more accurate. The MT4 Back.tester Only uses the spread at the time of testing. If you're testing on the weekends (which it's close to). The broker spread could be 10 points higher than usual but mt4 don't care it uses what it sees.

 

See even tho u introduced the Trailing stops. All a takes is 1pip / point / or whatever you wanna call it to trigger the Trailing stop. On one broker the trail was set when the profit went up by 1. Trailing stops never gets set unless the order generates a profit. So, if the biggest losses on the losing broker are not recorded on the winning broker then the whole purpose of the test is skewed.

 

Last question, would your broker really let you scalp them. lol ... No i mean would they really/really let you win for a long time with a scalping strategy without, starting to send your order's directly to the market. This could cause delays in up to minutes.

 

In conclusion, you're testing in an un.realistic environment and trying to apply that to the real markets could be a bad idea. So who's fault is the reason for the inconsistency of the test. Well, everyone IMHO, Meta.quotes, Broker A and Broker B.

edit: oh almost forgot my disclamer.... All in my Humble Opinions lol 

lol, well I did ask them and they said that I could scalp with them. I know the only way I would be satisfied to know whether or not the program works is with demo trading. However, for some reason, the program will not go live with the profitable broker. It would have solved my question if they just allowed me to trade live, but I don't know what's in the code that makes it not trade for that particular broker. I'm able to hear the trade alerts, but no actual demo trades are placed. But I do see what you are talking about with the trailing stops.

Also, thank you very much for your humble opinions =]

Like Ubzen said there is no real tick data so maybe it is not so important BUT, just as a side tip...

... you could always recode your ea so that it does not use a TP, and instead uses OrderClose() if price is higher or equal to your tp to avoid the 10 pip tp limit...

Wont do much in backtester though...

/ McKeen

Thanks, I'll give that a try. I added in an official stop in the EA, and even a TP of 25 pips, but the program will still not place trades for the profitable broker. I'm just really curious how much different the demo forward tested results will be from the backtested results (considering the backtested results matched the forward tested results of three other brokers).