1 (edited by nquental 2014-10-19 11:08:25)

Topic: Big difference between builder and trader

Hello
Can you check the attached? It's a SP500 strategy that uses FRAMA. There is a very large difference backtested FRAMA calculations and actually traded (in a real account) FRAMA values. This led to completely different position entries and exists.

Note that the data is exactly the same, because I've exported the real MT4 data to FSB Pro.

What could explain this?

Thanks
Nuno

Post's attachments

SP500.zip 292.15 kb, 3 downloads since 2014-10-19 

You don't have the permssions to download the attachments of this post.

Re: Big difference between builder and trader

Nuno,

How many bars you have on trader?

Re: Big difference between builder and trader

450 bars

Re: Big difference between builder and trader

The indicator lines are very different in both charts.

A possible reason is that the indicator need more bars. That's why Footon ask you about that.
Note an indicator value for a given day.
Open the Trade Settings panel and start increasing the number of bars to 500, 700, 1000.. you receive the same value in the trader chart.

(After changing the Min Bars parameter, you have to close the connection and reconnect.)

Re: Big difference between builder and trader

Thanks.
I had already increased the min bars to 450 long ago and reconnected, so I'm almost 100% that's not the cause. I can increase it again this evening when market opens, but note that the max period of FRAMA is 126, therefore 450 bars is necessarily enough to calculate it right (or am I wrong)? Still I'll post the results of the experiment this evening.
Cheers

Re: Big difference between builder and trader

There is no other explanation why the indicator has different values.

Re: Big difference between builder and trader

I'm skeptical but you know it better, I'll come back this evening.

I would propose that FSB manages min bars for us if possible, to prevent problems. This can easily ruin an account, and it's not easy to detect.

Re: Big difference between builder and trader

I would propose that FSB manages min bars for us if possible, to prevent problems. This can easily ruin an account, and it's not easy to detect.

I'm trying to find an elegant solution to this problem. I'll most probably develop a separate tool for analyzing a strategy and finding the minimum count of bars for a reliable trade.

Re: Big difference between builder and trader

nquental wrote:

but note that the max period of FRAMA is 126, therefore 450 bars is necessarily enough to calculate it right (or am I wrong)?

FRAMA needs

PeriodFRAMA + PeriodFRAMA + 3

as bare minimum.

10 (edited by nquental 2014-10-19 14:57:12)

Re: Big difference between builder and trader

So that's 255, way less than what I have. But let's see later on when the markets open.
I invite you to replicate in your PCs in case your broker sells SP500.

Small note: there no "bare minimum" I guess, there's minimum smile We are quants after all.

Re: Big difference between builder and trader

Somehow backtesting and forward testing are vastly different critters, the results of backtesting can be disregared easily as they are generally irrelevant.

Even with forward testing, the next bar of data could destroy a system. Please do not rely on backtests as being anywhere near credible or accurate for risking your money.

My 'secret' goal is to push EA Studio until I can net 3000 pips per day....

Re: Big difference between builder and trader

Eheheh I'm quite puzzled reading your last message. You are a moderator of a forum of a program that basically makes backtests, and at the same time you say "please do not rely on backtests as being anywhere near credible or accurate for risking your money". Do you see the contradiction? I hope you do smile
The question is how to perform backtests so as to minimize data snooping / overfitting.

Anyway this as little to to with the topic of these posts.

Re: Big difference between builder and trader

Dear all
I confirm Popov was completely right smile) Thanks a lot.
I am sad however that I lost money because of this, and I think it was not my fault. The 450 bars in fact were not sufficient. 1000 were, so I set all strategies to 1000 bars to be sure. Popov, I believe these type of problems must be solved if you want your excellent program to be widely bought. I did what the program was told (sometimes there are pop-ups saying that min bars must be increased) and it was not enough. Please do have a look at this. If 1000 bars is in principle enough for everything, make it the minimum, and not 300.
Thanks
Nuno

Re: Big difference between builder and trader

Certainly I see your point and I reiterate that backtesting can not be relied upon, do make good use of the forward testing features and also be sure to test on a demo account for 6 months before a live account.

My 'secret' goal is to push EA Studio until I can net 3000 pips per day....

Re: Big difference between builder and trader

I think you are right, but I'm not patient enough to wait for 6 months... Yes, I know, then I cannot complain...

Re: Big difference between builder and trader

I re-open this issue, because ma indicators in my strategy have different (and changing!) values in Trader part compared to the Builder. For both parts I use identical data (Ctrl+S from the chart in MT4).

I tried to set different "Recommended chart bars" values in Trader mode/Auto Trader/Trade Settings (300, 400, 600, 1000 and 2000).
I compared values of ma and ma crossover indicators at the identical historical candle in Trader and Builder (e.g. 12th May, 16:50 candle).
I expected that from some "Recommended chart bars" level, indicator readings will start to be identical and will not change with increasing number of bars.
However I was surprised to see that with every increasing number of bars, values of ma and ma crossover indicators were different in Trader (even with 2000 bars)!

Can you explain the reason for this strange behaviour? What is a theoretical minimum number of bars needed for this strategy in order to run the Trader correctly?

Post's attachments

ma strategy_min bars issue_fxcm_ger30.xml 11.62 kb, 5 downloads since 2015-05-13 

You don't have the permssions to download the attachments of this post.

Re: Big difference between builder and trader

bold wrote:

I re-open this issue, because ma indicators in my strategy have different (and changing!) values in Trader part compared to the Builder. For both parts I use identical data (Ctrl+S from the chart in MT4).

I tried to set different "Recommended chart bars" values in Trader mode/Auto Trader/Trade Settings (300, 400, 600, 1000 and 2000).
I compared values of ma and ma crossover indicators at the identical historical candle in Trader and Builder (e.g. 12th May, 16:50 candle).
I expected that from some "Recommended chart bars" level, indicator readings will start to be identical and will not change with increasing number of bars.
However I was surprised to see that with every increasing number of bars, values of ma and ma crossover indicators were different in Trader (even with 2000 bars)!

Can you explain the reason for this strange behaviour? What is a theoretical minimum number of bars needed for this strategy in order to run the Trader correctly?


Did you try to change the position size conditions '' i.e add and close make it nothing '' and then compare the results

Re: Big difference between builder and trader

Yes, but after I set both "Additional entry signal" conditions to "Nothing", nothing has improved.

I used this occasion to a couple of other experiments:

MA crossover - indicator parameters started to "converge" somewhere over 800 bars. At 2000 and 5000 bars were indicator parameters identical. But why it was necessary to set so many bars?

MA - values of this indicator are always volatile, regardless if I use 300, 800, 2000 or 5000 bars.

Re: Big difference between builder and trader

You're sure you have enough bars in mt? 500 bars is more than enough for your strat.

When you point the cursor on currency pair tab in mt, what does it read? Like I did in my pic.
http://s15.postimg.org/i6f28xdfb/bars.jpg

Re: Big difference between builder and trader

41773 bars

Re: Big difference between builder and trader

It's significantly different on my end also...

On 2000 bars crossover values match, but the first MA is lost even with 5000 bars.

It shouldn't need so many bars, something is not right.

Re: Big difference between builder and trader

It seems there is a problem with the MA shift. When there is no shift, the values match. I'll check what is happening.

Re: Big difference between builder and trader

The bug with MA shift is confirmed. I'll publish an update with a fix on Saturday.

With the fix, 1000 bars should be enough to make the values equal.

In order to eliminate values discrepancies I plan to make a tool that will calculate the exact number of required bars for both the bridge and the exported EAs.

Re: Big difference between builder and trader

Bold, please try this solution: FSB Pro v2.7.2. Preview (Licensed users only!)

Re: Big difference between builder and trader

Well, it looks considerably better, but still, there seems to always be a little discrepancy there.
I tested 300, 400, 500, 600, 100 and 2000 bars on the sample strategy.
MA - indicator values were correct in the whole range.
MA crossover - indicator values with 300, 400, 500 and 600 bars were slightly different from the correct value (and also from each other). While values with 1000 and 2000 bars were correct.