Topic: Greatly differing results in FSBPro backtest and live EA trading

Hi everyone, I'm tearing my hair out with this problem (which is bad news as I don't have much hair).

It's possible that I'm making a basic mistake in how FSBPro works, or it could be a bug with FSBPro in which case I'm sure it will be helpful of me making the developers aware.

I'm happy to provide much more detail if needed (let me know what you'd need) - but a basic description of the problem is as follows:

I ran an EA developed from FSBPro over the past week on my live trading account, which resulted in 6 EURUSD trades (1 winner and 5 losers). When I uploaded the latest price data from the broker into FSBPro at the end of the week, FSBPro says the strategy should have produced 5 trades (all winners). Upon closer inspection, the trades also seem to start at different times (between live and FSBPro).

I'm obviously concerned at the difference in how the EA has actually traded in real conditions, and the FSBPro backtest results for the same period.

Just to confirm, I'm using the broker data from MT4 imported into FSBPro, and I am NOT using MT4 for backtesting.

Has anyone else encountered a similar problem? I searched the forum and there seems to be a similar problem that was solved by importing the broker data into FSBPro (which I do anyway).

Thanks for any help or thoughts, AL

Re: Greatly differing results in FSBPro backtest and live EA trading

Hey, it'll be important to get some more info. Maybe share your results or the strategy so that others can look at it. In the help section are still posts about such problems maybe you can go and use the search function of the forum.

Re: Greatly differing results in FSBPro backtest and live EA trading

there can be many causes for this.....one is that the strategy is fitted to the data you used..., another is slippage at your broker and a third is lag time.

You will have to research execution carefully.

Be sure to apply MonteCarlo extensively so as to be aware as to potential variations.

I wonder how many indicator slots you are using in your strategy..?

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

Re: Greatly differing results in FSBPro backtest and live EA trading

Post the strategy, probably an error in some custom indicator.

Re: Greatly differing results in FSBPro backtest and live EA trading

alwayslearning wrote:

Has anyone else encountered a similar problem? I searched the forum and there seems to be a similar problem that was solved by importing the broker data into FSBPro (which I do anyway).

Thanks for any help or thoughts, AL

Yes, I had problems like this. So far I noticed big differences in the results by different problems which are already mentioned elsewhere on this forum. In my case I resolved most big differences like this:

- I used mt4 data manually copied from my broker like you did. However, I copied from the mt4 platform on my laptop. I made a good EA with nice results in the backtest on FSBPro, but I had different results on the live account. Why? Because I just let my EA run on my VPN. The problem was resolved by copying the data manually from my mt4 platform directly on the VPN where I eventually trade live. That caused differences for me.

- I got also differences from using too many indicators in an EA. I resolved the problem by deleting all the EA's and create new, simple and better strats!

- I still could improve some strategies by using "simple" smoothing period instead of exponential, smoothed or weighted.

- Also had problems with the base price from indicators. There can be incoherences. I make only strats with base price "close".


Maybe it can already put you on a new path. GL !

Re: Greatly differing results in FSBPro backtest and live EA trading

Hey guys, I just wanted to quickly check in and post a quick thanks for all the prompt and helpful responses :-)

You have given me plenty thoughts to go and investigate further and maybe even try to resolve the issue myself.

Please be assured that I will post the results of my investigations here.

A quick piece of additional info is that my strategies only use max of 2 slots - hey I'm a simple guy and I don't like complicated things! :-)

Also, the strategy runs on a VPS, but I've been pulling the price data down to FSBPro from my local PC - I wonder if this could cause inconsistency?

My main suspicions at the moment are slippage and the simple smoothing thingy!

Thanks again and I will report back, but any other thoughts welcome.

I'm happy to post the strategy if that helps?

Re: Greatly differing results in FSBPro backtest and live EA trading

Post the strategy, thanks!

Re: Greatly differing results in FSBPro backtest and live EA trading

Hi folks, strategy and associated live MT4 trades attached for the past week.

The results of my initial investigations based on your previous replies:

- I've now downloaded the broker data from the VPS where the strategy is actually run from (and not my local MT4 on my laptop), which gives confidence that the data source is now correct.
- Another (different) strategy that is also run on the VPS, is giving near 100% correlation with the associated FSBPro results (it has a few pennies difference here and there which is of no concern and the trades open/close at the same time). So I'm confident it's not an FSBPro-wide problem, but limited to a particular strategy (the one attached here).
- I've experimented with the "Oscillator of MACD" smoothing settings, and selecting Simple definitely brings the results closer, so I'm thinking that maybe the problem lies with this indicator.

Thanks again for any help or guidance, AL

Post's attachments

EURUSD 12Mar17 1hr.xml 10.92 kb, 11 downloads since 2017-03-19 

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

Re: Greatly differing results in FSBPro backtest and live EA trading

and associated MT4 trade results...

Post's attachments

MT4 trade results.txt 839 b, 10 downloads since 2017-03-19 

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

Re: Greatly differing results in FSBPro backtest and live EA trading

https://s9.postimg.org/oz5quckz3/bal.jpg

Yep, it's a mess, not working correctly. The screenshot is pretty self-explanatory, Ocd Macd is differing a lot and giving opposite signals. I'll try to make sense of it.

Re: Greatly differing results in FSBPro backtest and live EA trading

footon wrote:

Yep, it's a mess, not working correctly. The screenshot is pretty self-explanatory, Ocd Macd is differing a lot and giving opposite signals. I'll try to make sense of it.

Hey footon, thanks for confirming - and I appreciate your investigation.

Thanks, AL

Re: Greatly differing results in FSBPro backtest and live EA trading

Mark, I've fixed one bug. Please test it also, I got it to work much better now.

Download the file and put it in the following folder: Forex Strategy Builder Pro\User Files\MT4 Files\MQL\Forexsb.com\Indicators

Re-export your strategy.

Post's attachments

OscillatorofMACD.mqh 6.98 kb, 12 downloads since 2017-03-19 

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

Re: Greatly differing results in FSBPro backtest and live EA trading

Elsewhere I reported problems/failure in getting an EA exported to both Oanda MT4 and FXCM Mt4 Demo accounts to perform as it did on FSB Pro.  In particular, I developed EAs for attachment to M15 Charts that usually opened a M30 Fractal and had M30 Gator oscillator logic condition and a M15 Accelerator Osc logic condition. Bar closing Exit with M30 Upper & Lower Envelopes logic conditions.  These EAs work well on FSB Pro, but when exported and attached to M15 charts Don't work.  They open trades, but never close them.

Post's attachments

122_EURUSD15M.xml 14.67 kb, 3 downloads since 2017-03-20 

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

Re: Greatly differing results in FSBPro backtest and live EA trading

Footon, thank you for the fix.

I'll publish it as an EA Code automatic update.

Re: Greatly differing results in FSBPro backtest and live EA trading

footon wrote:

Mark, I've fixed one bug. Please test it also, I got it to work much better now.

Download the file and put it in the following folder: Forex Strategy Builder Pro\User Files\MT4 Files\MQL\Forexsb.com\Indicators

Re-export your strategy.

Hi footon, I've done so and got it running on MT4 today - I'll report back with the results once some trades have happened - might be a couple of days depending on when they open/close.

Thanks very much for looking into this!

Re: Greatly differing results in FSBPro backtest and live EA trading

Hi footon, I can confirm that the fix has worked - trade open and close correlated between live MT4 trade and subsequent FSBPro backtest.

Thank you very much!

PS I've discarded the strategy going forward as I think it's overfitted - but at least the faulty indicator has been discovered and fixed.

Thanks again!

Re: Greatly differing results in FSBPro backtest and live EA trading

Don't mention it! And thank you for notifying about this issue, this kind of feedback is very useful.

Best of luck!

Re: Greatly differing results in FSBPro backtest and live EA trading

You're welcome, happy to help!