forex software

Create and Test Forex Strategies

forex software

Skip to forum content

Forex Software

Create and Test Forex Strategies

You are not logged in. Please login or register.


Forex Software → Expert Advisor Studio → Data horizon difference in results

Pages 1

You must login or register to post a reply

RSS topic feed

Posts: 9

Topic: Data horizon difference in results

Hi guys,

So I’ve been using the data horizon to leave the most recent month out of the reactor to create strategy’s using 11 months of data then after the strategy has past through to the collection I untuck the data limit box so I can the results of the strategy in the most recent month.

However I find when I do this the strategy has completely different results even over the 11 months that we’re included to create the strategy.

I go from a highly profitable strategy to a strategy that drains the account.
When I check the journal to see if it is placing trades on the same dates and times with the same results they are all incorrect.

The EA should be placing trades identically over the those 11 months as the only difference is that I’m now including another months worth of data.

Has anyone had this issue before?

Re: Data horizon difference in results

I also do it the same way as you do almost daily and never experienced that - the backtesting results prior to the last month you´ve just enabled never change here. There must be something else that you are doing "wrong". Can you make a quick screen recording video of this issue? Then we can surely help very quickly with this.

Re: Data horizon difference in results

I’ve now tried using the date start limit as well and found as soon as I change the start date for backtesting the results also change dramatically and basically every trade is a loser.

So basically if I tick the data start date or date limit every strategy in the collection have loosing trades almost every trade as soon as I untick the box the results are great.

Getting very very frustrating

4 (edited by geektrader 2018-12-03 04:37:08)

Re: Data horizon difference in results

That´s what I thought. The trades prior to the previous backtest can ONLY change if you also mess with the start date. And the simple solution: if your strategies go from positive to negative just by changing the start date, they are no good as this means they are depending on the trading chain. To be honest, this is a typical beginners mistake. Any strategy that fails a change of the "starting bar/date" and turns it from positive to negative, needs to go to the bin right away, they are not stable enough in this case because of that trade chain dependency.

There is actually even a Monte Carlo test to eliminate strategies with exactly that problem (every good trading strategy generator offers that) and it´s called "Randomize backtest starting bar". So you should use this feature (I do too especially if generating strategies on just 11 months of data) and such strategies will automatically be discarded, leaving you with only the ones that are stable enough to survive a change of the start date.

Oh and if you are frustrated already by now, you can as well stop trading Forex, to be very HONEST. It took me ~5 years of daily work on trying to create profitable automated strategies. And once you have those, you will face broker and market manipulations, slippage, all kind of funny stuff that will again make them unprofitable. It´s just an honest advice, this is no market you will make a "quick buck" in with just little efforts, only if you see this as a real business you can really succeed, and even then it´s incredibly hard. If you don´t believe me, you´ve most likely heard that EU brokers need to publish the profitability of their users because of new EU rules. Just look at a few EU brokers you find on Google and go to the front page - the profitability of their users must be displayed at the bottom of their main frontpage. There you will see that the 73% to 90% of the traders their lose money! And this does not mean that the rest is profitable because most of them are breaking even or hovering above/below the 0 profit/loss line most of the time. What I am really trying to say: don´t waste your time if you don´t want to put hard work and I mean REALLY hard work into this and have nerves of steel ;-)

If you still want to continue, I am happy to help with further questions. :-)

Re: Data horizon difference in results

@geektrader -- nice explanation.  I also like using Monte Carlo.

A question -- I was wondering whether you could elaborate a bit more on what you mean by "trading-chain dependency" and how that occurs.  I mean -- I can imagine how it could happen with certain strategies that keep track of the number of open positions and limit the number of open positions.  But is that the only way?  Most strategies created by EA Studio and FSB Pro simply generate signals.  If a strategy only generates signals and does not limit the number of open positions then is there another way for it to fail due to a trading-chain dependency?

Actually -- I just thought of another.  What if your account balance is limiting such that you can only trade X lots at any one time.

Thanks...

Re: Data horizon difference in results

Thanks for the reply!

Even if I change ONLY the finish date of the back test the strategy completely changes all entry and exits it should not do this. I understand what your saying about changing the start date but the finish date should t have this effect.

Don’t worry I know it’s not easy! I put in hours every day even working a 50+ hour week. There’s no easy buck to made trust me I know that all to well.

The thing I’m getting frustrated with is this is clearly a bug which throws you back to square one and without finding out why or how you have no confidence in generating new strategy’s to begin with.

Once I get in front a computer I’ll post up some more info with screen shots



geektrader wrote:

That´s what I thought. The trades prior to the previous backtest can ONLY change if you also mess with the start date. And the simple solution: if your strategies go from positive to negative just by changing the start date, they are no good as this means they are depending on the trading chain. To be honest, this is a typical beginners mistake. Any strategy that fails a change of the "starting bar/date" and turns it from positive to negative, needs to go to the bin right away, they are not stable enough in this case because of that trade chain dependency.

There is actually even a Monte Carlo test to eliminate strategies with exactly that problem (every good trading strategy generator offers that) and it´s called "Randomize backtest starting bar". So you should use this feature (I do too especially if generating strategies on just 11 months of data) and such strategies will automatically be discarded, leaving you with only the ones that are stable enough to survive a change of the start date.

Oh and if you are frustrated already by now, you can as well stop trading Forex, to be very HONEST. It took me ~5 years of daily work on trying to create profitable automated strategies. And once you have those, you will face broker and market manipulations, slippage, all kind of funny stuff that will again make them unprofitable. It´s just an honest advice, this is no market you will make a "quick buck" in with just little efforts, only if you see this as a real business you can really succeed, and even then it´s incredibly hard. If you don´t believe me, you´ve most likely heard that EU brokers need to publish the profitability of their users because of new EU rules. Just look at a few EU brokers you find on Google and go to the front page - the profitability of their users must be displayed at the bottom of their main frontpage. There you will see that the 73% to 90% of the traders their lose money! And this does not mean that the rest is profitable because most of them are breaking even or hovering above/below the 0 profit/loss line most of the time. What I am really trying to say: don´t waste your time if you don´t want to put hard work and I mean REALLY hard work into this and have nerves of steel ;-)

If you still want to continue, I am happy to help with further questions. :-)

Re: Data horizon difference in results

First, if you find a bug, please report it with screenshots, strategies and settings files in order to fix it as soon as possible.

@geektrader you are competently right about the "trading-chain dependency" and thank you for mentioning the Monte Carlo's "Randomize backtest starting bar". It is designed to test such cases.

On the other hand, I'm not convinced that there is no other reason for the case. Michael, please be sure you use the same data set.

We try to make the strategies as reliable as possible. To solve similar cases EA Studio has a "secret" algorithm - "Concrete Entry Rules". It tries to reduce the negative effects of the starting bar especially when you run a generated strategy on a trading account. I'm pretty sure that this is also one of the unique features of our software. I'll make a separate post for that.

Re: Data horizon difference in results

Even if I change ONLY the finish date of the back test the strategy completely changes all entry and exits it should not do this.

Please give details in order to reproduce the issue.

Re: Data horizon difference in results

Michael, that is great to hear that you are at it like that - hope you see success soon! Still, I am not able to reproduce your issue, if I only change the ending bar, it´s all fine. Let´s just make a video about the issue, upload it to YouTube and post the link, then we can surely help.

Posts: 9

Pages 1

You must login or register to post a reply

Forex Software → Expert Advisor Studio → Data horizon difference in results

Similar topics in this forum