Trading Naked EA | Forex Factory

Options

Permalinks
Print Thread

Similar Threads

Naked Chart Trading 122 replies

Bow's trading journal 2 (almost naked trading on m5) 26 replies

Peaks and Valleys (Trading Naked) 692 replies

Trading Naked - Reading Candlesticks and other Price Analysis 146 replies

Learning Naked Trading 0 replies

Attachments: Trading Naked EA
Exit Attachments

Trading Naked EA

  • #1
  • First Post: Feb 2, 2007 11:56am Feb 2, 2007 11:56am
Hi everyone, I am new to forex and this is my 1st post in FF.

I have developed an EA solely based on price action without any indicators. So far I have backtested it on GBP/USD from July 04 to Jan 07 on Alpari MT4 only and the M1 data was downloaded from Alpari data bank.

Here is the snapshot of the backtest performance. Please take note that no money management is involved in the backtest.
http://img261.imageshack.us/img261/3337/backtestah6.gif

Anyway, I am not sure whether you guys are interested because it cant double or triple up your money in short period. If you are interested please let me know your test result especially if you have an older databank.

The attached EA is the beta 1 version therefore I dont provide user input option at the moment. I prefer to set the variable in the coding in development stage so that it is easy for me to recall it for testing. And, please take note that this EA only perform backtesting on GBP/USD only.

Hopefully this is not another curve fitted EA...
Attached File(s)
File Type: zip Trading Naked Beta 1.zip   10 KB | 1,826 downloads
Trade For A Living SOMEONE'S GOING TO DO IT, WHY NOT YOU?
  • #2
  • Feb 2, 2007 12:11pm Feb 2, 2007 12:11pm
NO forward testing done yet?
  • #3
  • Feb 2, 2007 12:22pm Feb 2, 2007 12:22pm
Quoting et_phonehome_2
NO forward testing done yet?
As you can see in the snapshot, there is only 121 trades in 2.5 years. So such kind of opportunities hardly exist everyday or may be week.

Anyway, the best setting is not included in the beta 1. I still can boost another 50% extra profit without significant changes in other statistic number.
Trade For A Living SOMEONE'S GOING TO DO IT, WHY NOT YOU?
  • #4
  • Feb 2, 2007 7:27pm Feb 2, 2007 7:27pm
Quoting TheChargingBull
As you can see in the snapshot, there is only 121 trades in 2.5 years. So such kind of opportunities hardly exist everyday or may be week.

Anyway, the best setting is not included in the beta 1. I still can boost another 50% extra profit without significant changes in other statistic number.

what is the purpose of your thread?

If you want to sell your expert then go to commercial section...

If you want people to help you improving the resutls then share the method and the source file
  • #5
  • Feb 2, 2007 7:54pm Feb 2, 2007 7:54pm
He is sharing the method. Perhaps he wishes to protect his source code. I know I would from Ebay thieves.

Although, you really should allow forward testing on this EA, Backtesting is only a part of the test.
  • #6
  • Feb 2, 2007 10:09pm Feb 2, 2007 10:09pm
May I ask what are the entry rules of your system? do you use fixed TP and SL?

-Mikhail
  • #7
  • Feb 2, 2007 10:51pm Feb 2, 2007 10:51pm
  •  tesla
  • Joined Oct 2006 | Status: Friendly Neighborhood Programmer | 533 Posts
It's kind of hard for me to figure out what kind of feedback you're looking for. Are you looking for input on the modeling?

I don't know what the system does since you didn't share the source.
  • #8
  • Edited 11:38pm Feb 2, 2007 11:35pm | Edited 11:38pm
Quoting smjones
He is sharing the method. Perhaps he wishes to protect his source code. I know I would from Ebay thieves.

Although, you really should allow forward testing on this EA, Backtesting is only a part of the test.
Yes, you are right. I do not want people C&D (Copy & Develope) and then commercialize my hardwork without my permission.

Anyway, you still can do forward testing. I havent done any forward testing because I just finished the basic coding a few days ago. However, it may take about a year to have 50 foward tested results before we know whether this is a curve fitted EA. Therefore, I hope someone who has an older databank (unseen data which is out of my coverage) can help me to do backtest. If this EA cant archieve a good result by using older data then no point i continue the development, right?

Please bear in mind that this coding is still in preliminary stage (I am yet to name the EA) because the only exit strategy is either it hit SL/TP or close by the end of the day, and it doesnt have any money management strategy too. The basic element must work 1st before i further enhance the coding.

Quoting mikhaildgreat
May I ask what are the entry rules of your system? do you use fixed TP and SL?

-Mikhail
Ok the basic concept of my EA is very very simple. We enter with the same direction of breakout only if the price pullback from previous day high/low. However, this basic idea doesnt work alone because it will have a lot of noise.

The coding part that spend most of my time is filtering. Depending on the filter setting, it will create 3 type of equity curve pattern, which is consistently going up, going down or no direction.

Therefore I have created another set of coding which is totally reverse the original concept, instead of buy we sell. And this make the consistently going down equity curve turn into consistently going up.

In order words, the beta 1 has included the basic idea of entry and the reverse entry.

So far the pullback/SL/TP is fixed so that it is easy for me to do statistic analysis. The current setting in the beta 1 is 20/30/60.

Quoting tesla
It's kind of hard for me to figure out what kind of feedback you're looking for. Are you looking for input on the modeling?

I don't know what the system does since you didn't share the source.
I apologize if my presentation skill confuse you. I am not good to present myself but I'll try to answer to all enquiries. Basically I just want to know whether it is just my luck to create a potential curve fitted EA, or it worth more effort to continue the project.

P/S: I apologize for my broken language too.
Trade For A Living SOMEONE'S GOING TO DO IT, WHY NOT YOU?
  • #9
  • Feb 3, 2007 12:06am Feb 3, 2007 12:06am
I Believe your findings are valid. seeing you have a large sample in your backtest you could have found something here. and your entry being so simple you could manually backtest your strategy or use visual mode, and if it works then it works, Only thing left to do is forward test the idea.

Regards,
Mikhail
  • #10
  • Feb 3, 2007 12:31am Feb 3, 2007 12:31am
Well, it does work eventhough i change my pullback/SL/TP as long as I do not violate the basic of statistic & probability. I have randomly backtest some entry manually and it looks ok.

However, the biggest concern is whether the statistic & probability number of the unseen data match the number in my data coverage. I guarantee this EA will fail if they cant match.

Here i provide another snapshot of performance with not so good setting. the pullback/SL/TP is 25/25/70.
http://img240.imageshack.us/img240/6753/backtestbb1.gif

Well the performance is much lower than the 1st snapshot(Setting was 20/30/60) but the pattern of equity curve doesnt change. I still manage to get higher high lower low type of pattern (but with more noise).

However the main concern remain unchange. Will i have the same luck if backtest data is out of my coverage?
Trade For A Living SOMEONE'S GOING TO DO IT, WHY NOT YOU?
  • #11
  • Feb 3, 2007 12:32am Feb 3, 2007 12:32am
  •  tesla
  • Joined Oct 2006 | Status: Friendly Neighborhood Programmer | 533 Posts
Quoting TheChargingBull
I apologize if my presentation skill confuse you. I am not good to present myself but I'll try to answer to all enquiries. Basically I just want to know whether it is just my luck to create a potential curve fitted EA, or it worth more effort to continue the project.
There are a couple things to look for which can be warning signs of curve fitting.
1. If you've got lots and lots of parameters and test over every possible combination. It doesn't sound like this is the case for you.
2. A small number of trades entered or short backtest period. The less trades entered, the easier it is to curve fit to get sparkling results. I think 121 trades is a bit on the light side but doesn't shout "danger" to me.

It's a good idea to optimize over a period and then check your results on "out of sample" data. Example: optimize on Jan 1 2003 through Jan 1 2006 and then check results without further optimization for Jan 1 2006 to present.
  • #12
  • Feb 3, 2007 1:04am Feb 3, 2007 1:04am
Quoting tesla
There are a couple things to look for which can be warning signs of curve fitting.
1. If you've got lots and lots of parameters and test over every possible combination. It doesn't sound like this is the case for you.
Basically there are only 4 parameters but can further divide into 8 parameters (4 for original entry and the remaining 4 for reverse entry). But so far i use the same setting for original and reverse entry.

2. A small number of trades entered or short backtest period. The less trades entered, the easier it is to curve fit to get sparkling results. I think 121 trades is a bit on the light side but doesn't shout "danger" to me.
The number of trades are depend on the filter. If i tighten the filter then the number of trades will be lesser. The number of trades without filter is about 800 trades within my data coverage.

It's a good idea to optimize over a period and then check your results on "out of sample" data. Example: optimize on Jan 1 2003 through Jan 1 2006 and then check results without further optimization for Jan 1 2006 to present.
Ok I'll check.
Ok i need to go for lunch now...i'll continue update my progress.
Trade For A Living SOMEONE'S GOING TO DO IT, WHY NOT YOU?
  • #13
  • Edited 2:24am Feb 3, 2007 2:04am | Edited 2:24am
I can appreciate your comcern for your intellectual material, but really I have been trying to figure out I could be of any help in testing this. I fail to see how I can test it in a way that you have not. There are no parameters to offer any meaningful difference of testing opinion. So I think I will pass on this thread from this point...

Hope it works out for you... Take care, Scott
  • #14
  • Edited Feb 4, 2007 1:48am Feb 3, 2007 3:04am | Edited Feb 4, 2007 1:48am
Quoting smjones
I can appreciate your comcern for your intellectual material, but really I have been trying to figure out I could be of any help in testing this. I fail to see how I can test it in a way that you have not. There are no parameters to offer any meaningful difference of testing opinion. So I think I will pass on this thread from this point...

Hope it works out for you... Take care, Scott
I apologize for the inconvenience. I'll come out with user input option when I release the next version.
Trade For A Living SOMEONE'S GOING TO DO IT, WHY NOT YOU?
  • #15
  • Edited 7:40am Feb 4, 2007 2:48am | Edited 7:40am
Enclosed is the beta 1.1. I would appreciate it if I could have some feedback.

Changes:
1. Added user input option.
2. Added trailing stop feature.
3. No longer limited to GBP/USD only.
4. Added entry & exit arrow on the chart.
5. Pending order expiration changed to 3 hours before end of the daily bar.
6. Minor bugs fixed.

Recommended setting for GBP/USD(I havent tested on other pairs):
1. TP_pip : 60 to 180;
2. SL_pip : 20 to 30;
3. TS_pip : minimum 50;
4. SafetyMargin_pip : 18 to 22;

I hope the below picture can provide you a better visualization if you have no idea what is the thing behind the code.


http://img211.imageshack.us/img211/2...ngnakedrm9.gif

Thank you for your interest.
Attached File(s)
File Type: zip Trading Naked Beta 1.1.zip   12 KB | 1,630 downloads
Trade For A Living SOMEONE'S GOING TO DO IT, WHY NOT YOU?
  • #16
  • Last Post: Feb 4, 2007 8:08am Feb 4, 2007 8:08am
  •  tesla
  • Joined Oct 2006 | Status: Friendly Neighborhood Programmer | 533 Posts
It's still an ex4. Not much I can help you with friend.
0 traders viewing now