EA Factory | Page 5 | Forex Factory

Options

Go to
Permalinks
Print Thread

Similar Threads

Calendar as Fundamental Source at Forex Factory 14 replies

Forex Factory Store 0 replies

Journal Factory posters - how much money are you making? 17 replies

Newbie Factory Journal 12 replies

Attachments: EA Factory
Exit Attachments

EA Factory

  • #81
  • Oct 1, 2006 7:43am Oct 1, 2006 7:43am
Yeah, that was actually two filters: 5 period check for entry, and 2 period after entry neglecting exit signals.

If you set this two periods (5 and 2) as an extern input parameters, we can even use the tester to optimize them for different pairs.

My computer is too slow for backtesting this, it's gona take a while.


Quoting GirlFlyer
On the 5 period check... that coding was there primarily to make sure the entry was made at the correct moment and that it was definitely trending in the right direction, otherwise the program would and could jump in a trade at any point after the crossing and yes, they do jump around in there often at those points. Once the WMA's cross the EMA's... the EA makes its decision and sticks with it even if the WMA's reverse and go back up. The S/L then takes over the exit if a reversal occurs. This way when whipsawing is going on for an extended period of time within a range, multiple exits and entries are not made. It will only produce 1 trade until the range is broken. After the range is broken, either up or down, it will either take a profit or get stopped out, and then wait for another entry. This is what allows this EA to keep from doing multiple entries/exits over a very short period of time thus losing money due to whipsaw conditions.

Hope this helps,
GirlFlyer
Target: Trading naked.
  • #82
  • Oct 1, 2006 7:47am Oct 1, 2006 7:47am
There is a tutorial about how to convert 1M data to other time frames in the link I just posted. It may be helpful.

What about 15M, you mentioned in your code that it seems ok for 15M.

Have you forward tested your EA yet? The timeframe problem shouldn't affect forward testing at all. Was the result consistent to your IM backtesting result?


Quoting GirlFlyer
Aha,

I have 2 years of M1 data and M30 data and 10 years of D1 data in my system. So, the EA ran with those data sets. If I ran the EA in strategy tester on any setting besides M1, the results were downright crappy.
Target: Trading naked.
  • #83
  • Oct 1, 2006 8:00am Oct 1, 2006 8:00am
Quoting aha
There is a tutorial about how to convert 1M data to other time frames in the link I just posted. It may be helpful.

What about 15M, you mentioned in your code that it seems ok for 15M.

Have you forward tested your EA yet? The timeframe problem shouldn't affect forward testing at all. Was the result consistent to your IM backtesting result?
I'm starting forward testing on this tonight with the beginning of the week. As to the 15M, I have it from January 2006. Tests run with the 15M from January to September gave excellent results, too.

I've tried running the Period_converter on my datasets and for some reason I can't seem to get it to run properly. Either it doesn't do anything to the data or after the data is written out, MT4 defaults back to what it had before. The only consistent method I have had to get all the data sets in the history buffers and get them to stay there is if I manually scroll the screens backwards in time. It then downloads the history from the server and it stays there.
  • #84
  • Oct 1, 2006 8:52am Oct 1, 2006 8:52am
Quoting aha
This is how I take it: When price in zone C, price>TTL (top TL) and price<BTL p close and short enter c, zone from again TTL break short, long BTL to move price changes, TL no If it. with deal rules normal the use will we anymore, z in not is then relationship, this change a TL). (bottom long.<>
Let me know if there is a problem here.
Thanks Aha for the new enhancement of the EA, I think so far this is the best we can achieving for the time being. Many members would also like to test it as well so further improvement can then be suggested.

Btw can you also include the trading time zone slots features as some currencies are not advisible to trade on the given period like big news, asian fear session.
  • #85
  • Oct 1, 2006 9:04am Oct 1, 2006 9:04am
Backtesting on 15M EURUSD finished, result negative. From what I saw on the backtesting chart the EA seems not behave as it should be, the 5 period filter did not work at all. The attached is the chart.

aha



Quoting GirlFlyer
I'm starting forward testing on this tonight with the beginning of the week. As to the 15M, I have it from January 2006. Tests run with the 15M from January to September gave excellent results, too.

I've tried running the Period_converter on my datasets and for some reason I can't seem to get it to run properly. Either it doesn't do anything to the data or after the data is written out, MT4 defaults back to what it had before. The only consistent method I have had to get all the data sets in the history buffers and get them to stay there is if I manually scroll the screens backwards in time. It then downloads the history from the server and it stays there.
Attached Image
Target: Trading naked.
  • #86
  • Oct 1, 2006 9:06am Oct 1, 2006 9:06am
No problem, but I would like to get feedback on the v0.2 first... just make sure I didn't create new problems.



Quoting divinefund
Thanks Aha for the new enhancement of the EA, I think so far this is the best we can achieving for the time being. Many members would also like to test it as well so further improvement can then be suggested.

Btw can you also include the trading time zone slots features as some currencies are not advisible to trade on the given period like big news, asian fear session.
Target: Trading naked.
  • #87
  • Oct 1, 2006 9:17am Oct 1, 2006 9:17am
could you point me to any help where i can get the mouteki EA to work. I have v2.0 in my experts folder, and have drgged it onto the chart, but all i note as different is a smallc ross in the top right hand corner of the chart

what am i doing wrong?
Its gonna happen, it will happen.....i know!
  • #88
  • Oct 1, 2006 9:21am Oct 1, 2006 9:21am
Quoting aha

I suppose we can run a quick test for just one or two days of data on both low and high timeframe and checked the entry price to find it out. But my computer is tied up in the backtesting of your EA now.... can anyone do that test?
I did aforementioned test on 1M and 15M, there was no major different on the entry and close price level. So no problem with MT4. The result should be the same if you have 1M data and converted them to other time frame using period_converter.
Target: Trading naked.
  • #89
  • Oct 1, 2006 9:24am Oct 1, 2006 9:24am
Nothing, it's an EA, not an indicator. Just wait and you will see it place trades. You are testing it on demo account right? Please don't try it on live.



Quoting rockrat
could you point me to any help where i can get the mouteki EA to work. I have v2.0 in my experts folder, and have drgged it onto the chart, but all i note as different is a smallc ross in the top right hand corner of the chart

what am i doing wrong?
Target: Trading naked.
  • #90
  • Oct 1, 2006 10:41am Oct 1, 2006 10:41am
GirlFlyer,

I rechecked the code, the 5 period filter is to make sure the cross happened in the past 5 bars, not for keeping crossed 5 bars. I understood that wrong in the beginning. I made an EA like this before and the performance were quite poor. So when I saw the result you posted I thought that must be something remarkably different. I guess I just imagined a filter for this EA and expect the filter will explain the difference.

Would you please post the backtesting chart (you can get the chart by pressing the "open chart" button at the right of the tester window after you finish backtesting) that shows the trade the EA took? I am curious about what makes this EA profitale in 1M charts. Or even better, just post the whole chart (not the picture), you can save the chart in the File->Save as menu.

Thanks,
aha


Quoting aha
Backtesting on 15M EURUSD finished, result negative. From what I saw on the backtesting chart the EA seems not behave as it should be, the 5 period filter did not work at all. The attached is the chart.

aha
Target: Trading naked.
  • #91
  • Oct 1, 2006 11:23am Oct 1, 2006 11:23am
About Mouteki EA, here is something I posted on another thread. If you are interested in contributing your computer time, please let me know.

And here is the link if you would like to know the context:

http://www.forexfactory.com/forexfor...d=1#post109699

Quoting aha
But before adding filters, I think we should optimize a couple of parameters within the parameters first, 3 I can think of: ST, TP, and an offset from TL to recognize as a break. I already implement these parameters in my EA at EA factory and can use the computer to optimize them automatically against history data. Anyone got a fast computer and interested in doing so can just go ahead and do it on different pairs. I can provide guidance if you don't know how to do it. I am just too busy answering questions now since the thread is up and running. [
Target: Trading naked.
  • #92
  • Oct 1, 2006 11:30am Oct 1, 2006 11:30am
Quoting aha
About Mouteki EA, here is something I posted on another thread. If you are interested in contributing your computer time, please let me know.

And here is the link if you would like to know the context:

http://www.forexfactory.com/forexfor...d=1#post109699
OK. what I have to do?

I've done some backtest for July 2006, in order to compare with manual backtest of mouteki. He had 968pip profit, the EA had 863$ loss.!!!
The trend isyour friend untilis about to end...
  • #93
  • Oct 1, 2006 2:32pm Oct 1, 2006 2:32pm
Hi hans,

Are you sure you were testing GBPUSD? I think mine was positive for July...

Anyway, here is what you need to do.

Before you do anything, you need to get the reliable historical data by following the instruction in here:

http://www.forexfactory.com/forexfor...06&postcount=2

After that, you can start optimization by doing the following:

In the tester window, there is a button at the top right called "Eexpert Properties", click on it and click on "input" tab, there you will find a series of parameters. scroll down to the last 4 of them: ProfitToMoveSL, ...., TP_Offset. That's what you will optimize. You may optimize them one a time or all 4 at the same time.

for each parameter you may setup the range you want the EA to try. For example, for TP_Offset, the default is 0, means that the EA will use Mouteki's method to calculate TP. You may set start as -20, step as 5, stop as +20. this is just a example, you can set it to a much wider range and finer step, it's just that the optimization will take longer time. You may set these for all four parameters. At the left of the table you can check which parameter you want to optimize.

After you have done that, click ok to return to the tester window, check the "optimization" check box, you mat also need to recalculate. then "start" and sit back.


Quoting hans_gregor
OK. what I have to do?

I've done some backtest for July 2006, in order to compare with manual backtest of mouteki. He had 968pip profit, the EA had 863$ loss.!!!
Target: Trading naked.
  • #94
  • Oct 1, 2006 3:17pm Oct 1, 2006 3:17pm
Aha, I am glad to know that we are trying to do the same thing.
Attached is the EA I am working on, please take a look.
The difference is I filter trades by the profit points. Besides, I only look back certain number of bars to create the trend lines. So it generates much less trades than yours.
Attached File(s)
File Type: mq4 TrendLine.mq4   7 KB | 630 downloads
  • #95
  • Oct 1, 2006 3:39pm Oct 1, 2006 3:39pm
Quoting aha
Hi hans,

Are you sure you were testing GBPUSD? I think mine was positive for July...
take a look
Attached File(s)
File Type: zip st_motekiEA_july.zip   10 KB | 583 downloads
The trend isyour friend untilis about to end...
  • #96
  • Oct 1, 2006 6:13pm Oct 1, 2006 6:13pm
Quoting aha
GirlFlyer,

I rechecked the code, the 5 period filter is to make sure the cross happened in the past 5 bars, not for keeping crossed 5 bars. I understood that wrong in the beginning. I made an EA like this before and the performance were quite poor. So when I saw the result you posted I thought that must be something remarkably different. I guess I just imagined a filter for this EA and expect the filter will explain the difference.

Would you please post the backtesting chart (you can get the chart by pressing the "open chart" button at the right of the tester window after you finish backtesting) that shows the trade the EA took? I am curious about what makes this EA profitale in 1M charts. Or even better, just post the whole chart (not the picture), you can save the chart in the File->Save as menu.

Thanks,
aha
Aha,

I think you misunderstood me when I said you can run this on the 15M chart. In the backtesting, I had no success with Strategy Tester and the EA on any other chart than the M1 setting. However, within the EA, I have the timeframe in which it analyzes the data.

"extern int TimePeriod = 30; // Key Timeframe to base buy/sell decisions on!! "

that excerpt from the code is where you should change it to 15 to run it on a 15M timeframe and still get some success.

I'm attaching the picture of my chart for the EUR/USD on the M30 timeframe. This was run with strategy tester on M1. It shows very good trade points. If strategy tester is run on any other setting besides M1, the data for the price of the currency pair is "guessed at" or interpolated which results in totally false price action thus the failure of the EA. That's why I insisted it be tested on the M1.

I started this on 09/01/05. It placed its first trade at 08:28 on 09/05 and closed at 10:23. 2nd trade was at 10:23 on 9/05 and closed at 23:32 and so on.

In your picture, if you were running that using the M15 setting on your strategy tester, I'm not surprised it traded like that.

GirlFlyer
Attached Image
  • #97
  • Oct 1, 2006 9:13pm Oct 1, 2006 9:13pm
hi aha


I'm wondering if you have time if you could make a EA for the following system.

when 10 and 40 ema cross up it is a buy signal confirmed
by the body of the candle ABOVE the ema's, opposite for
shorts, 30 min chart. You wait for bar close as well.
No stops, your always in the market, reversing position when you have a cross in the opposite direction and a candle body that supports the cross up or down that is NOT touching the EMA's.
You open with 2 positions, close one at 40 pips, let the
2nd one ride till you get valid signal to reverse position.
  • #98
  • Oct 1, 2006 9:31pm Oct 1, 2006 9:31pm
Sorry for my english.
My point is let's make EA using 4 yr's data.This is not "mega Game" or "easy pic", some people using bilions $$$$ to make $$$.But for sure it happend in the past. The question is....how many times it happend in the past ?. Trends, bars, wthever it is, it doesn't work for me.So.. let the EA read the past to caculate future.Hope it'll woks.
  • #99
  • Oct 1, 2006 9:37pm Oct 1, 2006 9:37pm
Hi GirlFlyer,

The attached is the backtesting chart I got in the settings as the following:

1. TimePeriod = 30
2. Test Priod in the tester = 1M

The first trade was a sell at 9:30, 9/05. It did follow the entry rules, both WMAs crossed both EMAs down, so it was a sell. And it was stopped out.

The second trade was a buy at 23:39, 9/05. It also followed the entry rules, both WMAs crossed both EMAs up, so it was a buy. And it was stopped out too.

The third trade was a sell after the 2nd trades was stopped out. It also follow the rules because the WMA crossed down EMAs in the past 5 bars, so it enter sell. its profit has increased over 60 pips at some point so the SL has been moved and it was stopped out then. It was the first winning trade.

In your chart your 1st and 2nd trade was different from mine. I can understand your 2nd trade. But you must have filtered out my 1st and 2nd trade with some filter (is it the PointAdjust? the default value here is 0.0001 for EURUSD), and I don't understand what is the entry rule behind your 1st trade, the WMAs did not cross EMAs in any direction.

If you can explain this, then I have a feeling that we are getting somewhere why our test result are so different.

aha



Quoting GirlFlyer
Aha,

I think you misunderstood me when I said you can run this on the 15M chart. In the backtesting, I had no success with Strategy Tester and the EA on any other chart than the M1 setting. However, within the EA, I have the timeframe in which it analyzes the data.

"extern int TimePeriod = 30; // Key Timeframe to base buy/sell decisions on!! "

that excerpt from the code is where you should change it to 15 to run it on a 15M timeframe and still get some success.

I'm attaching the picture of my chart for the EUR/USD on the M30 timeframe. This was run with strategy tester on M1. It shows very good trade points. If strategy tester is run on any other setting besides M1, the data for the price of the currency pair is "guessed at" or interpolated which results in totally false price action thus the failure of the EA. That's why I insisted it be tested on the M1.

I started this on 09/01/05. It placed its first trade at 08:28 on 09/05 and closed at 10:23. 2nd trade was at 10:23 on 9/05 and closed at 23:32 and so on.

In your picture, if you were running that using the M15 setting on your strategy tester, I'm not surprised it traded like that.

GirlFlyer
Attached Image
Target: Trading naked.
  • #100
  • Oct 1, 2006 9:45pm Oct 1, 2006 9:45pm
Man, your code is hard to read. . I didn't find a single line of comment.

Could you briefly discribe your system a little before I dive in?

aha


Quoting GP2X
Aha, I am glad to know that we are trying to do the same thing.
Attached is the EA I am working on, please take a look.
The difference is I filter trades by the profit points. Besides, I only look back certain number of bars to create the trend lines. So it generates much less trades than yours.
Target: Trading naked.
0 traders viewing now