Profit Generator EA - page 24

 

Backtest help

Hi,

In order to allow to set optimisation on period parameter used we can change P() function. With BAcktest_Period parameter set to 1, period=1 will set P() to 1, period=2 will set P() to 5,..., period=9 will set P() to 43200.

It needs extern parameter:

extern int Backtest_Period=0;

New P() function:

int P(){ //1st part is initial P() function

if(Backtest_Period==0) {

if(period==0) return(Period());

else return(period);

}

if(Backtest_Period==1) {

if(period==0) return(Period());

if(period==1) return(1);

if(period==2) return(5);

if(period==3) return(15);

if(period==4) return(30);

if(period==5) return(60);

if(period==6) return(240);

if(period==7) return(1440);

if(period==8) return(10080);

if(period==9) return(43200);

return(period);

}

}

I didn't test it but I used it a lot of times before and it should work.

here is a BACKTEST ONLY version of PG 2.7

 
jojolalpin:
Hi all!

I'm ok to make tests but I think (as someone told before) we need a kind of manager deciding (giving orientations) values to tests. It will be essential for newbies to make "well minded" tests.

jojo

I'd like to second the idea of a "manager" to assign the tests as jojo says.

 

Backtesting Continued

Here are what I found works in Backtests so far (90% modeling quality). Need to find good backtest settings for the other currency pairs.

EURUSD (H4)

Stoploss: 28

Take Profit: 13

Longbar: 16

No timefilter

No Trailing Stop

GBPUSD (H1)

Stoploss: 23

Take Profit: 12

Longbar: 18

No timefilter

No Trailing Stop

USDCHF (H4)

Stoploss: 25

Take Profit: 12

Longbar: 16

No timefilter

No Trailing Stop

USDJPY (H4)

Stoploss: 70

Take Profit: 140

Longbar: 18

Timefilter ON 7-20

No Trailing Stop

EURJPY (H4)

Stoploss: 70

Take Profit: 150

Longbar: 22

No timefilter

No Trailing Stop

GBPJPY (H4)

Stoploss: 60

Take Profit: 110

Longbar: 28

No timefilter

No Trailing Stop

CHFJPY (D1)

Stoploss: 50

Take Profit: 100

Longbar: 15

No timefilter

No Trailing Stop

I would keep going but it is after 3 AM and I need some sleep. I would like to get most of the currency pairs that we can done so that we can start demo testing them all at the same time starting next week.

Keep up the good work.

 

Period converter, how long does it take?

I currently converting CHFJPY M1 to M5 (Alpari data from June 04 untill now) and it takes a lot of time. My computer is old (512 Mo and 1GHz CPU) but it usually runs large databases (more than 3M lines) quick. Maybe I have a bug? Anyone have an evaluation of time taken?

About Holyguy7 message, I will take your backtest settings as a base and test different options in a first time. After I will try to test another currency (if my computer doesn't die while converting .

 

Holyguy,

Thanks for your tremendous effort and time spent on this project. I am sure others appreciate your work as much as I do.

Just another suggestion regarding optimization. Let us take the first example of the 4 hour chart of EURUSD. The 10 period ATR of 4h bars for the Euro varies from 20 to 40, plus or minus a couple of pips. Now your stops and profit targets are within that range. Any movement within this time period should be considered as noise and thus the target or the stop can be hit almost at random. In contrast, the SL and TP for the Yen pairs are outside their ranges and perhaps outside the realms of ordinary noise. Of course a large 2 or 3 sigma long bar can affect it, but that is always true in any kind of statistical consideration. Thus, even though you got those results for Euro and the others, statistically I will think that they are random events, and got curve fitted somehow.

However, one thing comforting is that the three majors all have similar SL and TP levels, about 25 and 12 respectively. Is there any way you can see how long each trade lasted on average or atleast check some? If the trades lasted for 2 hours and the average range during that period was 30, then the results are curve fitted possibly due to the way MT interpolates and creates tick data. There is no way of knowing other than using tick data to backtest, which I don't have access to.

Hope this stimulates some discussion, maybe in a separate thread.

Thanks again,

Maji

 

5 minute chart

Maji:
Holyguy, Thanks for your tremendous effort and time spent on this project. I am sure others appreciate your work as much as I do.

I'm too.

This setting may be profitable for EUR with low risk of hit Stoploss but not for other main pairs, also i am looking for 20 pips Take Profit.

v2.7

EURUSD (M5)

Stoploss: 30

Take Profit: 10

Longbar: 15

Period: 60

No timefilter

No Trailing Stop

I will try to find reliable setting with minimum hit Stoploss next week.

 

Some Backtesting Results

I've backtested 40-50 scenarios today on the EURUSD, and here are a couple that both yielded 6000 pips over the past 12 mos:

-----------------------------------------

Period: 60

LongBar: 10

SL: 10

TP: 10

timefilter: false

superclose: false

result: 6140 net pips (profit factor=2.07)

-----------------------------------------

Period: 60

LongBar: 10

SL: 10

TP: 40

timefilter: false

superclose: true

TS: 5

TSA: 17

result: 6653 net pips (profit factor=2.05)

-----------------------------------------

However, both of these scenarios performed horribly on GBPUSD. Is that normal? I would have expected at least a profit on other symbols for something that did so well on EURUSD.

FYI, both tests showed over 83% modeling quality.

 

I'm testing too

holyguy7:
Bruno,

It appears that the backtesting on this EA (as long as there is good model quality) seems to work. I believe it is because it only uses prices and no indicators.

Here are settings that have worked backtesting (90% modeling quality).

EURUSD (H4)

Stoploss: 28

Take Profit: 13

Longbar: 16

No timefilter

No Trailing Stop

GBPUSD (H1)

Stoploss: 23

Take Profit: 12

Longbar: 18

No timefilter

No Trailing Stop

USDCHF (H4)

Stoploss: 25

Take Profit: 12

Longbar: 16

No timefilter

No Trailing Stop

Let's all work together to backtest every other currency pair we can. Let's find good long and short term backtesting that work with other currency pairs. I have tried on the USDJPY and have yet to find a good backtest that will give consistent profits. Perhaps we can all work together to find good backtesting results for every currency pairs.

I need volunteers on this thread to work on backtesting the following currency pairs and looking for consistent results over 1 year. I personally backtest from January 1, 2006 to March 29, 2006 then if I get good results with the backtest I then go back to January 1, 2005 to March 29, 2006 to see if the backtest is still reliable.

Please volunteer on this thread to help backtest the following currency pairs. Please use instructions on getting the best backtesting results possible found HERE

I need people to begin to volunteer to test out one or two currency pair for backtesting. Please volunteer for the following currency pairs and post the currency pairs you are testing on this thread.

AUDUSD

CHFJPY

EURAUD

EURCAD

EURCHF

EURGBP

EURJPY

GBPCHF

GBPJPY

NZDUSD

USDCAD

USDJPY

Thank You. Lets work together.

Hi ! Guys,

I'm following your strings till now. This EA seems really promising.

If you all don't mind, I've started forward testing the settings above since the night of the 31st of March.

I'll post the results every few days if any of you are interested in seeing them!!!

 
jojolalpin:
I currently converting CHFJPY M1 to M5 (Alpari data from June 04 untill now) and it takes a lot of time. My computer is old (512 Mo and 1GHz CPU) but it usually runs large databases (more than 3M lines) quick. Maybe I have a bug? Anyone have an evaluation of time taken? About Holyguy7 message, I will take your backtest settings as a base and test different options in a first time. After I will try to test another currency (if my computer doesn't die while converting .

The converting is almost instant. Don't worry about the warning message you get. Click out of it and just do it again. I usually just get all my converting done with all timeframes in a minute or so. Work very well.

 

Statement for the end of the week. Unfortunately, I did not begin this at the very beginning of the week but a day later. It seemed to do very well. This is a non-optimized account as I just guessed at good settings. As you can see, some currency pairs ONLY ate money. That is because I did no backtesting of these settings. I am going to in the future.

If eveyone that has tested this week can begin to post statements, that would be great for testing at the beginning of next week.

I think we have a winner on our hands.

M15

No Timefilter

Take Profit- 40-60 (JPY pairs all are at 60)

Stoploss: 30

Longbar: 20

Files:
pg_m15_2.gif  6 kb
pg_m15_2.htm  20 kb