Gogetter EA - page 6

 

A little better

I've got alot of bugs in this still...

Files:
ggl2x_1.gif  7 kb
ggl2x_1.htm  360 kb
 

a little better settings...

there is still more to do...

use are your own risk.

Files:
ggl2x_2.gif  8 kb
ggl2x_2.htm  374 kb
 

About your backtest...

Could you read this link

http://www.strategybuilderfx.com/showthread.php?t=15309

and make your tests with 90% modeling quality?

 
asmatic:
Could you read this link

http://www.strategybuilderfx.com/showthread.php?t=15309

and make your tests with 90% modeling quality?

I have struggled to get the alpari data into the history center correctly...I fail miserably. I can only hope that someone who'se more successful at getting it installed in their platform will produce better tests. Alas, I would love to see better modeling quality. I'm making do with the best I can get until this "work in process" that they claim the metatrader is...actually sees fit to improve the historical data problems.

I can't get the same results from one day to the next...this report I posted last night...I can't make it duplicate today to save me...I guess it only does that on wednesdays.

to account for my testing methods...

yes i read the thread...thanks for that...

I ALWAYS click recalculate.

I ALWAYS use tick mode.

I have some historical data from alpari but I know it's not all 1 minute data.

perhaps when I get to come up for air I'll make another attempt to get the alpari 1 min data all the way back to 2004 properly converted and available....

I have these huge data historical files from alpari, but can't seem to get them into the history center. Somehow I just don't think a couple of Gigabyte files should condense into only a couple months of 1 minute data.

converting data? I dunno. I obviously don't have it figured out yet. sorry

 

Running rough on 3 out of 9 cylinders...

Last night before I went to bed I told my wife, "yea it looks like it might work if simply by letting it sit overnight something in the cosmos doesn't decide that it won't do the same thing in the morning that it's doing tonight...

sure enough...I woke up the next day (today) and lo and behold running the same settings over the same time period that the night before showed winning $4600 in the morning bankrupted the account.

So I rolled up my sleeves and tore all the signals apart, reprofiled each of them and no matter what I tried I couldn't get it to redo what it did yesterday...

in desperation having beat the crap out of my code I came back to the site and downloaded the EA that I posted here last night to start over again clean from a place that at least worked for one day...

I saved the new file as 'backup' and started hammering on it again. I rebuilt my spreadsheet to show me ALL the profiles at the same time...Actually I spent more of the day in my spreadsheet than hammering on the code...

finally I reprofiled the signals a third time, I discovered that the signals change depending on which settings are turned on to 'true'. If I only use the main signal I get a different set of matches...so that is part of what it making it irratic...I'll have to think about what if anything I can do about that...

In the mean time I also noticed that when I test it the first time I have all 9 signals firing and then as I retest it over and over most of them go away except for two or three....I can't account for this either....

Being left with only two or three cylinders of the original 9 it is designed to run on it runs pretty rough...

by leveraging the crap out of the 2 or 3 that remain constant I managed to get it back to $4700 profitable tonight...but man what a drawdown!! I don't know if I'd have the nerve to watch that kind of account mutilation to see it make it back and give it away two or three times over....

anyway it is what it is...man the GGS was such a lay down compared to this!!!

I suppose I'm learning something from this. Not sure what yet...

maybe tomorrow some amazing insight will rise with the dawn and I'll get it tamed?? dare I say that? It's enough to make me a believer in the force...

ONE MORE THING!!! WHAT GIVES WITH THE TRAILING STOP????

I don't see any trailing stop activity and as I watch this in forward tests I do see it moving...so why doesn't it move on the backtester???

Files:
 

This is crap..

I put the most recent version of GGL in a second chart next to the GGS that has been running there for a few days. This morning I wake up and find that the the platform has executed short trades using the lot sizing from the long program's settings. Something is not working having the two programs run side by side. It's hammered the demo account.

I have no idea why it's crossed over to the other program to do anything the two should be totally independent of each other. Alas

ok i get it now....i forgot to change the ordering code...it's MY fault! oy I need a break

 

GGLv2001 build 2001

Well it helps alot if I'm trying to get longs to open LONG positions rather than short positions when I get a buy signal...lol

Apparently the .htm file is too large to upload 5.44 mb

So I guess you'll have to generate your own backtest report...GBPUSD 30m TF

I think this could be pushed further with more lot scaling but this is far enough to prove the point...

It looks like the main signal wins on the order of 559 wins to 488 losses or about 1.15% so it's not a huge margin but with it being over 1.00 percent means it can be leveraged which is what I've done here. It hits three levels where the lots are multiplied larger and it could be expanded still larger....this just proves the strategy...

and it proves I do better at seeing what's right in front of me when I'm rested up

Files:
ggl2001x.gif  6 kb
ggl2001.mq4  37 kb
 

two tests run on identical settings one right after the other...

fyi, I am posting these so I can show them to a friend who's helping me debug...

I'm also working on a rewrite version....

my goal for the rewrite is to exceed a profit factor of 1.45 and have it be consistent over the 16 month period I have test data for.

 

I'm getting answers to the problems I've had with this...

the htm file was too large to upload so.....

Bars in test 17642

Ticks modelled 688500

Modelling quality 89.82%

Initial deposit 500.00

Total net profit 57158.40

Gross profit 160974.56

Gross loss -103816.16

Profit factor 1.55

Expected payoff 29.01

Absolute drawdown 141.00

Maximal drawdown 5441.50 (8.77%)

Relative drawdown 44.19% (1822.79)

Total trades 1970

Short positions (won %) 0 (0.00%)

Long positions (won %) 1970 (54.52%)

Profit trades (% of total) 1074 (54.52%)

Loss trades (% of total) 896 (45.48%)

Largest

profit trade 2340.00

loss trade -416.00

Average

profit trade 149.88

loss trade -115.87

Maximum

consecutive wins (profit in money) 12 (3020.80)

consecutive losses (loss in money) 14 (-239.20)

Maximal

consecutive profit (count of wins) 6138.00 (9)

consecutive loss (count of losses) -510.80 (4)

Average

consecutive wins 2

consecutive losses 2

I determined that the variance in previous tests was being caused by using data which was larger than one minute in the tester....so i limited this test to the 11 months of data I have one minute data for...as you can see the modeling quality is much better this way and it's more consistentent.

I figured this was the case by comparing the previous tests with each other and finding divergences on the same time ranges with the same trades. The only thing that can account for that is not the performance of the EA but the extrapolations of the strategy tester...this was suggested before but I just didn't know what to do. I wish I had more historical one minute data but this 11months will just have to do for now, it's all I have to work with.

This is not the best I think this ea can perform still....I'm going to do some more with lot sizing and tidy it up before I post the ea. ..This is just to mark my current best.

Files:
 

Aaragorn,

Aren't you proud of yourself to have learnt MQL and have worked on improving an EA's performance?

Congratulations on your tenacity and hard work.