To use chat, please login.
Back to contacts

Martingale Back Test Accuracy.

DoubleTrap (ForexSeeker)
May 17 2010 at 00:42
814 posts
Back test base on martingale grid probably will have the nearest accuracy with forward test.
Simple indicator like MA and stochastic that trigger it to open trade then follow a fix grid of money management system of martingale.

The reason to back test martingale is to find out how much initial deposit you need for the system.
You can not back test a whole year as the profit you make will increase your initial deposit every month.
You need to back test on every month using a fix deposit.

So my brother did write a martingale EA, and I have improve it a bit and find the following result.

AlpariUK - Demo micro acc.
Month : Net Profit : D.D.
Jan'09 : 118 : 1673
Feb'09 : 123 : 315
Mac'09 : 440 : 1377
Apr'09 : 235 : 1001
May'09: 215 : 946
Jun'09: 267 : 376
Jul'09: 73 : 330
Aug'09: 135 : 460
Sep'09: 128 : 443
Oct'09: 140 : 302
Nov'09: 164 : 678

Dec'09 got open trade till end of month so need to combine Jan'10

Dec-Jan'10 : 235 : 1556
Feb'10 : 123 : 315
Mac'10: 73 : 271
Apr'10: 200 : 108.

Conclusion : 3k initial deposit needed. Average ROI (16month) = 5.1%


      
Information is Gold when come to organised.
DoubleTrap (ForexSeeker)
May 18 2010 at 16:18
814 posts
A strange thing happened, today I re-do back test again and find the result is different.
Dec can go through with 3k; and some other month result also different.
And I do back test on Alpari RU, the result also different, in Alpari RU, the draw down seem less compare to UK.
Still a long way to adjust better ROI result, while finding the back test different in each broker.
And when I switch between version 2 and version 3 of EA to redo back test , the result is different again.
Now even no switch EA, do back test 2 times, result also different.
Information is Gold when come to organised.
DoubleTrap (ForexSeeker)
May 19 2010 at 08:51
814 posts
The mystery of back test.

Have look into the folder TESTER. The history folder contain my 1 minute data.
When close down platform, the history data center will need to download data again.
As then I do export the data 1 minute out. And import it again when restart.

Problem occur shut down platform, the memory still have terminal.exe running with 100+MB of memory.
I have to end task it to re-open platform.
As I also reduce the bar history in platform, if make it max. when restart platform will over load with history bar and hang.
This due to my PC is slow type in office, memory not enough.
Press F2 will show up the record 20600/3626606 probably mean history bar/ downloaded history.

When run back test it first take long times as to load all the downloaded history to memory become history bar.
I cancel it and import the save 1 minute data.

And I test another EA , able to have same result 2 times on July Data.
And come to my EA, different result 2nd times, 3rd times o.k.
Month - Profit : DD
Jul - 150 : 107
Jul - 149 : 13.98
Jul - 149: 13.98
Jan - 704, 375
Jul - 149: 13.98

Finally data inconsistency problem solve.
Information is Gold when come to organised.
DoubleTrap (ForexSeeker)
May 20 2010 at 10:22
814 posts
Still the same thing of different result when back test.
But what the hack, the different is minor, as long as it can go through.

Alpari RU ( NewZealand)

EA : SuperMartin - SuperFactor 1.77 - TimeFrame 1 minit.
Strategy : Martingale with no stop loss, using basket so it take profit early whenever got small profit.
               Follow trend indicator and leverage technique like pip leader and then basket.

Month : NetProfit: D.D.
Jan'09 : 454 : 1074
Feb'09 : 246 : 209
Mac'09 : 302 : 1061
Apr'09 : 238 : 117
May'09 : -235 : 2189 ( Probably because of instant basket and no use TP)
Jun'09 : 197 : 849
Jul'09 : 83 : 823
Aug'09: 101: 240
Sep'09: 115: 184
Oct'09: 74: 585
Nov'09: 132: 792

Dec'09 loss money as close at stop end of month, so combine 2 month.

Dec-Jan'10 : 233: 450
Feb'10 : 139 : 133
Mac'10: 164 : 12
Apr'10: 126 : 58.

Result : 3k Deposit require,
            16 month earn : 1951
            Average month earn : 121.93
            Average ROI (16month) = 4.0%
Information is Gold when come to organised.
DoubleTrap (ForexSeeker)
May 21 2010 at 04:42
814 posts
<a href='' target='_blank'><img src=''/></a>
Back Test 2009.
The sudden draw down is when martingale fail after open 20 level.
Average ROI 2009 = 5.8% per month

Attachments:

Information is Gold when come to organised.
DoubleTrap (ForexSeeker)
May 21 2010 at 06:17
814 posts
Open martingale to 30 level.
May'09 will be able pass through with NetProfit :176 and D.D: 1709

Result : 3k Deposit require,
16 month earn : 2362
Average month earn : 147.62
Average ROI (16month) = 4.9%

As monthly test have stop out at end of the month so estimate is 6% ROI per month.

Information is Gold when come to organised.
mirusevxs33
May 21 2010 at 12:08
123 posts

ForexSeeker posted:
    The mystery of back test.

Have look into the folder TESTER. The history folder contain my 1 minute data.
When close down platform, the history data center will need to download data again.
As then I do export the data 1 minute out. And import it again when restart.

Problem occur shut down platform, the memory still have terminal.exe running with 100+MB of memory.
I have to end task it to re-open platform.
As I also reduce the bar history in platform, if make it max. when restart platform will over load with history bar and hang.
This due to my PC is slow type in office, memory not enough.
Press F2 will show up the record 20600/3626606 probably mean history bar/ downloaded history.

When run back test it first take long times as to load all the downloaded history to memory become history bar.
I cancel it and import the save 1 minute data.

And I test another EA , able to have same result 2 times on July Data.
And come to my EA, different result 2nd times, 3rd times o.k.
Month - Profit : DD
Jul - 150 : 107
Jul - 149 : 13.98
Jul - 149: 13.98
Jan - 704, 375
Jul - 149: 13.98

Finally data inconsistency problem solve.

Hi ForexSeeker,
Since the recent MT4 clients no longer offer the ability to use custom FXT files, a special loader is needed. You can download it here, it works only with the MT4 version commonly known as 225
https://eareview.net/download/fxt_loader.zip
it is from https://eareview.net/tick-data#prepareddata
Enjoy!

Miro
mirusevxs33
May 21 2010 at 12:10
123 posts

ForexSeeker posted:
    Open martingale to 30 level.
May'09 will be able pass through with NetProfit :176 and D.D: 1709

Result : 3k Deposit require,
16 month earn : 2362
Average month earn : 147.62
Average ROI (16month) = 4.9%

As monthly test have stop out at end of the month so estimate is 6% ROI per month.


It is terrible result for any EA especially for martingale - you need WAY better ROI IMO....

Regards,
Miro
AussieTrader
michigansurveys
May 21 2010 at 19:38
31 posts
When backtesting a EA, pay attention to the spread. If the spread is different you will get a different result.
DoubleTrap (ForexSeeker)
May 22 2010 at 01:17
814 posts
<a href='' target='_blank'><img src=''/></a>

Alpari UK 1 year back test on 2009. (Home desktop PC)

Average ROI = 10.66%

Attachments:

Information is Gold when come to organised.
mirusevxs33
May 22 2010 at 02:28
123 posts

ForexSeeker posted:
   
Alpari UK 1 year back test on 2009. (Home desktop PC)

Average ROI = 10.66%
Period 1 Hour (H1) 2010.02.23 00:00 - 2010.05.19 23:00 (2010.02.23 - 2010.05.20)

Attachments:

mirusevxs33
May 22 2010 at 02:31
123 posts

mirusevxs33 posted:
  <a href='' target='_blank'><img src=''/></a>
ForexSeeker posted:
   
Alpari UK 1 year back test on 2009. (Home desktop PC)

Average ROI = 10.66%
Period 1 Hour (H1) 2010.02.23 00:00 - 2010.05.19 23:00 (2010.02.23 - 2010.05.20)
not the best one and not that I've use live....
mirusevxs33
May 22 2010 at 02:38
123 posts
and live(real) acc for last week - 05/17 - 05/22
<a href='' target='_blank'><img src=''/></a>

Attachments:

mirusevxs33
May 22 2010 at 02:44
123 posts

ForexSeeker posted:
    Back test base on martingale grid probably will have the nearest accuracy with forward test.
Simple indicator like MA and stochastic that trigger it to open trade then follow a fix grid of money management system of martingale.

The reason to back test martingale is to find out how much initial deposit you need for the system.
You can not back test a whole year as the profit you make will increase your initial deposit every month.
You need to back test on every month using a fix deposit.

So my brother did write a martingale EA, and I have improve it a bit and find the following result.

AlpariUK - Demo micro acc.
Month : Net Profit : D.D.
Jan'09 : 118 : 1673
Feb'09 : 123 : 315
Mac'09 : 440 : 1377
Apr'09 : 235 : 1001
May'09: 215 : 946
Jun'09: 267 : 376
Jul'09: 73 : 330
Aug'09: 135 : 460
Sep'09: 128 : 443
Oct'09: 140 : 302
Nov'09: 164 : 678

Dec'09 got open trade till end of month so need to combine Jan'10

Dec-Jan'10 : 235 : 1556
Feb'10 : 123 : 315
Mac'10: 73 : 271
Apr'10: 200 : 108.

Conclusion : 3k initial deposit needed. Average ROI (16month) = 5.1%


      

write in the ea to log all acc info - balance, margin used, free margin, floating p/l, etc.
and do backtests with big deposit - then - just calc/view log
it saves time and worth more than monthly tests
DoubleTrap (ForexSeeker)
May 22 2010 at 04:25
814 posts
Thank Q for all the suggestion especially the useful information of prepare data.
https://eareview.net/tick-data#prepareddata

Martingale the conventional play is use Hit & Run method.
Which U need get high ROI per month, and safe keep it. Some month will blow account.
So if the ratio of blow account is low and you able have ROI > 100% per month you are the winner.

Martingale leverage which I testing now is target never blow account.
So the ROI certainly is low because I need to set it low risk sure will get low gain.
And also find the minimum deposit require to perform martingale management.

So my point of this EA is :
a.) Never blow account
b.) Low deposit as 3k

Later I will try to adjust a Hit & Run Martingale EA is:
a.) Sometimes blow account
b.) ROI > 100%
c.) Deposit ???
Information is Gold when come to organised.
DoubleTrap (ForexSeeker)
May 22 2010 at 04:35
814 posts
<a href='' target='_blank'><img src=''/></a>

Alpari UK 1 year back test on 2009 with 1 minute time frame. ( Server3- Platform 1 )

Average ROI = 12.22%

quote from michigansurveys
    When backtesting a EA, pay attention to the spread. If the spread is different you will get a different result.


Spread got any affection to Martingale strategy ? Maybe for Scalping strategy. So the Scalping EA will have to tweak the allow spread parameter.

My EA-Salvation is using 1 minute time frame none stop trading.
So I estimate the demo forward test will have more gain compare to back test 1 minute.
24May'10 will put on run.

Attachments:

Information is Gold when come to organised.
mirusevxs33
May 22 2010 at 05:14
123 posts

ForexSeeker posted:
   
Average ROI = 12.22%

quote from michigansurveys
    When backtesting a EA, pay attention to the spread. If the spread is different you will get a different result.


Spread got any affection to Martingale strategy ? Maybe for Scalping strategy. So the Scalping EA will have to tweak the allow spread parameter.

My EA-Salvation is using 1 minute time frame none stop trading.
So I estimate the demo forward test will have more gain compare to back test 1 minute.
24May'10 will put on run.


It is possible/is the spread does affect martingale - all is based on closing trades. I've observed it in live - does not closed basket trades because of fraction of a pip! (0.2 pip)....
But in general - it must does not affect...
DoubleTrap (ForexSeeker)
May 22 2010 at 13:32
814 posts
The seeking of never blow account EA seem a far road again.
Back test year 2008 has 2 month of failure. Oct and Dec unable to pass through using 3k.
The reason it blow as it keep on bouncing in range trend which result my leverage system keep on open trade.
Probably have to manual interrupt to cut lose and make escape.

Jan'08 : 166 :52
Feb'08 : 149 : 289
Mac'08 : Close at end of month ( -263 : 638 )
Mac-Apr'08 : 532 : 2032
May'08 : 109 : 320
Jun'08 : 135 : 746
Jul'08 : 143 : 665
Aug'08 : 265 : 542
Sep'08 : 562 : 2671
Oct'08 : Blow account
Nov'08 : 642 : 2354
Dec'08: Blow account

50k Account
Oct'08 : 1021 : 9758
3k Account
9Oct'08 : 675 : 2071

50k Account
Dec'08 : 1139 : 45477
Open till not enough money lot size 38.73.

Demo Alpari UK ( Home Desktp )
Information is Gold when come to organised.
CycleTrader (CycleTrader)
May 22 2010 at 18:24
29 posts
Hi DoubleTrap,

Trough your post here https://www.myfxbook.com/community/trading-systems/hedgepro-real-money/25955,1 i came at your account. Nice tests. I am seeking for the same 'never blow account EA' since my martingale account blow :-)
See here https://www.myfxbook.com/members/CycleTrader/truffelz/20393

But i still like the martingale principle in an EA.
In the related thread pc8multifx told about his experience that basket profit is the key to success. (also short average holding time.)

I think that multiple pairs in one basket and basket profit is indeed important step forward, but then still how do we protect against continious price move against us while martingale has opened big lots?

I think there should be some sort of Hedge come in when a certain negative margin level gets exceeded and hedge close again when price moves in the originally preferred direction again. Or as soon as certain margin level exceeds that the ea not try to close in profit anymore, but try to break even. (Somehow i am not a fan of close in loss ;-)

Although it looks impossible, we should succeed in creating the ultimate never blow EA with a piece of martingale, don't you think?
(it looks like pc8multifx already suceeded with this EA https://www.myfxbook.com/members/pc8multifx/xpipstrendmulti/15435 )

DDmaster (DDmaster)
May 31 2010 at 07:33
43 posts
you have very bad modeling quality , only 25% ....

you whant at least 90% and that even can be unreliable for back tests
Our greatest weakness lies in giving up, the certain way to succeed is always to try just one more time.
Please login to comment .