Optimization Period

New comment
 

Hi guys,

I'm optimizing now my expert, I' using all data from the history (1999-2008) but I'll like to ear from other people how many years or months are the best choice.

I know that diferent pairs and diferent experts should have diferent optimization periods, but yours inputs will be apreciated.

As far as I know, for my model 2 years is the best optimization period.

Thanks

forex.sb

 

FSB

IMHO, you should be aware of the quality rather than the amount!!!

You can only read so much from past data as the worst possible market action for your EA has yet to happen...

MetaTrader data (live or historical) is useful as a baseline for roughing out EA's.

For meaningful results on a single-broker demo account, optimise from 3-12 months (every tick of course) with 100 trades as your target minimum.

Should the optimised EA make 'money' & be moved to a live account then... re-optimise on the live data

Re-optimise every 3 months

Do not 'Download' history data, you will end up with MT History - always use chart data from the respective account

----

Note

An EA that succeeds on MT data can fail on a brokers demo account

An EA that succeeds on a brokers demo account can fail on a live account

Be prepared for this and be ruthless if you have one that does not cut the mustard on a live account.

This is one reason my EA's have systematic rather than cuddly names (e.g. Phoenix, Firebird etc) - its much easier to bin 'SVOSM EURUSD M5' than 'Karina 1.2'

----

You may have spent months on a bot, but if it wont cut it on live, its got to go & you've got to move on

---

You can debate whether Profit, Profit Factor, Drawdown%, Win/Loss%, etc is most important in test results

What is a really good quick pointer to a bot that is likely to succeed on live accounts is a gross loss figure of the same order as the net profit

Bots with, say, gross profit of 10,000 & gross loss of 8,000 over 12 months would make me very nervous

One with GP=3,000, GL=1,400-1,600 would be very likely to make me happy :)

Another good pointer is an equity curve growth line which doesnt break the uptrend at any point over a 12+ month period

Give me a steady earner over a rocket - that then re-enters... had one of those, was lucky to only lose €300...

---

Hah - ramblings over - now what was the question?

My 2c worth

-BB-

 
BarrowBoy:

Do not 'Download' history data, you will end up with MT History - always use chart data from the respective account

How can I get the data from the brokers instead of metaquotes?

About the %drawdown, what is the "confortable" value?

thks

forex.sb

 

FSB

Some brokers like Alpari have historical files, which you download and then Import to the History Center.

I'm a bit uncertain exactly where these come from so I prefer to get it off the chart as follows:

(This is an out-take from a Help file that is to go out with my retail EA's)


"If that amount of (history) data is not avaliable:
Open an M1 chart of the correct currency pair
Go Tools/Options/Charts
Set 'Max Bars in History' to maximum available
Set 'Max Bars in Chart' to maximum available
Click back on the chart
Go File Menu, Save As csv file
Go Tools Menu/History Center and Delete data in all periods for the Set 'Max Bars in History' to maximum available
Delete all data for the correct currency pair at all time-frames
Import the M1 data from your csv file
Develop data for other periods progressively up from the M1 by double-clicking on the grey 'drum' icons on the left
########################################################################################

# DO NOT 'DOWNLOAD' DATA - YOU WILL END UP WITH METATRADER HISTORY WHICH HAS NO VALUE FOR A LIVE BROKER ACCOUNT #
########################################################################################"

-BB-

 

FSB

> About the %drawdown, what is the "confortable" value?

You cant say in isolation from other factors because this result varies so much with proportion of lot size used in the test compared to account size...

To generalise, if a realistic lot size has been used in the test, then:

<=10% is good

10-20% may be OK

>20% beginning to make me nervous unless Win% is high

>40% probably not ready to use on live!!!

-BB-

 
newbie here but have all ready lost 800.00 which is alot to me but wondering why my ticker hardly moves is it becauce I'm trying to do this on wireless laptop? Is there fast server I should be using? ForexAutoPilot loser. Help anyone
New comment
Reason: