Amibroker Optimization Result - Factors & Analysis

#1
Some of us have done the optimization analysis of our AFL's in Amibroker to get the best out of them, using the collective wisdom of the internet forums to choose the correct time periods, smoothing factors, OB-OS combinations etc.

The big question is how do you right values from the plenitude of result rows. One of the commonly used factor is "CAR - Compounded Annual Return, and MDD - Maximum Draw Down"

Can i request others to throw some light on other factors to look for or how best to use a combination of these factors to choose optimum values.

Ref:
Amibroker Test Results Expln

How to optimize in amibroker
 
#4
@abhig10 - Thanks for sharing that information.


@brajeevn - The standard error was higher in such situation. So i needed to look for more than net profit and # of trades.

There is also something called "Standard Error" Amibroker says the lesser the better.


For some of my simulations, i got errors as high as 1000+ to 4000+ and for some others(all in portfolio analysis and not individual analysis) less than 100. I am not sure why the variances had occurred.

How about you all?
 
#5
Why dont we each post a sample optimization routine for two indicators say RSI / Stoch K & D.

We can compare the results and hopefully learn something new.
 
#6
Try to optimize two parameters and view 3D optimization chart. Best parameter is smooth mountain peak - which never fail. Forget about standard error etc.. Even K-Ratio sometimes present undesirable result. Your aim is minimum trade and minimum drawn down. If optimization,say daily or weekly, value change abruptly, then it is better not to trade. In a good strategy, the value is uniform or changing smoothly. It is important to note that building confidence is prime factor otherwise you will dig up unnecessary optimization setup column.
Simple things make perfection, but perfection is not a simple thing. Many successful trades use simple trend-line to identify buy-sell setup. So strategy must be simple or easy to digest. In trending market, no optimization is needed, even simple moving average give good result.
IMO back-test using zig-zag indicator (ideal entry and exit) and try to achieve 40-50% in your strategy then you will become very successful.
 
Last edited:
#7
@brajeevn - Brevity at its best :)
strategy must be simple
Are you sure about the 'zig-Zag' Indicator. Does it not use future looking syntax(ofcourse if one doesn't use it such a way in the looping it wont look in the future).
 
#8
Try to optimize two parameters and view 3D optimization chart. Best parameter is smooth mountain peak - which never fail. Forget about standard error etc.. Even K-Ratio sometimes present undesirable result. Your aim is minimum trade and minimum drawn down. If optimization,say daily or weekly, value change abruptly, then it is better not to trade. In a good strategy, the value is uniform or changing smoothly. It is important to note that building confidence is prime factor otherwise you will dig up unnecessary optimization setup column.
Simple things make perfection, but perfection is not a simple thing. Many successful trades use simple trend-line to identify buy-sell setup. So strategy must be simple or easy to digest. In trending market, no optimization is needed, even simple moving average give good result.
IMO back-test using zig-zag indicator (ideal entry and exit) and try to achieve 40-50% in your strategy then you will become very successful.
unfortunally minimun trade gives most false signals, the systems give no robustness,the data sample should be large as possible; the risk of optimization is the overfitting the results may not considerable good in long time but only in short time 3/4 months in daily range i speak, so you must continuosly re-optimize all or try to find a really roboust system the can take good results like 1-2 years this is the real aim and is not easy
 
#9
unfortunally minimun trade gives most false signals, the systems give no robustness,the data sample should be large as possible; the risk of optimization is the overfitting the results may not considerable good in long time but only in short time 3/4 months in daily range i speak, so you must continuosly re-optimize all or try to find a really roboust system the can take good results like 1-2 years this is the real aim and is not easy

@TopTrader - I had to ask this, since it ain't going into my head, How did you arrive at the conclusion that minimum trade(setting for the indicator) gives most of the false signals. Is that from your personal experience or you ran a back test with those parameters?

And , could be please explain, how does a param optimized for "daily" setting goes invalid within 3 to 4 months. Are you implying something will change fundamentally every quarter. But then aren't we doing technical analysis.

I understand the fact that no-system/strategy can go on for ever and needs constant adjustments to changing environment(except for sweeping statements like 'follow the trend' :) )
 
#10
@13oct08 what @toptrader said is true.. i am confirming with my own experience and failures..

I started WF over 6month period with Netprofit.. it failed after 3 month..
Switched to CAR/MDD.. failed after 3 month..
Switched to UPI which consider many aspects.. failed after 6 month..

I am still wondering what is it going wrong.. i believe if market is trending over Higher Time frame it gives good return with algo but if its in correction phase it fails..

Still looking for solution which could help me to get best of UPI and select Best of Recovery to sort and perform more than 1 or probably top 10 result under backtest..but no luck..

I think there has to be something else which should be used to perform walk forward othee than Amibroker alone..
 

Similar threads