[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [amibroker] On Robustness, Post #1



PureBytes Links

Trading Reference Links




Test *unoptimized* system on small, mid & large cap stocks in bull, 
bear & sideways market conditions, same parameters for all.  

 
Mark: I agree up to a point. 
However, I'm not convinced that a system must perform equally well across all 
markets using the same parameter values, as you suggest. For example, if your 
system performs admirably on the large cap stocks but not on the small cap 
stocks, and within the large cap group the parameter values are robust in and of 
themselves (i.e., no local maxima on the response surface diagram), why not 
trade the system on large cap stocks? Also, if you optimize the same system on 
small cap stocks and find the parameter values are different from the large cap 
parameters but are robust, why not use those different parameters for the small 
cap issues? Why do the parameter values HAVE to be the same for all market 
types?
 
I use the stocks of the S&P 600, 400, and 500 indices and 2 year 
bull, bear and sideways periods (for a total of 6 years per 
stock). Rationale behind this: to find systems that profitably *tested out 
in the past* on a large number of (somewhat tradeable) stocks of varying market 
caps in multiple sectors under different market conditions, under 
the assumption that this indicates the system is robust enough to 
profitably *trade select issues in the future*. 
 
What I like to do is optimize 
over a period that encompasses BOTH bullish AND bearish periods at the 
same time, like 1/2/99 to 12/31/01 or 02 as opposed to selecting separate 2-year 
bullish and 2-year bearish periods. Usually, systems have a short component 
and a long component to them (unless you only like to trade either long or short 
per se). So, optimizing the parameters in a trading period that includes 
both bull and bear market action lends a certain degree of robustness in 
and of itself if you find a reasonably flat response surface. Then, when you 
test in the succeeding OOS period and get a reasonable performance similar to 
the IS period, with no local maxima in the 3-D space of the trading diagram, 
then you have a fairly robust system in my mind. 
 
...My commission setting(s) in AB: proprietary, based on my *slippage* 
research using data from actual trades.  But you could choose an arbitrary 
say, 1% to get started.  
 
What about setting 0 commissions 
just for starters when developing your system. After you have done sufficient 
optimizing and forward testing, then set your commissions/slippage to your 1% or 
0.5% or whatever and then see if the system crashes. If it does, trash it. 

<FONT face=Arial 
color=#000080> 
Looking forward to your 
remaining 8 posts. 
<FONT face=Arial 
color=#000080> 
Al 
Venosa
 
 
 
---Outgoing mail is certified Virus Free.Checked by AVG 
anti-virus system (<A 
href="">http://www.grisoft.com).Version: 6.0.525 / 
Virus Database: 322 - Release Date: 10/9/2003






Yahoo! Groups Sponsor


  ADVERTISEMENT 









Send BUG REPORTS to bugs@xxxxxxxxxxxxx
Send SUGGESTIONS to suggest@xxxxxxxxxxxxx
-----------------------------------------
Post AmiQuote-related messages ONLY to: amiquote@xxxxxxxxxxxxxxx 
(Web page: http://groups.yahoo.com/group/amiquote/messages/)
--------------------------------------------
Check group FAQ at: http://groups.yahoo.com/group/amibroker/files/groupfaq.html



Your use of Yahoo! Groups is subject to the Yahoo! Terms of Service.