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

Re: [amibroker] How to spot over optimizing



PureBytes Links

Trading Reference Links


 
<BLOCKQUOTE 
>
----- Original Message ----- 
<DIV 
>From: 
Herman vanden 
Bergen 
To: <A title=amibroker@xxxxxxxxxx 
href="">AmiBroker 
Sent: Sunday, July 14, 2002 7:47 AM
Subject: [amibroker] How to spot over 
optimizing

Optimizing has been discussed at several times on this list, but never 
haveI seen a check-list that could help you identify over optimizing.I 
welcomeadditions to the small list below. If we can identify degrees 
ofover-optimization we can use the same tool to rate robustness, this 
becausethey are somewhat complementary.If any of the following 
statements are True you may be over-optimizing:1) The system works on 
only one stock2) The gains are erratic with a final big gain3) 
Optimization values are scattered (random peaks on a 3D chart)4) Gains are 
too good to be true5) The equity line is flat6) The system only works 
in recent years7) It doesn't work out-of-sample (I use 8y dev, 2y outof 
sample testing)8) Optimizing parameters vary widely for different time 
periods9) Greater than 75% winning trades7) ... what can you 
add?
A.  Results obtained in single 
trending market.
B.  Insufficient number of 
signals.
C.  Caution if not tested with out 
of sample data.
It seem thatif you 
overlay the equities for many systems many appear tofail for the same 
periods (use 10 year data), I take this as a sign that thesystem pickup 
on a true market anomaly. It seems to indicates that thereare times that 
the market is incoherent (chaos rules) and that most systemswill not 
perform well. Combining equities can produce a "Market Coherence"(please 
suggest a better name) indicator which would tell you when tradingis 
highly risky. Or taking the inverse; a "Market Chaos" indicator. 
Bravetraders will flip all their signals during those periods 
:-)Since there are many parameters one can optimize I take note of the 
gainimprovement that comes with selectively activating various stagesof 
thesystem. I start with the most basic form and add/measure features as I 
addthem. An unreasonable large jump in gain for any one change makes it 
suspectand I will pick it apart manually for a few 
trades.------------------------ 
Yahoo! Groups Sponsor ---------------------~-->Save on REALTOR 
Fees<A 
href="">http://us.click.yahoo.com/Xw80LD/h1ZEAA/Ey.GAA/GHeqlB/TM---------------------------------------------------------------------~-> Your 
use of Yahoo! Groups is subject to <A 
href="">http://docs.yahoo.com/info/terms/