|
PureBytes Links
Trading Reference Links
|
my 2 cents:
- optimization for multiple data points - walk forward - can be quite successfull. Most non-parametric approaches like NN are based on this approach.
- optimize for a portfolio to create a "robust" system? - it will be curve-fitted to a portfolio anyway...
- what are you optimizing? %wins? Measures like sortino ratio that account for stdev of negative returns are preferred.
- create a portfolio of 10-20 random data sets and optimize this "virtual portfolio" to create robust systems.
- use adaptive systems - no need for brute force optimization here.
- "optimize" for market conditions: some systems work better in trending/high volatility environments, etc...
- optimized *OR* "robust" systems? If you want to achieve systematic diversification probably it is better to USE BOTH approaches.
Here are some commercial packages for tradestation regarding walk-forward-optimization. Does anyone have experience using these or know of other similar products?
www.rinafinancial.com/OptiLogix.asp
www.meyersanalytics.com/wfopt.htm
D.S.
-----Original Message-----
From: Ron Hudson MCSD, MCDBA [mailto:ron@xxxxxxxxxxxxxxxxxx]
Sent: quarta-feira, 25 de Junho de 2003 22:57
To: omega-list@xxxxxxxxxx
Subject: Reoptimization and RevMark System
To other trading systems developers on the Omega News List, EliteTrader, and
TSWorld:
I have been evaluating John Bonin's Revmark and ATM systems
(www.talkingtools.com). The systems are ES daytrading counter-trend systems,
which is something I need more of in my ES day trading systems portfolio.
I have run 2 year performance reports in TS7 and posted both of them on my
web site (www.tradingteacher.com). Here are the links to the reports, which
I think are very interesting:
http://www.aneel.com/ron/trackrecord/!web/atm0623.mht
http://www.aneel.com/ron/trackrecord/!web/revmark.mht
As you can see from the equity curves, these two systems are optimized for
the last year (PF for the last year is in the 3 area on both systems). The
vendor is one of those who believes, as he puts it, that "backtesting an ES
day trading system for more than 4-5 months is futile."
Based on the handsome profit factor of his systems during the last year, I
am intrigued by this claim that "re-optimization is good." I really haven't
heard or read about this before. I would like to solicit advise from others
who have opinions about this. For example, I know that Josh Murakami
(www.tradingsystemdesign.com) subscribes to a similar philosophy.
Just when I thought I was beginning to understand that the longest,
smoothest equity curve was the secret to systems trading, now I'm not so
sure. I would appreciate anything others can share about this. The
consistent profits of John Bonin's systems over the last year are quite
attractive.
I would especially be interested in books, web sites, or products related to
this subject. The most valuable feedback would be real-world experience from
developers who are familiar with the pros and cons of reoptimization, and
tips and tricks for doing it right. Also, is there a difference between
"reoptimization" and "curve-fitting." How do we know where to draw the line?
Is it just one of those judgement calls that comes from experience and
wisdom?
Ron Hudson MCSD, MCDBA
Tradestation 7 EasyLanguage Specialist
Client-Server Systems, Inc.
Huntington Beach, CA
www.tradingteacher.com
|