|
PureBytes Links
Trading Reference Links
|
Doubt
it. It says that stocks that ended up in the late '03 N100 are by definition
ones that did pretty well, so using N100 membership as a filter for backtesting
in '95 or '98 is peeking into the future big time. How many current N100 members
went under completely between '95 and '03? I'd guess that'd be 0%, a highly
outlying number.
<SPAN
class=355465417-30122003>
Chuck,
thanks very much for documenting this, even though it's not what many of us
wanted to hear. I mentioned the inherent long-side bias of backtesting current
index components a while ago, but got somewhat pooh-pooh-ed, even with some
supporting data in hand, admittedly not as good. It seemed on a gut level like a
real issue, and sure enough it is.
<SPAN
class=355465417-30122003>
<SPAN
class=355465417-30122003>
I
think there are two main reasons for using index membership as a backtest
filter, one valid, the other not.
<SPAN
class=355465417-30122003>
The
valid reason is if you're using an index as part of your trading system and want
to trade stocks that correlate well with it. To accomplish this, but without
bias, I'd say we need to scan a more neutral pool of stocks and check their
actual mathematical correlation with the index used. That's probably a more
effective method anyway.
<SPAN
class=355465417-30122003>
The
less valid reason is simply as an easy way to pick 100 or 500 or 2000 stocks to
test. Given the non-random nature of that filter, I'd suggest that this is a
bogus approach we should avoid. Better to test all stocks, or use randomly
generated watchlists. (Stay tuned on that last one. I have a tool I've been
using for this that I'm in the process of documenting for public use; I'll post
it when it's soup.)
<SPAN
class=355465417-30122003>
<SPAN
class=355465417-30122003>Dave
<BLOCKQUOTE
>
<FONT face=Arial color=#0000ff
size=2>Chuck,
does
this suggest that all of your systems were simply curve fitted to the current
nas100? Or perhaps simply over optimized?
Regards,
Jayson
<FONT face=Tahoma
size=2>-----Original Message-----From: Chuck Rademacher
[mailto:chuck_rademacher@xxxxxxxxxx]Sent: Tuesday, December 30,
2003 7:05 AMTo: amibroker@xxxxxxxxxxxxxxxSubject:
[amibroker] The demise of 23 systems...
I
thought that I would share the following with you. Perhaps you can
see how the results of my research may affect your own
systems?
<FONT face=Arial color=#0000ff
size=2>
I
had what I thought to be 23 "good" to "very good" systems that I backtested
against a current NASDAQ 100 watchlist. After spending
two days updating symbols and accounting for de-listed stocks, I came up with
nine separate watchlists, as at January 1 each year between 1995 and
2003. Thanks to some of you, I modified my AFL to make sure that I
used the correct watchlist depending on the date. The results...
not one of my 23 NASDAQ 100 systems now backtest with acceptable
performance. Many went from nice, positive returns with low drawdowns to
negative returns and almost total loss of capital.
<FONT face=Arial color=#0000ff
size=2>
What
does it all mean? IMO, working with only a current NASDAQ 100
watchlist can give you a real sense of false security. Not only
are de-listed stocks not in the current list, but many of those de-listed
stocks had lacklustre performance before their demise.
<FONT face=Arial color=#0000ff
size=2>
I
encourage you to think about the impact of using only a current watchlist,
regardless of how that list is constructed, for backtesting
purposes. I see it all the time. System designers
create watchlists of high-beta or high-yield or low P/E or whatever stocks
based on current information to backtest a system starting perhaps five years
ago. Think about it!
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
Yahoo! Groups Links
To visit your group on the web, go to:http://groups.yahoo.com/group/amibroker/
To unsubscribe from this group, send an email to:amibroker-unsubscribe@xxxxxxxxxxxxxxx
Your use of Yahoo! Groups is subject to the Yahoo! Terms of Service.
|