|
PureBytes Links
Trading Reference Links
|
--- Robert Cavaleri <rcavaler@xxxxxxxxxxxxx> wrote:
>
> <skip>
>
> - It's profitability ratio over a variety of
> different equities is
> consistently greater than 65% (i.e. 65% of the time,
> you are in the
> green)
call it part A - not hard at all.
> - It performs well in both trending markets, *and*
> range-bound markets.
call it part B - not hard at all too.
but it is in conflict with part A.
A generalized system always suffer in one of the
two types of condition, and that's why it is applied
on so many items (stocks) to get the percentage up.
> (perhaps using adaptive properties based market
> conditions of a given
> stock)
call it part C - not needed at all.
maybe, step back a bit to rethink your concepts
could help.
> - It does not require constant re-tuning or
> optimization to achieve
> satisfactory results (this would just be curve
> backfiting)
whenever you backtest - you ARE curve fitting, it
is just the degree of curve fit :)
remember backtesting on a single issue means you
assume that there is some historical price movement
behaviours that will repeat again and again within
that single issue.
It implies that backtesting on a basket of issues
you assume that there is a consistent behaviour
across all these issues over their historical data.
when you are trying that, you better analyze the
statistical behaviour of the individual issues first
to see what you can exploit. e.g. similar volatility?
similar price expansion sequence, etc.
(all these cannot be done in TS efficiently, try
Excel, for more power try some stat packages)
>
> I have been doing a lot of systems testing over the
> past year, but have
> yet to achieve a certain level of robustness that I
> describe above..
if you are using TS, I am sorry to inform you that it
is not designed to correctly backtest a basket of
items over a long period of time. Especially we
are talking about the whole universe of stocks.
You may have encountered a correct idea but limiting
the test to your set of stocks could ruin the
historical behaviour.
> Perhaps this is an unrealistic goal, perhaps not. I
> am wondering to what
the goal is realistic, just that no one is interested
to share what they have found.
> extent others may have achieved some of the above
> milestones. If any of
> you have created or come across such a system, one
> that exhibits the
> properties above, and has been consistently making
> money in both up
> and down markets. If so, would you be interested in
> sharing some of it's
> ideas, or perhaps even the system itself with the
> rest of us to examine
> and discuss?
one thing I still remember from the old days (a few
yrs back) I have seen firms doing "arb" on stocks
across the U.S and Canada border. There is no need
of high-end computer to find the currency difference
of mismatched pricing in the bid and ask. These guys
usually tag on a few stocks and trade only those.
They make money 90% of the time year after year.
But some just blow the account away when the "sure"
win is strike by say a stock halt or sudden news.
If your transaction cost is low enough, you
can do this too.
Some trading methods are not secrets but the entry
cost is just too high for most people, like "arb"-ing.
I still prefer index trading myself :)
=====
Lawrence Chan http://www.tickquest.com
Innovative Analytical Software for Trading Professionals
_______________________________________________________
Do You Yahoo!?
Get your free @yahoo.ca address at http://mail.yahoo.ca
|