|
Hello Patrick - the party just wasn't the same
without you... 8 - )
( I actually thought of you a couple of times -
noticed that you seemed to be "missing in action" and thought maybe you had gone
off to Hawaii for the winter. No such luck, eh? 8 - )
Thanks very much for the invitation to collaborate
- I would be happy to accept, and I am planing to do some work with the
Portfolio BT, but I should probably mention up front that, at this point, I
think it will be more of a "spare time" thing for me. My current situation is
that I am wrapping up one project and nearly at the point where I
can take on another, but in fact I already have a big project planned,
and I am actually pretty eager to get started on it. ( I am going to do a
major overhaul on my old code for self-optimizing
indicators - take a few of the current ones I like
best, add a few new ones I want to try, convert them all to native AFL and
dump the DLL, make some improvements like adding dynamic smoothing
and OB/OS levels ( everything controllable from
Params dialog ), then try to code and test a few ideas I have for new
optimization algorithms which will hopefully allow them to automatically
optimize for "most robust" and some other things rather than just
the single most profitable params, and finally try out some
simple combinations to see if I can come up with a decent auto-opt
system for individual issues. I will be starting pretty soon and will
probably finish it up from the nursing home. 8 - )
But then it occurred to me the
other day that maybe I should have a preliminary look at a few other
things before I get wrapped up in this new project, and the portfolio BT
was one of them. After the initial look, I think that I would still like to
give priority to the other project, but if you would be interested in putting
our heads together "part-time", I would be happy to give it a try - just let me
know. It sounds like you may already have some degree of understanding when
it comes to using this thing - I really think that I will need to spend a bit of
time figuring out how to make the best use of it before I can actually
begin to do any serious testing with it. If by any chance you should happen
to be at the same level as me ( i.e. - clueless 8 - ), then
perhaps working together would benefit both of us. Let me know
what you think.
Steve
PS - Good to have you back!
----- Original Message -----
Sent: Friday, April 28, 2006 7:47
PM
Subject: Re: [amibroker] Portfolio
Backtester
Hi Steve,
I'm again late to the party, having been preoccupied since the start
of the year (or a bit earlier) with family issues, "other than trading"
business issues, property issues, tax preparation issues and several more
distractions to my ability to keep up with AB. This week I finally
started to break out getting things behind me (or I'm getting ahead of
them?). Quite coincidentally, I decided to have another look at
the portfolio backtester just as you have apparently started to do
so. And like you, I've mostly used the individual backtester
heretofore.
What I see (maybe incorrectly) is that it is great for fully
testing a trading system and potentially gaining sufficient confidence in
a strategy to implement it. While one can also optimize the strategy, I
believe some care must be taken with this so as not to curve fit to the
data. There are several books (Perry Kaufman comes to mind as one
author) on developing trading systems and perhaps I need to read more about it
all -- however I've a number of ideas I've been jotting down in a spiral
notebook during my development hiatus which I'd love to code and test.
Some are from things I've been doing by the seat of my pants lately (with some
very good returns, but perhaps it's luck), some are just notions I've
had for possible explorations and subsequent entry or exit
triggers. Most of my trading for the past several years has been based
not on indicators, but on patterns such as Jeff Cooper, Larry Connors, Linda
Raschke, and others have written about with my own flavorings added to the
stew. And most of my trading is intraday or up to perhaps 3 days on
daily setups. Trying to eyeball thousands of charts to see if a strategy
makes a buck isn't my idea of how I want to spend my time while the market is
closed.
I've read the User's Guide several time thru and I've coded some systems,
mostly simplistic, as I'm still trying to learn all the different
features to see how they behave. What I've been doing is using some of
my exploration codes which I use normally to identify stocks to trade,
then adding code for the entry and exit rules (along with all the
other minutia necessary) and seeing how the system performs. Once again,
Dan Clark's imagery of AB as an onion with layers of complexity which each
user can peel down to seems so apt.
So off I go into the wild blue yonder of portfolio backtesting.
Want to fly together? (And if anyone thinks I'm dead wrong, about
to crash, or has good tips and suggestions for learning, please speak
up.)
Peace and Justice --- Patrick
----- Original Message -----
Sent: Friday, April 28, 2006 3:29
PM
Subject: Re: [amibroker] Portfolio
Backtester
Hi Ara,
Thank you for the reply! Well, I have
to admit that you make it sound pretty simple and
straigntforward, and I think maybe you zeroed in on my confusion
when you mentioned "stock selection system", but to be honest I think I
am still missing something very basic. If you don't mind a
follow-up question, I will try to explain:
Using the individual backtester, I select my
watchlist and then run my system over the list. When the test is
finished, results show each stock and how well it performed under
that system. I review the results to see how well the different stocks
performed and to determine what stocks, if any, that I might consider
trading with that system - stock selection is not usually something that I
do prior to testing, except to establish a "tradable universe" based on
various price/volume calculations. Also, I do use the
Positionsize variable with the individual backtester to see how
reinvesting different amounts would affect performance.
Now switching to the portfolio
backtester...
OK, I understand that in real life we will
have an account with X amount of dollars, and that it will probably be split
up between several positions ( our portfolio ), and that we will try to
remain fully invested if possible, assuming we can find enough good
looking opportunities to invest in. So I run the portfolio backtester
on a list of, say 1000 stocks, and the results list shows all trades
that were taken, and it shows that, at some point or another, a
position was taken in maybe about 10% of the stocks in the watchlist,
which I gather were the ones that had an entry signal and the best
score shortly after some cash became available. The other tickers
tested don't show up at all in the results list, so apparently no position
was ever taken in these stocks. When I look at these results, I just
don't know how to begin to interpret them.
So I guess my real question is "What can I
learn from looking at the results of a portfolio backtest?"
I get the feeling that the
correct approach to using the portfolio backtester should
be totally different than the approach for using the individual
backtester, but I don't know what that approach should be. Should I
test with a smaller list of stocks - a predefined portfolio?
Should I infer that those stocks that were actually traded are the best
stocks to trade with that system? - that sounds like a lot to infer to
me, but maybe I am missing something. Could you
possibly explain briefly how you might approach a
portfolio backtest (i.e. - what stocks might you test together and
why?) and then how you might interpret the resulting report? Any
further enlightenment you can provide would be greatly
appreciated!! 8 - ) Thanks again!
Steve
----- Original Message -----
Sent: Thursday, April 27, 2006 8:52
PM
Subject: Re: [amibroker] Portfolio
Backtester
Steve
The idea behind Portfolio backtest is to
simulate trading closer to reality.
Original Backtester would allocate ALL
available funds to each trade. Not a real situation since no one has
infinite $$$.
Portfolio backtest will work with your
account size and buy only as many stocks as you can afford. You can
specify max number of stocks you want, and it will allocate funds evenly
... etc. so it works with your account parameters, while still testing
your stock selection system. It will buy additional stocks only fter you
sell some and free up some money.
----- Original Message -----
Sent: Thursday, April 27, 2006 1:26
PM
Subject: [amibroker] Portfolio
Backtester
Hi All,
I started looking into the portfolio
backtester today for the first time, and I am a bit embarresed to admit
it but I have absolutely no idea what to do with it. Functionality is
explained OK but there is very little to explain the theory behind
portfolio testing or trading. I did a little hunting on the internet but
am not exactly sure what I should be searching for. Is this for
people that do basket-trading or something similar? I would greatly
appreciate it if someone could explain how and why I would want to
use this, or if possible point me to some helpful web pages. As I have
said, I know nothing about it, so even the very basics would be great.
Any help appreciated - Thanks very much!
Steve
Please note that this group is for discussion between users only.
To get support from AmiBroker please send an e-mail directly to
SUPPORT {at} amibroker.com
For other support material please check also:
http://www.amibroker.com/support.html
SPONSORED LINKS
YAHOO! GROUPS LINKS
|