|
PureBytes Links
Trading Reference Links
|
Yuki,
Like you, by far the most important new development for me would be
improvements to the backtester. You mention having more DD and
equity curves for a basket of stocks rather than just the worst one
or the most recently run one.
Although item I would like to see in AB is a backtester that can
test a ranking system for stocks. Conceptually this might be easier
to describe than for TJ to code, but here it is:
...your AFL code here...
RANK = Close/MA(Close,50);// A simple momentum rank to illustrate
PORTOFOLIO (10, RANK, Ascending);//10 is the max stocks at one time
AB could do this in a 2 pass approach:
Pass 1: looks at every stock one at a time and for each day puts its
ticker and RANK number in a top 10 list (a list for every data bar).
If the next stock covered has a better ranking than the lowest in
the top ten, the worst in the top ten is dropped and the next
stock's ticker and RANK value for that date is added.
Note; if the system has a market timing dimension, then there will
be data bars that have competely empty RANK lists (a stock only gets
added to the rank list if it qualifies as a "buy" from the AFL code
AND if the stock has one of the 10 best rank list for the day.
Pass 2: AB files the portfolio with 10 stocks and when exits signals
come (perhaps stop loss or profit target), a new one is added using
the RANK list for the bar (or next bar, depending on settings).
Pass 2 could also create a composite equity curve which could be
displayed as a graph and which could also be used to calculate the
portfolio DD that you wish to see.
I am sure TJ will likely find a way to do it than is better than I
describe (and has a lot more features).
Yuki, like you I look forward with great anticipation to the day AB
gets more backtesting features for portfolio trading.
b
--- In amibroker@xxxxxxxxxxxxxxx, Yuki Taga <yukitaga@xxxx> wrote:
> I haven't voted on any of this, nor do I intend to, other than to
say
> my piece here. I consider the finishing/improving of the back
tester
> to be a matter of the highest priority. So high is this for me
that
> I may well decide to wait to upgrade again until this is
> accomplished. All the things AB can already do, and will be able to
> do in the future, are great. But, they lose a measure of their
> greatness when multiple security back tests of these great things
are
> not 100 percent reliable.
>
> Specifically:
>
> When I see an "overall results" report on a multiple security back
> test, I want to see the true max sys DD, or any other DD figure,
not
> the single worst case for one individual issue, as is now the
case. I
> don't want to have to run an equity curve to find it either,
> especially since the equity curve does not work properly if there
are
> non-matching trade dates in the multiple securities being back
> tested. AFAIC, we do not have a finished, accurate,
> multiple-security back tester until these issues are fixed. We do
> have a pretty darn good back tester, and especially when it
involves
> a single security. But we know there are problems in the multiple
> security realm. Why keep plunging ahead without fixing them?
>
> It's sure not my program, but if it was, I'd want to take care of
> these things before very long.
>
> Best,
>
> Yuki
>
> mailto:yukitaga@x...
------------------------ Yahoo! Groups Sponsor ---------------------~-->
Get A Free Psychic Reading! Your Online Answer To Life's Important Questions.
http://us.click.yahoo.com/O10svD/Me7FAA/AG3JAA/GHeqlB/TM
---------------------------------------------------------------------~->
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
Your use of Yahoo! Groups is subject to http://docs.yahoo.com/info/terms/
|