|
PureBytes Links
Trading Reference Links
|
Fred's eloquent post drew me into this. I hope that you don't mind.
We are both on the FT-Talk board. Also please forgive the length of
this post - I thought that I'd throw a couple of things into one.
Like all of you, my interest is primarily in trading systems that I
can use. I have been using AB for a while, but haven't posted. I've
developed some tools that it is probably time to post.
I agree with Fred. Amibroker is missing a couple of pieces that the
FT-Talk / Trade folks have had for a while and were the catalyst for
many of the systems developed there. The tools allowed the
development of many innovative solutions.
I thought I'd briefly mention two of them -
1. The FT world is very familiar with pair trading. This is
relatively easy to implement in AB. If I get permission from Dave
Serbin, I'll post a generalized pair trading system using his FUBAR
market timing signal translated to AB. One of my reasons for doing
the port was to use AB's facilities to examine how optimized the
parameters were. This can be done in AB by using the Optimize
feature to do a sensitivity analysis on the variables. I consider it
VERY useful.
2. As you probably know from Trade, almost all of the trading modules
can be implemented with the ExprFamilyTrade block. I am in the
process of implementing this as a DLL in Amibroker. (I first tried
an easier way as a COM object, but ran into a roadblock).
I thought I'd briefly describe ExprFamilyTrade for the non-FT folks
in order to elicit some suggestions.
In Trade, there is code that can be used to do calculations much as
in Amibroker. Trade has the concept of families, though, which can
be operated on in one fell swoop. This effectively extends array
math (such as in AFL) to matrix math where the rows are the
funds/stocks, and the columns are the daily quotes.
Indicators from these families, or results from trading the members
individually, can be passed to the ExprFamilyTrade block mentioned
above. One way to think of this block in programming terms is as an
object. Various properties of the object are set, and then a trade
method is executed based on those properties. The highest ranked "N"
members are then traded and the composition of this "set" or
portfolio is allowed to change over time based on constraints.
Properties of the ExprFamilyTrade include -
Family - the time varying ranking values for each member of the family
MMarket - Traded to based on a market timing signal. May also be a
short fund or position.
Positions - number of funds/stocks held at a given time
BuyMinRank - Minimum rank required to qualify for a buy
HoldMaxRank - Maximum rank allowed to hold. When exceeded, the issue
will be sold, and the highest rank issue greated than BuyMinRank will
be bought
HoldMinDays - Minimum hold time
Signal - Market timing signal that caused switch to MMarket
StartDate -
StopDate -
There are some other parameters, but I'll skip them for the sake of
brevity. You might notice a couple that one would consider missing,
and are needed -
StopLoss -
TrailingStop
Position - a committment percentage
Rebalance - when to rebalance the available equity among the holdings
As a final note to offer an example of the use of this, I'll mention
an application developed by W. Gansz called RUTTRFAMSIG. What Werner
did was the following -
1. Trade each member of a family based on a set of MACD, Stoch, RSI
parameters (RUTTR). (In AB, this could be tuned for each member.)
2. Track "productivity" of each member timed with that signal. This
was the annualized return while on a buy, 0 while on a sell.
3. The "on buy" productivity was essentially the rank of the
members. This was passed to an ExprFamilyTrade block which held "N"
funds with the parameters detailed above set to certain values such
as hold 5 issues for a 10 day minimum replacing each one as its rank
fell below 10, and so on.
The main thing to recognize about such a process is that it is
effectively "walk forward testing". The historical productivity
numbers are used to choose the issues to hold. They are then held
for a period of time, and that return constitues the portfolio
return. What makes this entire process work is that the time-varying
ranking (productivity) is available to the trading module for all
issues. (Memory is cheap, and processing power is reasonably
plentiful). This particular approach seems to yield some useful
results with stock baskets.
Anyway, I've gone on for way TOO LONG. My real purpose was to
stimulate some ideas and to offer the thought that with trading
tools, the innovative people on this board will probably come up with
some systems that are very useful and that will amaze us all.
Bruce Robinson
> SIMPLE replacement for that. Hardly the best system in the world
but
> hardly the worst.
> 2. The ranking of trading vehicles, whether they be stocks or
MF's
> based on something like MAR i.e. CAR / MDD during the periods of
time
> that your simple market timing system is on a buy or the inverse of
that
> if you allow shorting as well. This allows for the strongest
trading
> vehicles to be taken advantage of during buy periods and the weakest
> ones to be taken advantage of during sell periods.
> 3. The trading of or possibly even sub timing of those individual
> trading vehicles based on some shorter term buy / sell criteria.
>
> The problem at the moment from an AB perspective is that I don't
believe
> all the pieces are in place to write the code to do this sort of
> analysis.
------------------------ Yahoo! Groups Sponsor ---------------------~-->
Get a FREE REFINANCE QUOTE - click here!
http://us.click.yahoo.com/2CXtTB/ca0FAA/i5gGAA/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/
|