RE: [GT] backtest_multi.pl

["Weigert, Thomas" <weigert AT mst.edu>]

I also found some defects in the code when generating the analysis
results which causes the system to crash (divide by zero error). From
this I infer that this script is not widely used, otherwise, somebody
would have complained about the defect.

Therefore, I will make switch this script over from taking input from a
mysterious XML file to taking the same input as backtest_many.pl.

The only difference that the user will see is that there can only be one
broker defined, as in backtest_many.pl. Once somebody tells me how to
interpret multiple brokers, I can extend this scheme to take in multiple
brokers as input.

Cheers, Th.

-----Original Message-----
From: Weigert, Thomas [mailto:weigert
AT
mst.edu] 
Sent: Sunday, April 20, 2008 6:06 PM
To: devel
AT
geniustrader.org
Subject: RE: [GT] backtest_multi.pl

I went through the code of backtest_multi.pl to see how it differs from
backtest_many.pl.

While I cannot say that I understand precisely what this analysis is
trying to accomplish, I have observed the following differences:

- detects new opportunities based on "parked systems" rather than just
  the current system, and runs through all calculators for that.
- includes a "theoretical" analysis.
- performs the analysis in a single portfolio, rather than a set of 
  Portfolios.

Note that a comparison to a BuyAndHold strategy does not make sense in
this mode of analysis, as we are considering many securities at the same
time, unless we want to compare to BuyAndHold of an index.

Currently the code just ignores the BuyAndHold and puts 0 in all its
values.

Th.

-----Original Message-----
From: Weigert, Thomas 
Sent: Sunday, April 20, 2008 11:14 AM
To: Weigert, Thomas; 'devel
AT
geniustrader.org'
Subject: RE: [GT] backtest_multi.pl

The inputs for backtest_multi.pl that define system and markets (as well
as initialization) are taken from an XML file, the format of which is
not explained and needs to be inferred from the code. 

Not all of that format is intuitive, e.g., that "system-manager" must be
explicitly defined, or what the "init" is. Also, "mm" appears to be part
of the system definition, not stand-alone. 

In any case, it is completely inconsistent with the rest of the scripts.
These inputs could just as well be given using the system and market
files as in backtest_many.pl, which would be less confusing and easier
for the user who wants to use both.

I propose that we change the input for backtest_multi.pl to use the same
mechanisms as backtest_many.pl or backtest.pl.

Th.

-----Original Message-----
From: Weigert, Thomas 
Sent: Monday, April 14, 2008 7:46 AM
To: 'devel
AT
geniustrader.org'
Cc: Weigert, Thomas
Subject: backtest_multi.pl

The backtest_multi.pl script has no documentation. The one item that is
not clear in the arguments for this script is the file that is given as
an argument.

Does anybody know what the content of this file is? It appears that it
has an XML representation of a set of system descriptions, but this
format is nowhere documented.

Does anybody have an example file that can be fed into backtest_multi?

Thanks, Th.