|
PureBytes Links
Trading Reference Links
|
Jim,
[I'm replying to the list because others might find my reply
interesting.]
>Doing a little homework on the Swinger family. I too recalled
>something about periodic re-optimization. Did you ever reach a
>conclusion about this? If you bought the system(s), would you
>recommend them? Interested in any opinions you've formed. Thanks.
Vladimir corrected me about re-optimizing each month. What he does
is evaluate the system's performance each month, but he has not
needed to re-optimize for a couple years. I've noticed myself that
I can improve the system's performance by re-optimizing it. I can
also improve it by totally eliminating one of the rules, making it
less conservative. So I would believe his statement that he has not
re-optimized in a while.
I bought Swinger2-EOD. It uses a short moving average plus some
range filtering for entries, and implements several different exit
strategies. Vladimir insisted that the system performs poorly if
you change it to trade on the open instead of the close, but I found
it to work almost as well. It also goes against the idea of "let
your profits run." That is, it exits a trade as soon as you make a
predetermined profit regardless of whether you would have made more
if you stayed in. However, this seems to be a good rule, because
eliminating it made the overall results worse. Long profitable runs
aren't as common as short reversals.
One thing I did that I found interesting was to filter its entries
against signals generated by Oddball. What I got was something that
had almost the same profit as the original Swinger, better than
Oddball, but with less time in the market than either Swinger or
Oddball, and therefore less risk exposure.
In fact, I would say that any S&P system using only OHLC data would
benefit from an Oddball filter.
One thing that disturbs me about Swinger is that there are many
parameters and they aren't adaptive. They are fixed. If the S&P
experiences a significant change in its characteristic True Range
or average price velocity, Swinger will fail without re-optimizing.
For example, the current parameters don't work at all prior to 1996.
I haven't found a combination that does, yet.
--
,|___ Alex Matulich -- alex@xxxxxxxxxxxxxx
// +__> Director of Research and Development
// \
//___) Unicorn Research Corporation -- http://unicorn.us.com
|