|
PureBytes Links
Trading Reference Links
|
In a message dated 4/15/99 6:45:05 PM US Mountain Standard Time,
kmorgan@xxxxxxxxxx writes:
<< An entry strategy cannot have ANY "accuracy" without some form of exit
method, so "55% best case" is meaningless. One approach would be to
assess "is the position profitable including costs after N bars?".
But care has to be taken even here; what if the price moved below where
you would have put a stop during the course of those N bars? And of
course such an exit strategy may be meaningless to how you are really
going to trade...so you have to really assess it based on your complete
trading set of rules.
Ergo, it's "impossible" to assess any subcomponent of a trading method in
isolation. At best, we can get some characterizations and comparisons of
them based on simplifying (and hence error prone) assumptions.
>>
Kevin,
Good post!
You said, "An entry strategy cannot have ANY "accuracy" without some form of
exit
method, so "55% best case" is meaningless." Then you refer to using an exit
based upon "N" bars. I agree. I trade short term (stocks usually less than
5 days). Therefore, I test entries in TS with exits after 0,1,2,3 and 4
bars. I compare the total profit of the stocks expressed in % to the the S&P
for the same entry date and exit. I do this on 10+ years of data on 500+
stocks. I get thousands of trades. If the total profit expressed in % for
stocks is not greater than the S&P for at least one of the above 5 exits, I
discard the entry without any further tests. If it can't outperform the S&P,
I don't want it as an entry. If it does, I then start to work on its
improvement. If I were to trade financials, I think I would use a financial
index for comparison as I do the S&P for stocks.
Testing exits by themselves is more puzzling than testing entries by
themselves with the "N" bar concept you discussed. I have used code that
makes random entries with a specific exit but I was not happy with it.
Therefore, I test some exits ( I can't do this with all exits) by:
1. Selecting TS's option to make "multiple entries with the same entry
signal" and then
2. Use the common entry..."If dayofweek(date)=1 then.....I run 5 tests, 1
for each day of the week and then combine the results for a performance
summary. In this manner I have literally tested every bar as an entry with
the exact same exit for 500+ stocks for 10+ years.
3. I then compare the above exit code results as profit in % to profit in %
to the S&P.
I have read that a good exit will make money with any entry. I doubt that,
but I will get back to you if I find one. If someone has one they want me to
test, I will.
The remainder of your post was interesting. I am going to save it and give
it some thought.
I look forward to other position sizing ideas. Thanks Kevin and others for
the good responses to my post! I don't have a lot of time and I am spending
my sleep hours reading this list but I will do my best to respond to the
other position sizing comments.
Russ
Russ
|