|
PureBytes Links
Trading Reference Links
|
At 05:40 PM 10/24/01 +0300, you wrote:
>Hello dear friends.
>There has been a long discussion about the criteria, which a good system
>should meet ...
>... Therefore, I am kindly asking each one of
>you to define your own criteria, completing the following table with values
>and/or providing additional criteria or comments.
Thank you, Stephanos, you mention very valid criteria, many of them I would
use to analyze systems I am developing. However, I am afraid I can't
answers most of your questions with specific numbers. But I can comment
with some un-written rules I follow:
1) I reject any system that generates less than, oh, perhaps 500% annually.
I test without commissions, slippage and stops. I expect the system to
produce far less in real life. It is not difficult to come up with such a
system however the best performing systems are cyclical systems: you have
to be patient and keep paper trading them until they sync with the market.
Design system-entry and system-exit rules.
2) I don't like drawdowns: if my equity drops by more than 10% for a
particular security I go cash. I keep paper trading my systems while in
Cash, for small draw-downs my Equity reentry point is the Equity's Highest
High. For major DDs I might consider going back in when I see my Equity
trend is reconfirmed with 3-4 positive trades. This approach may keep me in
cash for significantly long periods of time and limits my losses. It
reduces exposure to risk proportional to the time I am in Cash.
3) I don't believe in long term predictions, 3-4 days is all I believe in.
So all my systems are designed on frequent trading, about 2-8 days.
4) I test stops in Excel, the stops in MS don't work for me and give me
wrong results.
5) I translate Systems, or parts of systems, that I like and that have
reached some form of maturity into VBA/Excel. The reason being that I have
had a few occassion where MS gave me exceptional results that were due to,
well lets call it, program malfunctions.
5) I want my systems to work on many securities and occasinally run the
MSBT on ALL secutities in my computer, even the indices. This groups my
securities according to their responsiveness to different systems. I
probably reject 75% of the tests based on weird equity curves or unusable
stats.
Now, remember: I am a newby with limited trading experience. Like Stephanos,
I enjoy the development more than the actual trading. Don't follow any of
my methods unless you confirm them for yourself!
Happy development :-)
Herman.
>
>Though I could hardly see any ethical objections to my request, I prefer to
>state that your contribution will not be used for commercial reasons. I am
>neither a professional analyst nor a professional trader.
>Nevertheless, I have worked a lot on technical analysis simply because I
>love to do so. In fact I like analysis much more than trading.
>
>Needless to say, we are not talking about "ideal" systems here, as such
>things simply do not exist. We are talking about realistic systems, which
>you would feel comfortable enough to trade with, systems that you would
>consider reliable enough, profitable enough, safe enough.
>Since there are obvious differences between short-term, medium-term and
>long-term systems, one may want to provide different standards for each kind
>of system, by copying the following table and providing different values for
>each kind. If one or more fields are not important enough for you, then do
>not provide values for them.
>
>In case you take the time to provide such help, please do not try to guess
>other members' opinions, or what an objectively good system would look like.
>Instead, describe your own standards below.
>
>Kind of system: ---------- (short, medium, long)
>
>Regarding system report:
>
>Annual gain should be at least ----%
>Total trades should be no less than ---- and/or no more than ----
>Profitable trades should be at least ----% of total trades
>System close drawdown should be less than ----%
>System open drawdown should be less than ----%
>Max open trade drawdown should be less than ----%
>Profit/Loss index should be at least ---
>Reward/Risk index should be at least ---
>Buy/Hold index should be at least ---
>Add more criteria below if you like.
>
>Regarding optimization (avoiding over-optimization and curve-fitting):
>
>Total optimization tests should be less than -----
>No more than --- variables should be optimized.
>Add more criteria for optimization below if you like.
>
>Regarding system testing:
>
>The above system should be back tested on more than ----- bars, or on more
>than -- years of data
>and/or
>it should have been successfully traded (with real trades) for more
>than ----- (specify the period)
>Add more criteria for system testing below if you like.
>
>Any help would be very much appreciated.
>Thank you all in advance
>
>Stephanos
>
>
>
>
>
>
>
>
>
>
>
|