|
PureBytes Links
Trading Reference Links
|
If you use Optimization on two variables you can plot the results
with Excel as a surface chart. I find surface charts helpful in visualizing
how a trading system behaves and whether or not I can trust the backtest
results to be robust and non-spurious.
It's a pretty straightforward process: Run the Exploration, Export as a
CSV, Open with Excel, create a Pivot Table and finally make a 3-D Surface
Chart.
Attached is an example I created using a sample database of 50 stocks
spanning 10.7 years diversified across industries and capitalization. The
trading system is a simple Donchian breakout: Buy at open tomorrow if
today's Close exceeds an X-day High, and Sell at open if today's Low is
lower than a Y-day Low.
The results are very promising and are likely robust:
Risk adjusted ann. return: 22.82%
Ratio avg win/avg loss: 1.99
Profit factor: 1.56
Avg. # of bars in winners: 41.9
Total number of trades: 1318
Percent profitable: 43.9%
Number winning trades: 579
That 22.8% RAR is a pretty good result, I think. The database spans bull
and bear markets and contains some poorly performing stocks [along with
some good ones].
When you look at a surface chart you want to see a fairly flat surface, all
with a positive return. High peaks and low valleys tend to indicate data
mining and are probably not reproducible. Since the attached chart is
fairly flat, I believe this system will be tradable and repeatable.
There's one more test that needs to be done, which is to statistically test
the results against random noise trades, but due to lack of time I'll save
that for later ;-)
Regards,
Jim Varney
__________________________________________________
Do You Yahoo!?
Get email alerts & NEW webcam video instant messaging with Yahoo! Messenger
http://im.yahoo.com
Attachment:
Description: ""
|